ijsint vol. 1, no. 16... · web viewjournal of science, innovation and new technologyvol. 1, no. 16...

&

Vol. 1, No. 16July, 2016

Printed ISSN: 2223-2257Online ISSN: 2225-0751

INTERNATIONAL JOURNAL OF

SCIENCE,

INNOVATION

NEW Technology

www.ijsint.org

1

INTERNATIONAL JOURNAL OFSCIENCE, INNOVATION AND NEW TECHNOLOGY

Vol. 1, No. 16, July, 2016Printed ISSN: 2223-2257 - Online ISSN: 2225-0751

Editors Co-editorsProf. Dr. Kozeta SEVRANI Dr.Edlira MARTIRIProf. Dr. Fatmir MEMAJ Dr.Edlira KALEMI

Editorial Advisory and Academic Board Agni DIKA, University of South Eastern Europe, Macedonia Anita MIRIJAMDOTTER, Linnaeus University, Sweden Arjan GJONÇA, LSE, UK Artur BAXHAKU, University of Tirana, Albania Bashkim RUSETI, University of Tirana, Albania Betim ÇIÇO, Polytecnic University of Tirana, Albania Dhimitri TOLE, University of Tirana, Albania Edmond HAJRIZI, University for Business and Technology, Kosovo Fatmir MEMAJ, President of ASET, Albania Francesco PROTA, University of Bari, Italy Carles Gispert-Pellicer, Universitat Autònoma de Barcelona, Spain Giuseppe TARDIVO, University of Torino, Italy Gudar BEQIRAJ, Academy of Sciences, Albania Heinz-Dieter WENZEL, University of Bamberg, Germany Ilia NINKA, University of Tirana, Albania Iraj HASHI, Staffordshire University, UK Kozeta SEVRANI, University of Tirana, Albania Krzysztof KOMPA, Warsaw University Of Life Sciences, Warsaw, Poland Kurt MATYA, Vienna University of Technology, Austria Larry STAPLETON, Waterford Institute of Technology, Austria Lule AHMEDI, University of Prishtina, Kosova Marcus HUDEC, University of Vienna, Austria Mehtap HISARCIKLILAR, Staffordshire University, UK Neki FRASHËRI, Research and Development Center, Albania Norbert JESSE, Dortmund University, Germany Peter KOPACEK, Vienna University of Technology, Austria Rodica PRIPOAIE, Danubius University Galaţi, Romania Silvana TRIMI, University of Nebraska, USA Sang LEE, University of Nebraska, USA Shanggeun RHEE, Kean University USA Vittorio NICOLARDI, University of Bari, Italy Zamir DIKA, University of South Eastern Europe, Macedonia

The International Journal of Science, Innovations and New Technology is published under the auspice of ASET (Albanian Socio-Economic Think Tank) in collaboration with the Department of Statistics and Applied Informatics, University of Tirana, and the Department of Computer Science, University of Durres. IJSINT is indexed in EBSCO Host databases.The views presented in the Journal present opinions of the respective authors. The views presented do not necessarily reflect the opinion of the editors, editorial and academic board or staff. All rights reserved by ASET. No part of this journal may be reproduced or used in any form or by any means without written permission from the publisher, except for noncommercial, educational use including teaching purposes.

Publisher: ASETRr. e Dibrës, Pall. 487 Shk.2, Apt.22, P.O. Box 1506, Tirana, ALBANIATel/Fax: ++355 (4) 258 171, www.aset-al.com, E-mail: [email protected]

Correspondence and questions: E. Martiri: [email protected] E. Kalemi: [email protected]

i

INTERNATIONAL JOURNAL OFSCIENCE, INNOVATION AND NEW TECHNOLOGY

Vol. 1, No. 16, July, 2016Printed ISSN: 2223-2257 - Online ISSN: 2225-0751

Table of Contents

1 A model proposal for the electric energy valorization in a PV power plant equipped with CAES systemKliton BYLYKBASHI, Roberto CAPATA, Federico TESTA 1

2. A fuzzy-based system for improving reliability in MANETMirjeta ALINCI, Vladi KOLIÇI, Evjola SPAHO 11

3. Men and Women Wage Differences in Germany and PolandDorota WITKOWSKA, Aleksandra MATUSZEWSKA-JANICA 17

4. Analysis and comparison of DES cryptographic algorithm and AES cryptographic algorithm in different CPU Florim IDRIZI, Agon MEMETI, Burhan RAHMANI 25

5. An Application of Multilevel Latent Class Analysis with Adolescent GamblersEmil FRASHERI, Prof. Assoc. Dr. Besa SHAHINI 29

6. Code and message footprints of generated code for embedded devices as components of SOAKujtim HYSENI, Neki FRASHERI 37

ii

A MODEL PROPOSAL FOR THE ELECTRIC ENERGY VALORIZATION IN A PV POWER PLANT EQUIPPED WITH CAES

SYSTEM

Kliton BYLYKBASHI1, Roberto CAPATA1, Federico TESTA1 1 University of Roma “Sapienza”, Dept. Of Mechanical and Aerospace Engineering, Rome, Italy

[email protected], [email protected], [email protected]

AbstractIn this article, an analytical method is evaluated and implemented; to assess the possible electricity sales strategies produced by a 3 MW photovoltaic power plant, connected to a 250 kW CAES (Compressed Air Energy Storage) system, with a storage capacity of 750 kWh. The presented model combines a different numbers of parameters and variables, relevant for the system optimization. Several simulations of various system configurations have been carried out, to explore and evaluate the economic and technical feasibility of the plant, specifically it has been valued tow case of study: CASE 1 the system is not incentive; CASE 2 the system is incentive. In the end of paper it has been rated the Leveled Cost of Energy (LCOE) and specified how the investment could become affordable in the foreseeable future.Keywords: CAES, Photovoltaic System, Energy Accumulation Systems, Energy Power Exchange, LCOE

1. IntroductionThe renewable energy plays an important role for a sustainable progress, but by their nature, these sources do not allow, a continuous energy production. Sets of technologies are capable to accumulate the excess of energy to give it back when requested, commonly known like accumulation systems. These technologies are useful in the new configuration of the smart grid, that providing energy from renewable micro generation increasingly closer to the final consumer. Generally there are many storage systems, with different characteristics and specifications: hereinafter the ESA graphic determines the relations between their power rate range and their discharge time, for the different storage system.

Figure 1. Caratteristiche delle varie tecnologie di batterie di accumulo (fonte ESA)

1.1 Hydroelectric pumped storagePrinciple of operation: a conventional hydroelectric power plant is used to generate a cyclical flow of water between two reservoirs at different elevations. The possible operating phases are two: the pumping phase (when the price of energy is lower) and the generation phase (when the price of energy is higher). Application areas and typical dimensions: the global pumping capacity amounts to approximately 200 GW. In addition, they represents around the 99% of the global stored capacity. The typical size occupies a range that varies from the order of MW to the order of GW. Strong points: high efficiency (70%), strengthened technology, reliability, very fast charge/discharge periods.

1

Journal of Science, Innovation and New Technology Vol. 1, No. 15 – March, 2016

Critical issues: need for placement in geo-morphologically favorable sites, relatively high investment costs.1.2 Electrochemical storage (Na-S batteries)Principle of operation: the sodium/sulfur battery belongs to the group of high temperature batteries in which the two electrodes are in the molten state, physically and electrically isolated from each other by a ceramic separator that allows the ion passage and performs the functions of electrolyte. Application areas and typical dimensions: large-scale electric network regulation (i.e. Grid Energy Storage), based on the MW scale (the figure shows a Japanese plant of approximately 30 MW); aerospace applications (e.g. Space Shuttle). Strong points: high energy density, high charge/discharge efficiency (89-92%), long life cycles, potential low cost in the application on a large scale, relatively mature technology, good environmental compatibility. Critical issues: the high temperatures do not allow the application to electric vehicles.1.3 Mechanical storage (flywheels).Principle of operation: the flywheels operate accelerating a rotor up to a very high speed and maintaining the energy in the system in the form of kinetic rotation energy. When energy is extracted from the system, for the principle of energy conservation, the flywheel speed decreases. Application areas and typical dimensions: network control service (i.e. Power Quality), the storage capacity of the order of tens of kWh, transfer power ranging between 10 and 20 kW.1.4 Conventional CAESCAES (Compressed Air Energy Storage), indicates a configuration which provides: a sequence of compressors with inter - and post - refrigeration stages (eventually aiming at reducing the compression work and maximizing the magnitude of the storage capacity). The storage can be a storage cave; a combustion chamber where the stored air is canalized and then attains the function of combustive agent of the natural gas; a turbine and a generator.Application areas and typical dimensions: the only two applications in the world have been so far realized in Germany (1978, 290 MW) and USA (1991, 110 MW). Both of the plants use saline caverns as storage tanks. Currently in the world there are several plants of this type still in design phase or construction phase. Strong points: high reliability, sufficiently mature technology, compressor and turbine operate in two different instants. Critical issues: placement is needed in particularly rare sites (such as salt caverns or porous formations).1.5 Comparisons between the different technologiesIn the following diagram, a list of several technologies of storage systems is provided (for most electrochemical systems) classified according to the specific energy capacity as a function of the specific transfer power. The oblique lines represent the charge/discharge rate.

Figure 2. Diagram Power/Energy – charge/discharge rate

2. Innovative CAES systemLightSail Energy (LightSail) is in Berkeley, California. This company has developed a compressed air energy storage technology, which may be used for grid-scale storage. The main innovation is the injection of a mist of water spray into a compressed air system, so the spray rapidly absorbs the heat energy of compression and provides the energy during expansion. The system comprises a reversible mechanism to compress and expand air, one or more compressed air storage tanks, a control system, one or more heat exchangers and in certain embodiments of the invention, a motor-generator.

2

A model proposal for the electric energy valorization in a PV power plant equipped with CAES system

Figure 3. CAES plant layout [patent No: US 8,240,142 B2]

The reversible air compressor-expander uses mechanical power to compress air (when it is acting as a compressor) and converts the energy stored in compressed air to mechanical power (when it operates as an expander). The compressor-expander comprises one or more stages, each stage consisting of pressure vessel (the “pressure cell”) partially filled with water or other liquid. In some embodiments, the pressure vessel communicates with one or more cylinder devices to exchange air and liquid with the cylinder chamber(s) thereof. Suitable valves allow air to enter and leave the pressure cell and cylinder device, if present, under electronic control. In a more detailed way, the system includes a cylinder device (21) defining a chamber (22), a piston device (23) in the chamber and a pressure cell (25). The cylinder (21) and pressure cell (25) together form a one-stage reversible pressure compression/expansion mechanism (24). Air enters the system (20) via pipe (10), passes through a filter (26) and enters the cylinder chamber (22) via pipe (30) where it is compressed by the action of the piston (23). Before compression begins, a liquid mist is introduced into the chamber (22) using an atomizing nozzle (44). The volume of mist injected into the chamber (22) is predetermined to be the volume required to absorb all the heat generated during that piston stroke. As the mist condenses, it collects as a body of liquid (49e) in the cylinder chamber (22). The compressed air/liquid mixture is then transferred into the pressure cell (25) through outlet nozzle (11) via pipe (51). That is when the critical heat exchange occurs, followed by storage of the air: in the pressure cell 25, the transferred mixture exchanges the captured heat generated by compression to a body of liquid (49f) contained in the cell. The air bubbles up through the liquid and on to the top of the pressure cell, and then proceeds to the air storage tank 32, via pipe 33.In conclusion, the LightSail’s system is more efficient because it captures and stores both the mechanical energy and the thermal energy used in compressing air. Specifically, a water mist is infused into the compression chamber as the air is compressed. Water can hold 3,300 times as much heat as the same volume of air, and as such, it is able to capture the heat generated by the process more effectively. Both potential energy in the form of pressurized air and the heated (and therefore higher-energy) water can be stored. When the captured, pressurized air is released back through the system, the heated water is re-infused into it. That heated air can return more of the energy stored by the system than can other CAES processes.2.1 Plant specificationsApplication areas: electrical energy generation systems, both opened (connected to the grid) and closed (isolated from the grid). Typical dimensions: each module has a nominal transfer power (Power Unit, P.U.) of 250 kW and a nominal storage capacity (Storage Unit, S.U.) of 750 kWh. Advantages: high efficiency in comparison with the other CAES systems (the global transfer efficiency is 70% against 25÷30 % of traditional CAES). The system is modular: there is the possibility to adapt the system to the specific demands, varying the number of S.U. and P.U.. Innovative introduction of vaporized water injection for cooling purpose during the compression. Low

3


cost of maintenance. Use of air as energy vector (with zero impact on operative costs). Critical issues: high installation costs.2.2 The Electrical Stock Market (Italian Power Exchange – IPEX)The Electrical Stock Market is a telematic marketplace where electrical energy supply meets demand; it defines the amount and the price of the traded electrical energy. It represents an essential instrument for the creation of a competitive market. It has been created with the aim of facilitating the emerge of efficient balanced prices, which allows the producers and consumers to sell and buy energy when there is a greater economic profit. Some other functions are the stimulation of competition between the operators, the market’s stabilization support, the incentive of new power plants and new electrical grids construction and the stimulus for new operators entrance in the Market. The IPEX was established on 1° April 2004 and it is now managed by the GME (“Energy Market Manager”). It is divided in two Markets: the MSD (“Market Services Dispatching”) which is divided into the MI (“Intraday market”) and the MGP (“Day-Ahead Market”), on which we will operate in this case study. The MGP is the location for the most of the electrical energy exchange transaction. There is the exchange of hourly energy stocks for the following day. The operators present their offers in which they establish the amounts of energy for sale and the minimal and maximum price at which they are willing to sell electrical energy. The session of MGP ends at 12.00 of the day before the electricity delivery. The results of MGP are communicated by 12.55 of the day before the electricity delivery. The offers are accepted after the end of the session, based on the economic subject and on the respect of the exchange limits between the zones. The MGP is an auction market and is not a continuous bargaining market. The accepted offers are referred to the PUN (“National Unique Price”), which is the medium of the prices of the geographical zones, weighted with amounts of energy purchased in those zones. The GME acts as a central counterparty.

3. Case studyThe aim of this project is to realize a calculation code that operates as a simulator of sales strategies, to be applied to a PV system connected to the power grid and to the L.S.E. CAES plant. The purpose is to evaluate the economic convenience of the application of this system that, thanks to its innovative modularity, can be adapted to various power levels. In particular, it will be evaluated the application to a 3 MW photovoltaic system.3.1 System hypothesisHP 1. There is no possibility to sell at the same time the energy produced by the PV plant and the energy stored (CAES tank). The action number 1 of the flowchart guarantees the observance of this rule: the branches related to the respective sales have been separated.HP 2. There is a price threshold at which the electricity sale is more advantageous than the electricity storage. This price threshold is called δ [€/MWh] and amounts to 57.14 €.HP 3. This value has been calculated through the equation:

(1) MGP :δ=ƞ :100 %Where:

MGP is the Minimal Guaranteed Price: it’s the marginal sale’s price of electrical energy produced by renewable sources. This is an incentive condition for these types of technology. If the energy is sold to the grid in a moment in which the PUN is lower than 40 €/MWh, there is the guarantee to sell it anyway at this price.

Ƞ is the global efficiency of the CAES plant (70%).

HP 4. The plant has a loss factor of 1.15. For the absorption of 250 kWh the CAES plant needs 250 kWh·1.15 = 287.5 kWh for each hour, so that after 3 hours it has absorbed 862.5 kWh accumulating only 750 kWh of electrical energy, losing 112.5 kWh. Similarly, during the emptying phase the plant sell to the grid 250 kWh of electrical energy. In this case the plant has to lose 250 kWh·0.15 = 37.5 kWh for each hour.

3.3 Calculation codeThe purpose is to realize an iterative algorithm, able to automatically decide if it’s more economically advantageous the sale of the energy produced by the PV plant or the storage of that energy for selling it in a second moment with a better price. During the structuring of the calculation code it was

4


necessary to consider several parameters, from whose interaction is possible to obtain a simulation of the energy and economic operation of the system. In the following paragraph these parameters will be introduced and classified according to their nature and, for each of them, it will be given a short description.Assessment’s parameters:PUN [€]: it’s the average national price of the electrical energy; it has been obtained from the website of GME (Energy Market Manager).SPREAD of the day before [€]: it’s the difference between the value of the PUN related to the day before and the value of the MGP (Marginal Guaranteed Price). This parameter is an index of the profitability of the choice of selling energy. The arrows on the left of each value have the purpose of underline this profitability evaluation. Reclaiming the parameter δ , the following scheme is obtained:

: SPREAD > δ: the sale is very profitable. : 0 < SPREAD < δ: the sale is moderately profitable. : SPREAD < 0: the sale isn’t profitable (the price will be anyway 40 €/MWh).

Table 1. Assessment parameters

G: Solar hourly radiation (2005) [Wh/m2]: it has been obtained through an estimate given by the website “Solar Radiation Data (SoDa) – Solar Energy Services For Professionals”.EPV: Hourly energy production of the PV plant [kWh]: it’s been calculated with the following equation (in which STOT , PV is the total surface of the PV plant, ƞPV is the PV plant efficiency (14%):(2) EPV=G∗STOT , PV∗ƞPV

System’s variables:It’s possible to modify these three variables. Consequently the calculation code will produce different results.

PP,PV: Peak PV Power [MW]: it’s the design nominal power of the PV plant.Number of Storage Units: number of tanks of the CAES plant.Number of Power Units: number of units used for the energy transfer.

Other bound variables:CS.U.: Nominal capacity of a Storage Unit [MWh]PS.U.: Nominal Power of a Power Unit [kW]CCAES: Total capacity of the CAES plant [MWh]PCAES: Total transfer power of the CAES plant [kW]ηG,CAES: CAES plant efficiency [%]ECAES: Stored energy [MWh]: it’s the amount of energy stored as compressed air in the CAES tank.Strategies:

5


SELL ENERGY PRODUCED BY PV PLANT: The energy instantly produced will be directly sold to the grid.STORE : The energy instantly produced will be temporarily stored and released in a second moment.SELL ENERGY STORED IN CAES: The energy earlier stored in the tank will be released and sold to the grid.

Figure 4. Flowchart of the calculation code

Table 2. Code operationsNUMBER

OFCHOICE

TITLE MEANING FORMULA NOTES

1 PV PRODUCTIONThe PV plant is now producing electrical

energy?

EPV >0 EPV is the energy produced by PV plant.

2 PROFITABILITY OF SALE

The sale of electrical energy produced by PV plant is profitable at this moment or not?

SPREAD>δThe PV plant is now producing

electrical energy [1]: we must decide what we should do with that electrical

energy, if selling it or not.

3 FULL TANK The CAES tank is totally full?

ECAES=CCAES

The energy sale is not profitable: so, we should insert in the tank the energy

instantly produced [5]; but if the tank is full we must sell it anyway.

4 CALCULATION OF THE NEW INCOME

The new income comes from the sale

of the electricity produced by PV

plant.

MAX {MGP , PUN }·( EPV −(1 .15 ·∆ ECAES ))The amount (1 .15 ·∆ ECAES ) is

the electrical energy hourly stored (HP 4); the term

MAX { MGP, PUN }is in consideration of HP 2 and HP 3.

5 TANK REPLENISHMENT

The tank is replenished. ECAES i+1

=ECAESi+∫

0

t

PCAES dtThe amount ∫0

t

PCAES dt in one hour

is equal to 250 kWh (HP 4).The subscripts i and i + 1 refer to the

6


instants 0 and t respectively.When the tank is full the energy is sold

to the grid.

6 EMPTY TANK The tank is empty? ECAES<∫0

t

1 . 15 · PCAES dt

I this case we analyse the tank functioning.

Considering the HP 1 the energy sale from PV and CAES cannot be done

simultaneously.If there is sufficiently energy in the tank, the sale (and the emptying [8])

can take place.

7 PROFITABILITY OF SALE

The sale of electrical energy earlier stored

in the tank is profitable at this moment or not?

SPREAD>δ The parameter δ has been described in the HP 2 and HP 3.

8 TANK EMPTYING The tank is disburdened.

ECAES i+1=ECAESi

−∫0

t

1 .15 · PCAES dtThe term ∫

0

t

1 . 15· PCAES dt is the

amount of energy that we can get from the tank in the period t. If the period is

one hour we can get 287.5 kWh.

9 CALCULATION OF THE NEW INCOME

The new income comes from the sale

of the electricity earlier stored in the

tank.

MAX {MGP , PUN }·∫0

t

PCAES dtThe revenue from CAES energy sale is

updating.This iterative algorithm evaluates the

economic profit that we can obtain whit the combination PV + CAES plants, in

the given specifications.

4. Economical view of project. The economic feasibility of the project has been studied, performing the analysis of two specific cases: in the first case no market incentive is considered, in the second one market incentive is considered. For both cases, some simulations of different discount rates for evaluating the different NVP and IRR are shown. Finally, the Leveled Cost of Energy is evaluated, and, is defined when the PV + CAES technology become economically solid.4.1 Generality of the simulation: hypothesisHereinafter the fixed and variables parameters of the case of study are defined:Fixed parameters:

PV and CAES System; Power of PV plants: 3,000 kW; Power of CAES System: 250 kW; Capacity of CAES System: 750 kWh; Year energy production by the Plants: 4,485,120 kWh; CAPEX 3,000,000 €; OPEX per year 17,940 €; Inflation rate 3%; Life plant: 21 year; Implementation time of the system: 2 years.

Variables Parameters Energy Price:

o 0.06 €/kWh no incentive market;o 0.313 €/kWh incentive market.

4.1.1 Case 1This case represents the current situation of the energy market in Italy, where the IPEx establishes the price to sell the energy, that now is 0.06 €/kWh. In the following table, is evaluating the trends of Net Present Value for different discount rates (15%, 12%, 10%, 8%, 6%, 4%) and Internal Rate of Return (IRR%).

7


Table 3. Economic analyses of the investment: CASE 1

The dr is the Real Discount Rate, in which is considering the inflation; the equation is the following:

(3) dr=[ 1+dn

1+r−1]

So there are the summary table of dn and dr:Table 4. Nominal and Real Discount Rate

For all cases considered the investment is inconvenient, because the IRR% is low 8.73 % in case without inflation and 3.46% in the worse case, considering a 3.00 % of inflation and 41.3 % of taxes, as well as in the Italian marketplace. Therefore, the photovoltaic and CAES system considering, will be not-convenient.

4.1.2 Case 2The Case 2 considering the energy selling price of 0.313 €/kWh (this is a case with government incentives like was in Italy some years ago); the business becomes highly affordable, the NVP and IRR are very convenience.

Table 5. Economic analyses of the investment: CASE 2

The investment is economic, in the worst case where the market is inflated and taxed, the IRR is 27.96 % and the NPV 15% is 3,675,449.36 € after 21 years from the investment, more then 23% of initial capital. In the best case the IRR is 42.08% and the NVP 4% is 26,305,669.72 € more than 777% of initial capital.

4.2 Conclusions: LCOE: Levelled Cost Of Energy (Electricity)The two cases of study showed two opposite market situation, really distance from a reasonable investment. The question is: what is the profitable right price to sell the energy for this plant? The Levelled Cost Of Energy or Levelled Energy Cost (LEC) answers this question. LCOE is a convenient summary measure of the overall competiveness of different generating technologies. It represents the per-kilowatt-hour cost of building and operating a generating plant over an assumed financial life and duty cycle. Components for the calculation of LCOE are capital costs, fuel costs, fixed and variable operations and maintenance (O&M) costs, financing costs, and an assumed utilization rate for each plant type.

8


(4) LCOE=∑t=1

n C t+Ot+F t+Ot

(1+r )t

∑t=1

n E t

(1+r )t

LCOE: Levelized Cost Of Energy Ct: capital cost Ot: Operation (fixed and variable) and maintenance cost Ft: Fuel cost Ot: Other cost Et: Energy produced n: life of plants r: rate of discount

It will consider hereinafter as is varying the LCOE with the changing of plant’s life, and how this system could be economic. The nominal discount rates for this model are 15%, 12%, 10 %, 8%, 6%,4%; the energy produced by the plant every year is 4,485,120kWh, the Capex is 3,000,000 €, Opex is 31,396 € every year. There aren’t another costs of plant for this model, and there aren’t revamping in the all life of this plant (the revamping is including in operation year cost). It is observed that the prices vary from a maximum of 0.40 € / kWh to a minimum of 0.06 € / kWh, where the NPV is 4% and the stakeholder admits payback time in 21 years from the investment. But no one good economical investor wants to do an investment to have a return at the end of the plant’s life (in the better case) and no one clever state policy invests on loss industry, therefore the considering PV plants and CAES system are not momentarily affordable. The PV plants is reasonable just for the power micro generation for users that pay the energy above 0.20 €/kWh. This is the case of Italy, where the GSE (Italian National Grid Operator) recognizes the SEU (Users Efficient Systems) for operators of PV energetic establishment. In this case the price of electricity could be more than 0.20 €/kWh. If is looking the equation (4), the price of energy will be lower in two way: or decreasing the numerator or increasing the denominator.

Figure 5. LCOE variation

In the first case it will be possible that the technology will become less expensive (Opex, Capex and other costs will be cheaper). The denominator is composed by the energy producing that is:

(5) E=∫t 1

tf

Pdt=P ∆ t=P (t f −t1 )

9


Where tf=t hours of plant’s operation and t1=0. The power of photovoltaic system depends on the yield factor η, the irradiance I0 (W/m2), the surface S (m2), the angle of inclination of the module with respect to incident solar radiation ¿:

(6) P=η∙ I 0 ∙ S ∙ sin α (7) E=η ∙ I 0 ∙ S ∙sin α ∙t

In the present state of technology, to increase the energy, it should increase the performance of the plant. The yield Factor of the actual plant is 19% ; so the equation is:

(8) E=η ∙K=¿4.485 .000=0,19∗K=¿ K=23.605 .263 kWh(9) K=I 0 ∙ S ∙ sin α ∙ t

K is the potential of the energy that the plant can produce in one year if the η could be 100%. Is provided below the equation (4) introducing the equation (8) and (9), is looking:

(10) LCOE=∑t=1

n C t+Ot+F t+Ot

(1+r )t

∑t=1

n η∙ K(1+r )t

Therefore, if the efficiency of the plant will be increasing, the LCOE will be decreasing, making economically the photovoltaic system. In the same time, the CAES system could be a good business if it will be combined with PV system.

Reference [1] A. N. Negri; The electrical energy storage; RSEview, prima edizione, 2011.[2] P. Savoldelli; Investigations on technologies and materials from a thermal perspective in CAES plants;

RSE, 2013.[3] M. Benini; The electric energy storage: economic point of view; RSE, 2011.[4] Fong, A. Danielle, E. S. Crane; Compressed Air Energy Storage system utilizing two-phase flow to facilitate

heat exchange; US 8,240,142 B2, Lightsail Energy Inc., Berkeley, CA (US), 2012.[5] M. Cresta, F. M. Gatta, A. Geri, M. Maccioni, A. Mantineo, and M. Paulucci; "Optimal operation of a low

voltage public network with renewable DG by storage systems and demand response: a case study in a trial site," in 3rd Renewable Power Generation Conference (RPG 2014), Paper n°0192.

[6] D. Favrat; Advanced Concepts in the design of CAES and TEES systems, EPFL, 2013[7] G. Buzzi Ferraris, F. Manenti; Fundamentals and Linear Algebra for the Chemical Engineer: solving

numericals problems; Wiley Vch Verlag Gmbh[8] GSE; Technical standards for implementation of the regulations related to the integration of the storage

systems in the national electrical system; 08/04/2015[9] J.Simmons, A. Barnhart, S. Reynolds; Study of Compressed Air Energy Storage with grid and photovoltaic

energy generation, Arizona Research Institute for Solar Energy[10] L. Gori; Numerical computation; V edition 2006[11] GSE; Guida blu for the photovoltaic energy: CEI standards, 9/2015[12] ABB; Technical application papers nr 10: Photovoltaic plants[13] R. Westerfield Jordan; Fundamentals of Corporate finance, tenth edition[14] E. R. Urban, L. Gugliermetti; The economical cycle of dreding, April 2016[15] Exxon Mobil, 2015: The Outlook for Energy: A View to 2040 [16] Chicago University.2008: Levelized Cost of Energy. [17] Brealey, R. A. 2011: Priciples of Corporate Finance, MGH

Online reference [1] http://www.greentechmedia.com/articles/read/LightSail-Gets-37.5M-from-Thiel-Khosla-and-Gates-for-

Compressed-Air-Grid, (10.06.15)[2] http://www.soda-is.com/eng/index.html, (14.06.15)[3] http://www.autorita.energia.it/allegati/docs/13/486-13.pdf, (20.06.15)[4] http://www.gse.it/it/salastampa/news/Documents/Regole%20Tecniche%20Sistemi%20Accumulo.pdf,

(26.06.15)[5] http://www.mercatoelettrico.org/It/download/DatiStorici.aspx, (26.06.15)

10


[6] http://www.mercatoelettrico.org/It/Mercati/MercatoElettrico/MPE.aspx, (28.06.15)[7] https://www.mercatoelettrico.org/It/Mercati/CV/CosaSonoCv.aspx, (02.07.15)

11

A FUZZY-BASED SYSTEM FOR IMPROVING RELIABILITY IN MANET

Mirjeta ALINCI, Vladi KOLIÇI, Evjola SPAHO1

Department of Electronics and TelecommunicationPolytechnic University of Tirana

Mother Teresa Square, No. 1, Tirana, AlbaniaE-mail: [email protected], [email protected], [email protected],

AbstractMobile ad hoc networks (MANET) can be defined as a collection of large number of mobile nodes that form temporary network without aid of any existing network infrastructure or central access points, each node participating in the network acts both as host and router and must therefore is willing to forward to packets for other nodes. The characteristics of MANETs such as: dynamic topology, node mobility, provides large number of degree of freedom and self-organizing capability of that make it completely different from other network. For efficient routing and monitoring, the network is divided into manageable entities called clusters. Each cluster has a particular node elected as cluster head (CH) based on a specific metric or a combination of metrics such as identity, degree, mobility, weight, density, etc. Due to the nature of MANETs, the design and developed secure routing is a challenging task for researcher in an open and distributed communication environment. To define and improve the reliability of cluster nodes in MANET we present a Fuzzy based approach. We consider security parameter as reliability. We found that by selecting nodes with high security rate, the nodes are more secure and the system performance is improved.Key words: cluster head, MANET, secure, performance

I IntroductionMobile ad-hoc wireless network (MANET) is a wireless communication network in which the nodes not within transmission range establish communication with other nodes help to forward data. It operates without fixed infrastructure, support user mobility coming under the general scope of multi hop wireless networking. In such networks, nodes move arbitrarily without prediction. Each node acts as a host and locates network routes to aid the network form a complete communication route. Thus MANETs are dynamically establishing wireless networks, maintaining routes through network, forwarding packets to others to ensure multi-hop intra node communication not in transmission range. In MANET, secure communication is more challenging task due to its fundamental characteristics like infrastructure less, wireless link, distributed cooperation, dynamic topology, lack of association, resource constrained and physical vulnerability of node [1], [2].For efficient routing and monitoring, the network is divided into manageable entities called clusters. Each cluster has a particular node elected as cluster head (CH) based on a specific metric or a combination of metrics such as identity, degree, mobility, weight, density, etc. The cluster head plays the role of coordinator within its substructure. Each CH acts as a temporary base station within its cluster and communicates with other CHs [3], [4].Structuring a network is an important step to simplify the routing operation in MANETs. The clustering consists of dividing the network into a set of nodes that are geographically close. It is an efficient solution to simplify and optimize the network functions. In particular, it allows the routing protocol to operate more efficiently by reducing the control traffic in the network and simplifying the data routing. Several algorithms based on clustering techniques have been proposed in the literature. To define and improve the reliability of cluster nodes in MANET can be used several methods such as logic, probability methods for uncertain, optimization, neural networks, fuzzy logic, etc.We present a Fuzzy based approach for improving the security of cluster nodes in MANETs.The rest of the paper is organized as follow: In Section II, we present Fuzzy logic. In Section III, we present our proposed system. We show simulation results in section IV. In section V we give the conclusions.

12

mailto:[email protected]

A fuzzy-based system for improving reliability in MANET

II Fuzzy logic Fuzzy logic idea is similar to the human being’s feeling and inference process. Unlike classical control strategy, which is a point-to-point control, fuzzy logic control is a range-to-point or range-to-range control. The output of a fuzzycontroller is derived from fuzzifications of both inputs and outputs using the associated membership functions. A crisp input will be converted to the different members of the associated membership functions based on its value. From this point of view, the output of a fuzzy logic controller is based on its memberships of the different membership functions, which can be considered as a range of inputs [5], [6].To implement fuzzy logic technique to a real application requires the followingthree steps:1. Fuzzification – convert classical data or crisp data into fuzzy data orMembership Functions (MFs)2. Fuzzy Inference Process – combine membership functions with thecontrol rules to derive the fuzzy output3. Defuzzification – use different methods to calculate each associatedoutput and put them into a table:

A. FuzzificationThe fuzzy set is a powerful tool and allows us to represent objects or members in a vague or ambiguous way. The fuzzy set also provides a way that is similar to a human being’s concepts and thought process. However, just the fuzzy set itself cannot lead to any useful and practical products until the fuzzy inference process is applied. To implement fuzzy inference to a real product or to solve an actual problem, three consecutive steps are needed, which are: Fuzzification, fuzzy inference and defuzzification. Fuzzification is the first step to apply a fuzzy inference system. Most variables existing in the real world are crisp or classical variables. One needs to convert those crisp variables (both input and output) to fuzzy variables, and then applyfuzzy inference to process those data to obtain the desired output. Finally, in most cases, those fuzzy outputs need to be converted back to crisp variables to complete the desired control objectives [7]. Generally, fuzzification involves two processes: derive the membership functions for input and output variables and represent them with linguistic variables. This process is equivalent to converting or mapping classical set to fuzzy set to varying degrees [14].In practice, membership functions can have multiple different types, such as the triangular waveform, trapezoidal waveform, Gaussian waveform, bell-shaped waveform.

B. Linguistic VariablesA concept that plays a central role in the application ofFL is that of a linguistic variable. The linguistic variables may be viewed as a form of data compression. One linguistic variable may represent many numerical variables. It is suggestive to refer to this form of data compression as granulation. [7], [8]. The same effect can be achieved by conventional quantization, but in the case of quantization, the values are intervals, whereas in the case of granulation the values are overlapping fuzzy sets. The advantages of granulation over quantization are as follows:• it is more general;• it mimics the way in which humans interpret linguistic values;The transition from one linguistic value to a contiguous linguistic value is gradual rather than abrupt, resulting in continuity and robustness.

C. Fuzzy control rulesFuzzy control rule can be considered as the knowledge of an expert in any related field of application. The fuzzy rule is represented by a sequence of the form IF-THEN, leading to algorithms describing what action or output should be taken in terms of the currently observed information, which includes both input and feedback if a closed-loop control system is applied. The law to design or build a set of fuzzy rules is based on a human being’s knowledge or experience, which is dependent on each different actual application [8], [9].

13


A fuzzy IF-THEN rule associates a condition described using linguistic variables and fuzzy sets to an output or a conclusion. The IF part is mainly used to capture knowledge by using the elastic conditions, and the THEN part can be utilized to give the conclusion or output in linguistic variable form. This IF-THEN rule is widely used by the fuzzy inference system to compute the degree to which the input data matches the condition of a rule.

D. DefuzzificationThe conclusion or control output derived from the combination of input, output membership functions and fuzzy rules is still a vague or fuzzy element, and this process in called fuzzy inference. To make that conclusion or fuzzy output available to real applications, a defuzzification process is needed. [10], [11]. The defuzzification process is meant to convert the fuzzy output back to the crisp or classical output to the control objective. The fuzzy conclusion or output is still a linguistic variable, and this linguistic variable needs to be converted to the crisp variable via the defuzzification process [13], [15].These defuzzification techniques are commonly used: Tsukamoto’s Defuzzification Method; The Center of Area (COA) Method;The Mean of Maximum (MOM) Method; Defuzzification when Output of Rules are Function of Their Inputs.

III Proposed systemTo define and improve the reliability of cluster nodes in MANET can be used several methods such as logic, probability methods for uncertain, optimization, neural networks, fuzzy logic, etc..We present a Fuzzy based approach for improving the security of cluster nodes in MANETs.Determining how the parameters affect the calculations of outputs depends on the application and is the subject of reasoning system builder.For this reason adaptive algorithm Fuzzy Logic Controller (FLC) is a good method for this system [12]. It describes the process control algorithm as a link between information about Fuzzy process conditions to be controlled and the process output.The proposed system is shown in Figure 1.

Figure1: Proposed system

We consider four parameters as inputs: Number of Nodes in a Cluster (NNC), Spent Power (SP), Packet Loss (PL) and Security (SC) to decide the reliability output parameter. These four parameters are not correlated with each other, for this reason we use fuzzy system. The membership functions for our system are shown in Figure 2. The term sets of NNC, SP, PL and SC are defined respectively as:NNC = {Few, Middle, Many}={Fe, Mi, Ma};SP = {Low, Middle, High}={L, M, H};PL = {Few, Middle, High}={Few, Mid, Hi};SC = {Low, Middle, High}={}and the term set for the outputis defined as:

14


R / L=¿ ( Remain ¿ ) (Weak Re main¿ ) ( Not Re mainNotLeave ¿ ) (WeakLeave ¿ ) ¿¿

¿¿

IV Experimental results

In this section are presented the results obtained as a result of multiple simulations. In these simulations we consider the NNC and SC asa constant parameters. In Figure 3 we consider the NNC value 0.5 units and SC value 0.1, in figure 4 consider NNC=0.5 and SC=0.5 units, and in figure 5 consider NNC=0.5, SC=0.9.With increasing of SC, the R/L is increased. When the PL increases, the R/L is decreased. From the figure 3 we see that, when NNC is 0.5 and SC is 0.1, with the increasing of PL the R/L decreased. From the figure 4 we see that, when NNC is the same and SC increased, the R/L also increased. From the figure 5 we see that, when the SC is higher, the R/L is increased much more. We consider as reliability parameter the security. We conclude that with the increasing of security parameter the performance of the system is improved and the clusters are more reliable.

(a) Number of Nodes in a Cluster (b) Security

(c) Packet Loss (d) Spent Power

15


(e) Remain/Leave

Figure 2: Membership functions of our system

Figure 3.Simulation results for NNC= 0.5 and SC=0.1 Figure 4.Simulation results for NNC= 0.5 and SC=0.5

Figure 5.Simulation results for NNC= 0.5 and SC=0.9

V ConclusionsStructuring a network is an important step to simplify the routing operation in MANETs. The clustering consists of dividing the network into a set of nodes that are geographically close. It is an efficient solution to simplify and optimize the network functions. In particular, it allows the routing protocol to operate more efficiently by reducing the control traffic in the network and simplifying the

16


data routing. To define and improve the reliability of cluster nodes in MANET we present a Fuzzy based approach. From the simulation results, we conclude that with increasing of SC from 0.1 to 0.9 and NNC =0.5, the reliability is increased. When SC is higher, the output R/L is increased much more.In the future work, we will consider more different values of parameters to improve the security of the proposed system.

Literature[1] R. Agarwal, R. Gupta, and M. Motwani, “Review of Weighted Clustering Algorithms for Mobile Ad Hoc Networks,” Computer Science & Telecommunications, vol. 33, no. 1, pp. 71–78, 2012.[2] H. S. Chauhan, “Comparative Study of Clustering Based Algorithm in Mobile Ad-Hoc Network,” International Journal of Recent Technology and Engineering (IJRTE), vol. 2, no. 3,pp. 110–113, 2013.[3] A. Savyanavar and M. Borate, “Survey of Clustering Schemes in Mobile Ad hoc Networks,” International Journal of Science and Research (IJSR), vol. 3, no. 11, pp. 2407–2410, 2012.[4] M. Saxena and K. Mathai, “Analysis of Clustering Algorithms for Creation Of Energy Efficient Mobile Ad Hoc Network,” International Journal of Advance Foundation and Researchin Computer (IJAFRC), vol. 1, no. 11, pp. 17–25, 2014.[5] H.-J. Zimmermann, “Fuzzy Set Theory and Its Applications,” Springer Science & Business Media, 1991.[6] F. M. McNeill and E. Thro, “Fuzzy Logic: A Practical Approach,” Academic Press, 1994.[7] L.A.Zadeh, J.Kacprzyk, “Fuzzy Logic For The Management of Uncertainty",John Wiley & Sons, Inc., 1992.[8] T.Munakata, and Y.Jani, “Fuzzy Systems : An Overview", Commun. of ACM,Vol. 37, No. 3, pp. 69-76, March 1994.[9] B. Philo Chaythanya, M. M. Ramya, “Fuzzy Logic Based Approach for Dynamic Routing in Manet”, International Journal of Engineering Research & Technology (IJERT), ISSN: 2278-0181, Vol. 3 Issue 6, June – 2014[10] S. N. Sivanandam, S. Sumathi and S. N. Deepa, “Introduction to Fuzzy Logic using MATLAB”, © Springer-Verlag Berlin Heidelberg 2007[11] Ying Bai, Dali Wang“Fundamentals of Fuzzy Logic Control – Fuzzy Sets, Fuzzy Rules and Defuzzifications”[12] Ren´eJager, Fuzzy Logicin Control[13] VaibhavGodbole, “Performance Analysis of Clustering Protocol Using Fuzzy Logic for Wireless Sensor Network”, IAES International Journal of Artificial Intelligence (IJ-AI), Vol. 1, No. 3, September 2012, pp. 103-111,ISSN: 2252-8938[14] Vikas Thakur, LaxmiShrivastava, SaritaS.Bhadauria, “Performance Evaluation of AODV Using Fuzzy Logic to Reduce Congestion in MANET”, Asia-pacific Journal of Multimedia Services Convergent with Art, Humanities, and Sociology Vol.5, No.5, October (2015), pp. 1-9[15] Sujata V. Mallapur, Siddarama R Patil, ”Fuzzy Logic Based Trusted Candidate Selection for Stable Multipath Routing” I.J. Information Technology and Computer Science, 2015, 06, 12-21 Published Online May 2015 in MECS

17

MEN AND WOMEN WAGE DIFFERENCES IN GERMANY AND POLAND

Dorota WITKOWSKA1, Aleksandra MATUSZEWSKA-JANICA2

1Full professor Ph.D., Faculty of Management, Department of Finance and Strategic Management, University of Lodz, Lodz, Poland

2Ph. D., Faculty of Applied Informatics and Mathematics. Department of Econometrics and Statistics, Warsaw University of Life Sciences, Warsaw, Poland

E-mail: [email protected]. [email protected]

AbstractWomen are almost half of the workforce but Eurostat estimated that in 2014 they earn 16.1% less than men. There are many factors influencing wages which are connected with individual characteristics of employees and employers, together with the level of the economic development. The aim of our research is to determine factors influencing man and women wages in Germany and Poland applying econometric models estimated on the basis of microdata from Structure of Earning Survey. We also analyze if wage determinants are the same in both countries and for both genders. Keywords: labor market, wages, gender pay gap, econometric model

IntroductionWomen are almost half of the workforce and they receive good education on all levels. However Eurostat estimated that in 2014 the difference between average gross hourly earnings of male and female employees as percentage of male gross earnings was 16.1% on average for 28 members of the European Union1. Obviously the situation on labor market is not the same in different states, and it depends on many factors including history and tradition, level of economic development and legislation among others (see Table 1).

Table 1. Comparison of women employment rate and gender pay gap [in %]Women employment

rate Gender pay gap

year 2006 2011 2012 2006 2011 2012EU 27 61.2 62.3 62.5 17.7 16.2 16.4Poland 53.1 57.2 57.5 7.4 5.5 6.4Germany 65.0 72.4 72.2 22.7 22.2 22.4

Source: own elaboration based on Eurostat data.

Changes of economic and political systems in so-called Soviet bloc, which started in Poland in 1989, have influenced not only domestic condition in transformed states but also international situation, to mention breakup of Yugoslavia, Czechoslovakia and the Soviet Union. The sudden exposure of former centrally planned economies to competition from developed countries together with a breakdown of traditional export markets, destroyed national economies in all Central and Eastern European states. The different situation was observed in German Democratic Republic (GDR) which became a part of united Germany and followed completely different way of transformation than other post-communist states.After more than 25 years of transformation the former socialist countries, the situation in Europe completely changed. There are 28 European Union member states, with 11 countries being in transition. Here the question arises if level of wages is created in similar way in developed and transitioned states. Therefore, the aim of our research0 is to diagnose factors that influence the level of remuneration both men and women in Poland and Germany i.e. two the biggest European countries, representing transitional and developed economies. Another important reason to select these two states is to investigate if labor market in Eastern lands (- former GDR) is more similar to Western part of Germany or to Poland. It is also worth mentioning that one of the biggest gender pay gap is observed in Germany and one of the smallest - in Poland. The current study contains the results of the econometric models estimation based on the evidence from Survey of Earnings Structure.

1 Gender pay gap statistics [online], Eurostat, http://ec.europa.eu/eurostat/statistics-explained/index.php/Gender_pay_gap_statistics [30.06.2016] 0 Research is conducted in frame of project: Changes of women’s position in the labour market. Analysis of the situation in Poland and in the selected European Union States in the years 2002-2014, financed by National Science Centre grant No. 2015/17/B/HS4/00930, Poland.

18



Men and Women Wage Differences in Germany and Poland

Literature reviewThere are numerous research concerning the wage determinants. The problems concerning gender influence on wages are discussed, in the papers: Grajek (2003), Adamchik and Bedi (2003), Blau and Kahn (2000 and 2006), Newell and Socha (2005), Hirsch (2010), Grimshaw and Rubery (2007), Blau et al. (2012), Słoczyński (2013) and many others.

There are many factors that affect wage, which can be divided into three groups. Individual characteristics of employee, for instance: age, gender, length of service (job seniority), the

type and level of education, occupation, working profile (full or part-time job), job contract, family economic status, preferences, type of employment;

Enterprise characteristics such as: type of industry, public or private sector, size of the enterprise, activity of the trade unions, etc.;

Characteristics of the environment, for example: economic situation in the region or/and the country, the structure of the labour market, family policy e.g. external solutions for the care of children and elder people, the legal solutions and the activities of various institutions against discrimination in the labour market, cultural background and stereotypes.

The differences in employment of men and women and also gender pay gap could be explained on the basis of a number of theories which can be classified as follows (see Vöörmann 2009): theories focusing on the characteristics of the employee (neoclassical theory, human capital theory); theories focusing on job characteristics (institutional theory, theory of market segmentation); non-economic theories (e.g. devaluation theory, discrimination theory, preference theory).

Comparative analysis of men and women remuneration can be carried out in many ways. The most often it is provided applying: analysis of the wage distributions (see e.g. Kot [ed.] 1999, methodological comments are presented in the work Domański [ed.] 2005) and analysis of the level of the wage gap measured by the gender wage gap ratio (see e.g. Oaxaca and Ransom 1994, Blau and Kahn 2000, Anderson et al. 2001, Blau 2012, Witkowska 2013). The most popular and very useful method is decomposition of the wage inequality proposed by Oaxaca (1973) and Blinder (1973) (see e.g. Blau 2012, Śliwicki and Ryczkowski 2014, Goraus and Tyrowicz 2014). Other methods that can be used in gender wage gap analysis are: matching (see Nõpo 2008, Goraus and Tyrowicz 2014), spatial econometrics tools and classification and clustering methods (see Fernández-Avilés et al. 2010), classification trees (see e.g. Matuszewska-Janica and Witkowska 2013).It should be mentioned that the gender inequality (also in wages) has both social and economic dimension (see e.g. Klasen 1993, Klasen 1999, Seguino 2000, Blecker and Seguino 2002, Morrison et al. 2007, Löfström 2009). Therefore, governments and various organizations have been still undertaking many actions that has been focused on fair treatment of men and women. Research concerning gender equality in the labour market was initiated at the beginning of the second half of the 20th century. Then in Europe the Treaty establishing the European Economic Community (TEEC) (called Treaty of Rome) was signed on 25.03.1957, and in the United States president John F. Kennedy signed Equal Pay Act (EPA) on 10.06.1963. According to the US Bureau of Labor Statistics, the US gender wage gap had been decreased from 38% to 20% in 2004, since the adoption of the EPA. Gender pay gap (GPG) in European Union was estimated at the level 16.4% in 2012. At present, policy of equality is fully reflected in the formulated strategies (see e.g. Horizon 2020, Strategy for equality between women and men 2010-2015).The regional differences have a great impact on the structure of wages. In Europe there is a particularly clear division into Eastern and Western Europe since incomes influencing the level of life are essentially different in both parts of Europe. In 2014 the first 18 positions in the ranking of European states according to GDP per capita (PPP) were hold by countries located to the West of Poland and Scandinavian states. Among “old” European Union states only Portugal and Greece were ranked on the 25-th and the 26-th positions respectively 0. Discussing the results obtained for Western Europe states, part of the literature shows that wage differentials are mainly explained by the female segregation into low-wage jobs (Daly et al. 2006), but it has been also documented the existence of significant inter-industry wage differentials0 in all countries for both sexes (Gannon et al. 2007). There are also studies that support the idea that gender pay gaps are typically bigger at the top of the wage distribution and that the gender pay gap differs significantly across the public and private sector wage distribution of each country (Arulampalam et al. 2004).In the communist countries in Eastern Europe and in the former Soviet Union (before transformation of economic and political systems) the gender equality in the labor market was committed, at least nominally (Brainerd 2000). Government policies such as relatively high minimum wages and generous maternity leave and day care benefits encouraged women to work, and female labor force participation rates were very high 0 List of European countries by GDP per capita [online], Statistics Times, http://statisticstimes.com/economy/european-countries-by-gdp-per-capita.php [30.06.2016]0 Plasman et al. (2006); Gannon et al. (2007), Du Caju et al. (2011) document inter-industry wage differences.

19


compared with those of other countries. Although women remained over-represented in such areas as health and education, they fared at least as well as their counterparts in most developed and developing countries in terms of female-male wage differentials. However, in terms of occupations and industry branch, women and men were segregated in similar way as in the West. It is worth reminding, that in the centrally planned economy wages were assigned according to occupational wage scale within each industry. The enterprises operating under no competitive pressure were left with little impact on wage rates and wage differentials. There was no unemployment in the sense of joblessness, however efficiency of work was very low and many job positions were completely useless. Women were accorded a wide range of rights and privileges at work, such as: fully paid maternity leave, legal protection from overly physical and dangerous work during pregnancy, nursery schools and health care facilities that were located in larger enterprises. It is worth mentioning that in the majority of centrally planned economies, women’s labor market participation was higher than in Western states. The main reason of that fact was low earning of single employee which was not enough to maintain a basic living standard thus both adults in a nuclear family had to work. Therefore, relatively few women held senior positions since women undertook a very large share of domestic duties thus incurring a double burden and leaving them less time to pursue a career than men. Also, the revolution in gender relations in the West, which has brought about a slow but fundamental shift in the household division of labor, did not happen in the communist countries. Transformation toward market economy changed the situation in labor market dramatically. Due to UNICEF (1999) women’s labor market participation has been falling in many transitional economies since 1989. The scale of the collapse in participation during transition period was very large. For example, in Poland, about one and a half million female jobs disappeared between 1989 and 1994 (Newell and Barry 2001). Situation in post-communist countries is discussed by Klasen (1994), Grajek (2003), Keane and Prasad (2006), Lehmann and Wadsworth (2001), among others.

Data DescriptionAnalysis is provided upon the European Union Structure of Earnings Survey0 (SES) data collected from 2006. We use individual data, that Eurostat calls microdata. The SES is a survey providing information on relationships between the level of remuneration, individual characteristics of employees and their employer (economic branch, age, occupation, job seniority, size of enterprise, collective pay agreement, type of employment contract among others). The statistics of the SES refer to the enterprises with at least 10 employees. To achieve our goal, we conduct our research separately for six regions presented in Table 2 due to NUTS classification0. There are four regions representing West Germany and one region representing East Germany in NUTS 2 classification level, and one region representing Poland in NUTS 0 classification level.

Table 2. Investigated regions No. of region Code of Region Description

1 DEU West Germany, NUTS 2 regions: (DE3) Berlin, (DE5) Bremen, (DE6) Hamburg, (DE9) Lower Saxony, and (DEF) Schleswig-Holstein

2 DEV West Germany, NUTS 2 regions: (DEA) North Rhine-Westphalia3 DEW West Germany, NUTS 2 regions: (DE7) Hesse, (DEC) Saarland, and (DEB) Rhineland-Palatinate4 DEX West Germany, NUTS 2 regions: (DE1) Baden-Württemberg, and (DE2) Bavaria

5 DEY East Germany, NUTS 2 regions: (DE4) Brandenburg, (DE8) Mecklenburg-Vorpommern, (DED) Saxony-Anhalt and (DEG) Thuringia

6 PL Poland, NUTS 0 regionSource: Own elaboration based on NUTS classification.

MethodologyIn the present analysis we estimate several exponential models describing gross hourly wages in Poland and in five selected regions in Germany. For each region the models are estimated based on the three samples, containing all employees, employed women (denoted by F), and employed men (denoted by M). The dependent variable in all models is defined as lnHEi - natural logarithm of gross hourly wages (obtained by all employees, women or men). The explanatory variables0 are defined as sets of dummies because they represent qualitative characteristics of the employees and the employers. There are eight features which are selected as potential factors influencing wages.

0 Structure of Earnings Survey has been provided every four year since 2002. There are some differences between metadata because of imported correction in every survey. For example, in each survey (2002, 2006 and 2010) was different definition (SES 2006 and SES 2010) or different range (SES 2002 and SES 2006) of economic branches. So in presented analysis is used database from 2006. SES is a survey conducted in accordance with the Council Regulation No. 530/1999 and the Commission Regulation No. 1916/2000 as amended by Commission Regulation No. 1738/2005. The SES for 2006 is the second of a series of four yearly.0 NUTS (from the French version Nomenclature des Unités territoriales statistiques) is a geographical nomenclature subdividing the economic territory of the European Union.0 In the models were exploited characteristics for which information were available in the database and those that were given for all the analyzed regions.

20


Table 3. Selected dummy variables and theirs optionsDummy variable Options ( j )Branch(economicsector)

NACE ( j )i

C - Mining and quarryingD - manufacturingE - electricity, gas and water supplyF - constructionG - wholesale and retail trade; repair of motor vehicles, motorcycles and personal and household goodsH - hotels and restaurantsI - transport, storage and communicationJ - financial intermediationK - real estate, renting and business activitiesL - public administration and defense; compulsory social securityM - education (reference option)N - health and social workO - other community, social, personal service activities

Size ofEnterprise

¿( j¿)i

SIZE(0) - between 10 and 49 employees (reference option)SIZE(1) - between 50 and 249 employeesSIZE(2) - 250 employees and more

Age

AGE ( j )iAGE(1) - between 20 and 29 years (reference option)AGE(2) - between 30 and 39 yearsAGE(3) - between 40 and 49 yearsAGE(4) - between 50 and 59 yearsAGE(5) - 60 years and over

Occupation

OCCUP ( j )iOCCUP(1) - Legislators, senior officials and managersOCCUP(2) - ProfessionalsOCCUP(3) - Technicians and associate professionalsOCCUP(4) - ClerksOCCUP(5) - Service workers and shop and market sales workersOCCUP(7) - Craft and related trades workersOCCUP(8) - Plant and machine operators and assemblersOCCUP(9) - Elementary occupations (reference option)

Education

EDU ( j )i

EDU(1) - Pre-primary and primary education - levels 0-1EDU(2) - Lower secondary education – level 2 (reference option)EDU(3) - Upper secondary education - levels 3EDU(4) - Post-secondary non-tertiary education - level 4EDU(5) - Tertiary education - levels 5A and 5BEDU(6) - Tertiary education - level 6

Type of contract

TYPE ( j)i

TYPE(A) - Indefinite duration (reference option)TYPE(B) - Fixed term (except apprentice and trainee)TYPE(C) - Apprentice or trainee

Source: Own elaboration based on Structure of Earnings Survey 2006: Eurostat’s arrangements for implementing the Council Regulation 530/1999, the Commission Regulations 1916/2000 and 1738/2005.

1. NACE ( j )i - branch described by dummy variables defined for NACE0 codes: j = C, D,…, O, that refers to variants of variable presented in the Table 3.

2. ¿( j¿)i - group of dummies describing the size of the enterprise measured by number of employees, j = 0, 1, 2 (variants of variable see in Table 3).

3. SEXi - gender of employee; 0 = male, 1 = female.

4. AGE ( j )i - the age group described by dummy variables distinguished for six intervals of age, j = 0, 1, 2, 3, 4, 5 (variants of variable see in Table 3).

5. OCCUP ( j )i - occupation described by dummies defined for ISCO-88 major group codes0: j = 1, 2, 3, 4, 5, 6, 7, 8, 9 (variants of variable see in Table 3).

6. EDU ( j )i - level of education of employees that is considered as binary variables, defined for 6 levels of education (see Table 3).

7. JOBi - full-time or part-time employee; 0 - if person is full-time employed 1 - if person is part-time employed.

8. TYPE ( j)i - type of employment contract, j = A, B, C (variants of variable see Table 3).

For each explanatory variable the reference variant is distinguished. For dichotomous variables the reference variant is represented by the feature equaled 0, for multivariate variables the reference values are bold in Tab 3.

Results0 NACE is the classification of economic activities in the Europen Union, which is derived from the French Nomenclature statistique des activités économiques dans la Communauté européenne. Various NACE versions have been developed since 1970. 0 ISCO is the International Standard Classification of Occupations.

21


All models are estimated applying general last square (GLS) method with heteroscedasticity correction. The results obtained for each estimated model are presented in Tables 4 and 5. Interpretation of the parameter estimates for qualitative variables represented by the set of dummies is provided according to so called reference variant of the variable i.e. the one not presented in models. In the models estimated on the base of the whole samples (for each region), presented in Table 4, all parameters are statistically significant at the significance level =0.05. The parameter estimates standing by the variable representing gender are negative. In other words, women earn less about 10% on average than men in all examined regions since this parameter represents differences between men and women gross hourly wages in investigated samples. The biggest value of the parameter estimate is obtained for Poland (-0.155) and the lowest for North Rhine-Westphalia (-0.087). Obtained results indicate that difference in men and women wages in Poland is bigger than in Germany but the direct comparison of the parameters estimated for different models is not justified0. In the group of variables representing economic sectors (NACE ( j )i) we can notice a clear difference in values of estimated parameters. Reference branch in this group is educational sector (M). In Poland and in East Germany employees of the majority of economic sectors earn less than in education (with exception of construction (C), electricity (E) and finance (J). While in West Germany all parameters (but sector H in DEV region) standing by dummies representing economic branches are positive – i.e. employees earn significantly more that in education. Table 4. Parameter estimates for the models estimated using the samples with all employeesVariable DEU DEV DEW DEX DEY PL

Parameter estimatesconst 1.7404 1.9316 1.7957 1.8139 1.9359 2.1314NACE(C) x x x x 0.1448 0.3384NACE(D) 0.2858 0.2770 0.2822 0.3321 -0.0213 -0.1400NACE(E) 0.3997 0.4162 0.4281 0.3949 0.2421 0.0563NACE(F) 0.2097 0.2090 0.2663 0.2894 -0.0266 -0.0969NACE(G) 0.1618 0.1880 0.2235 0.2496 -0.0468 -0.1894

NACE(H) 0.0348 -0.0157 0.1635 0.1620 -0.1873 -0.1274

NACE(I) 0.1357 0.1470 0.1568 0.1901 x -0.1227NACE(J) 0.3612 0.3320 0.3918 0.3615 0.1628 0.0773NACE(K) 0.1232 0.1770 0.2302 0.2226 -0.1816 -0.2252NACE(L) x x x x x -0.1055NACE(N) 0.0954 0.1053 0.1388 0.1411 -0.0262 -0.2638NACE(O) 0.1223 0.1151 0.1635 0.1649 -0.0998 -0.1567SIZE(1) 0.0836 0.0938 0.0663 0.0813 0.0823 0.0901SIZE(2) 0.1980 0.2041 0.2214 0.2030 0.2073 0.1844

SEX(1) (female) -0.1019 -0.0874 -0.1102 -0.1101 -0.0961 -0.1551

AGE(2) 0.3404 0.1528 0.2751 0.2472 0.1766 0.1229AGE(3) 0.4736 0.2782 0.4041 0.3779 0.2456 0.1629AGE(4) 0.5337 0.3231 0.4853 0.4390 0.2872 0.1852AGE(5) 0.6282 0.4419 0.6131 0.5259 0.3952 0.1546OCCUP(1) 0.5973 0.6939 0.6196 0.6369 0.6518 0.7716OCCUP(2) 0.3690 0.5398 0.2785 0.4990 0.5048 0.5680OCCUP(3) 0.3032 0.3058 0.3159 0.3354 0.3494 0.3972OCCUP(4) 0.2441 0.1739 0.2136 0.2079 0.2783 0.2490

OCCUP(5) 0.0061 -0.0097 -0.0074 0.0131 0.0617 0.0614

OCCUP(6) -0.0379 0.0118 -0.0020 0.0098 0.0888 0.0956OCCUP(7) 0.1246 0.1253 0.1309 0.1149 0.1946 0.1902OCCUP(8) 0.0309 0.0580 0.0614 0.0236 0.1237 0.2359EDU(1) x x x x x -0.0121EDU(3) 0.1023 0.1252 0.1167 0.1219 0.0389 0.0778EDU(4) 0.2638 0.1130 0.3099 0.1420 0.1895 0.2388EDU(5) 0.4013 0.2630 0.4719 0.3191 0.3101 0.3088EDU(6) x x x x x 0.5388JOB(1) 0.0540 0.0199 0.0578 0.0473 0.0151 0.0082

TYPE(B) -0.0327 -0.0411 0.0068 -0.0256 -0.1570 -0.1833

0 In fact, actual GPG calculated by Eurostat for 2006 (according the SES methodology) was 7.5% in Poland and 22.7% in Germany on average.

22


TYPE(C) -0.9061 -1.0651 -0.9447 -0.9440 -1.0895 x

R^2 0.5890 0.6802 0.5675 0.6217 0.6847 0.5488cor. R^2 0.5889 0.6802 0.5674 0.6217 0.6846 0.5488Source: Own estimation. Note: bold letters denote parameters significant for =0.05, x - lack of variable in the model because data are not available.

In all models parameters that represent size of enterprises are significantly positive. This finding indicates that there are significant differences in average salaries according to size of the enterprise, i.e. eemployees in bigger firms earn more. This is consistent with the conclusions presented by Oi and Idson (1999) and Mortensen (2012).Affiliation to the age group is also relevant. Remunerations of elder employees are higher. This is connected with professional experience acquired during work. The ranges of the parameter values in the models estimated for German employees are higher than in the model for Poland. Significant differences in remuneration can be observed also in the occupational groups. High-qualified employees obtain better payment. In comparison to the elementary occupations managers and professionals earn much more. Also higher level of education promotes higher level of salaries. In all models average hourly wages part-time employees are significantly higher that full-time employees. Obtained results confirm the general trends in the labor markets.In the models, estimated separately for men and women employed in investigated regions, the tendencies have been partly preserved (Table 5). For example, salaries for both men and women in education - sector M in West Germany are usually lower than in other economic sectors in opposite to Poland, and female wages in East Germany although it is not true for men working in the region DEY. In both groups (women and men) wages increase with obtaining higher position of employment, with higher levels of education or working for the larger companies (this are natural tendencies, independent of sex).Beside this we could observe also some interesting phenomena. Firstly, if we pay attention to the professions we can noticed that for variable represented managers (OCCUP_1) in the "men’s" equations, the parameters have higher value than in the "women’s" ones, in contrast to the variable that represents professionals (OCCUP_2) where parameters in "men’s" equation parameters have lower value than in "women’s" ones in all analyzed cases. It means that differences between wages obtained by managers and elementary occupations (which is reference variant of that variable) for women are smaller than for men while the differences between professionals and elementary occupations are higher for women than for men.Secondly, in Poland the "size of enterprises" effect is stronger for male employees than female employees, what is especially visible for enterprises with 50-249 employees. Values of parameters standing by SIZE_2 variable in "German" equations oscillate around the 0.2, while in Poland in "men" equation is close to 0.3 and in "women" one - close to 0.1. Values of these parameters describe the difference in wages obtained by employees working in bigger companies in comparison to the small one i.e. employing from 10 to 49 employees.In West Germany both men and women part-time employed earned significantly more than full-time employees. In East Germany the situation is different from that in the rest of the country in the group of men since men part-time employed earn significantly less than full-time employees. In turn in Poland unless the situation for men is similar to that observed in West Germany, women employed part-time earned on average less than those employed full-time.In Germany employees with indefinite duration contract earn much more than apprentice or trainee. This is a typical situation mainly due to the larger professional experience of the employees with indefinite duration contract. In the case of Poland, the apprentice or trainee group was not considered due to lack of data. Employees with fixed term contract usually earn less than those with indefinite duration contract. In all the analyzed regions, these differences are greater for men.

Table 5. Parameter estimates for the models estimated using the samples containing male (M) and female (F) employees

VariableDEU DEV DEW DEX DEY PL

M F M F M F M F M F M F

const 1.797 1.629 1.893 1.898 1.792 1.729 1.796 1.750 1.969 1.839 1.993 1.995

NACE(C) x x x x x x x x 0.164 0.112 0.386 0.134

NACE(D) 0.382 0.197 0.362 0.195 0.392 0.170 0.441 0.225 0.050 -0.098 -0.057 -0.147

NACE(E) 0.488 0.273 0.486 0.264 0.519 0.297 0.478 0.279 0.280 0.192 0.113 0.010

NACE(F) 0.294 0.059 0.278 -0.013 0.358 0.103 0.374 0.107 0.024 -0.128 -0.026 -0.181

NACE(G) 0.263 0.081 0.280 0.087 0.327 0.139 0.360 0.162 0.002 -0.078 -0.069 -0.222

NACE(H) 0.073 -0.019 -0.016 -0.098 0.239 0.089 0.196 0.102 -0.183 -0.191 -0.026 -0.149

23


NACE(I) 0.217 0.065 0.231 0.056 0.250 0.079 0.289 0.092 x x -0.083 -0.087

NACE(J) 0.461 0.299 0.421 0.261 0.498 0.309 0.480 0.271 0.215 0.142 0.210 0.050

NACE(K) 0.205 0.060 0.241 0.096 0.300 0.166 0.311 0.127 -0.140 -0.205 -0.184 -0.182

NACE(L) x x x x x x x x x x -0.064 -0.094

NACE(N) 0.158 0.080 0.181 0.066 0.219 0.119 0.225 0.123 0.016 -0.029 -0.238 -0.253

NACE(O) 0.217 0.071 0.211 0.052 0.274 0.105 0.258 0.108 -0.027 -0.138 -0.044 -0.211

SIZE(1) 0.081 0.075 0.080 0.115 0.058 0.066 0.070 0.074 0.076 0.085 0.157 0.058

SIZE(2) 0.193 0.183 0.195 0.202 0.212 0.203 0.196 0.183 0.225 0.178 0.297 0.116

AGE(2) 0.240 0.391 0.165 0.143 0.214 0.285 0.199 0.265 0.146 0.201 0.129 0.126

AGE(3) 0.389 0.518 0.300 0.262 0.366 0.401 0.361 0.379 0.207 0.276 0.158 0.179

AGE(4) 0.458 0.575 0.358 0.301 0.448 0.478 0.433 0.436 0.225 0.322 0.161 0.216

AGE(5) 0.541 0.682 0.455 0.435 0.548 0.632 0.512 0.536 0.306 0.459 0.149 0.236

OCCUP(1) 0.625 0.521 0.703 0.614 0.621 0.578 0.662 0.571 0.646 0.628 0.840 0.719

OCCUP(2) 0.342 0.388 0.517 0.521 0.266 0.319 0.463 0.541 0.418 0.552 0.559 0.596

OCCUP(3) 0.318 0.294 0.336 0.278 0.316 0.323 0.344 0.330 0.302 0.383 0.373 0.421

OCCUP(4) 0.241 0.260 0.156 0.196 0.233 0.242 0.202 0.249 0.220 0.326 0.158 0.298

OCCUP(5) 0.057 0.015 0.127 0.001 0.070 0.009 0.106 0.026 0.115 0.085 -0.015 0.112

OCCUP(6) -0.055 -0.046 -0.032 0.009 -0.044 0.054 -0.011 0.018 0.018 0.133 0.084 0.122

OCCUP(7) 0.088 0.070 0.096 0.030 0.102 0.093 0.093 0.010 0.147 0.143 0.189 0.066

OCCUP(8) -0.004 0.022 0.025 -0.001 0.039 0.000 -0.002 -0.029 0.083 0.196 0.210 0.240

EDU(1) x x x x x x x x x x -0.027 0.013

EDU(3) 0.108 0.080 0.132 0.100 0.122 0.075 0.133 0.072 0.057 0.003 0.066 0.082

EDU(4) 0.264 0.244 0.116 0.104 0.313 0.271 0.173 0.087 0.221 0.148 0.207 0.246

EDU(5) 0.444 0.375 0.300 0.253 0.522 0.424 0.382 0.263 0.381 0.268 0.255 0.337

EDU(6) x x x x x x x x x x 0.444 0.578

JOB(1) 0.042 0.049 0.012 0.018 0.069 0.052 0.027 0.045 -0.013 0.017 0.050 -0.021

TYPE(B) -0.088 -0.001 -0.073 -0.015 -0.042 0.018 -0.041 -0.005 -0.168 -0.142 -0.200 -0.155

TYPE(C) -1.073 -0.775 -1.104 -1.038 -1.069 -0.862 -1.039 -0.870 -1.168 -1.024 x x

R^2 0.564 0.623 0.646 0.730 0.558 0.601 0.586 0.667 0.673 0.704 0.488 0.631

Cor. R^2 0.564 0.623 0.646 0.730 0.558 0.601 0.586 0.667 0.673 0.704 0.488 0.631

Source: Own estimation. Note: bold letters denote parameters significant for =0.05, x - lack of variable in the model because data are not available, M - male, F - Female.

Conclusions The comparison of labor markets in Poland, Germany and 27 European Union member states in terms of women employment rate and gender pay gap shows (Table 1) that in recent years, the women employment rate in Poland is lower than in Germany and European Union as an aggregate. In the same time gender pay gap is the smallest in Poland while the highest in Germany.

Present study contains the results of the econometric models estimation on the basis of microdata. It worth mentioning that all variables used in the models (economic branch, size of enterprise, age, level of education, profession, type of employment contract) significantly affect the amount of remuneration. However, the variability of the women’s remuneration is better explained than men’s salaries in the estimated models.The main differences between Germany and Poland are as follows. Firstly, values of the parameters standing by the variables AGE ( j )i indicate that the groups of older workers (compared to a group of younger workers) are on average better paid in Germany, especially in western lands, than in Poland since the differences of parameters estimated for the age groups has bigger range. This can be interpreted that the age bonus (which is usually connected with a greater work experience) is higher in Germany than in Poland. Secondly, women part time employed in Poland earn significantly less than full-time employed. This is quite opposite to the situation in Germany. Thirdly, employees with fixed term contract (in comparison to the ones with indefinite duration contract) are much worse remunerated in East Germany and in Poland than in West Germany. Finally, salaries for both men and women in education in East Germany and Poland are usually higher than in other economic sectors in opposite to the West Germany.

24


References1. Adamchik V.A. Bedi A.S., (2003), Gender Pay Differentials Turing the Transition in Poland, Economics of Transition, Vol.

11(4), pp. 697-726.2. Anderson T., Forth J., Metcalf H., Kirby S., (2001), The Gender Pay Gap, Final Report to the DfEE (UK Department for

Education and Employment).3. Arulampalam W., Booth A.L., Bryan M.L., (2007), Is There a Glass Ceiling over Europe? Exploring the Gender Pay Gap

across the Wage Distribution, Industrial and Labor Relations Review, ILR Review, Cornell University, ILR School, 60(2), pp. 163-186.

4. Blau F.D., (2012), Gender, inequality, and wages, Oxford University Press.5. Blau F.D., Kahn L.M., (2000), Gender Differences in Pay, Journal of Economic Perspectives, Vol. 14(4), pp. 75–99.6. Blau F.D., Kahn L.M., (2006), The U.S. Gender Pay Gap in the 1990s: Slowing Convergence, Industrial and Labor Relations

Review, Vol. 60(1), pp. 45–66.7. Blecker R., Seguino S., (2002), Macroeconomic Effects of Reducing Gender Wage Inequality in an Export-Oriented, Semi-

industrialized Economy, Review of Development Economics, Vol. 6(1), pp. 103-119.8. Blinder A. (1973), Wage Discrimination: Reduced Form and Structural Estimates, The Journal of Human Resources, Vol.

VII(4), pp. 436-55.9. Brainerd, E., (2000), Women in Transition: Changes in Gender Wage Differentials in Eastern Europe and then Former Soviet

Union, Industrial and Labor Relations Review, Vol. 54(1), pp. 138-162. 10. Domański Cz. [ed.], (2005), Analiza rozkładów płac i dochodów w Polsce w przekroju terytorialnym), Wydawnictwo UŁ,

Łódź (in Polish - Analysis of wages and income distribution in Poland).11. Daly, A., Kawaguchi A, Meng X., Mumford K., (2006), The Gender Wage Gap in Four Countries, IZA Discussion Paper No.

1921, Institute for the Study of Labor (IZA).12. Du Caju, P., Rycx F., Tojerow I., (2011), Inter-Industry Wage Differentials: How Much Does Rent Sharing Matter?, The

Manchester School, Vol. 79, pp. 691-717.13. Gannon, B., Plasman R., Rycx F., Tojerow I., (2007), Inter-Industry Wage Differentials and the Gender Wage Gap: Evidence

from European Countries, Economic and Social Review, Vol. 38(1), pp. 135-155.14. Goraus K., Tyrowicz J., (2014), Gender Wage Gap in Poland – Can It Be Explained by Differences in Observable

Characteristics?, WNE Working Papers No. 2014-11, Faculty of Management, University of Warsaw.15. Grajek M., (2003), Gender Pay Gap in Poland, Economic Change and Restructuring, Vol. 36(1), pp. 23-44.16. Grimshaw D., Rubery J., (2007), Undervaluing Womenʹs Work, EOC Working Paper No 53, Manchester EOC.17. Keane M. P., Prasad E.S., (2006), Changes in the structure of earnings during the Polish transition, Journal of Development

Economics, Vol. 80, pp. 389– 427.18. Klasen S., (1999), Does Gender Inequality Reduce Growth and Development? Evidence from Cross-Country Regressions,

World Bank Working Paper Series No. 7.19. Klasen S., (1994), Human Development and Women’s Lives in a Restructured Eastern Block: Lessons from the Developing

World, in: Schipke A., Taylor A.M., The Economics of Transition, Springer – Verlag, Berlin – Heidelberg, pp. 253 – 294.20. Kot, S. M. [ed.], (1999), Analiza ekonometryczna kształtowania się płac w Polsce w okresie transformacji, Warszawa,

Kraków, PWN (in Polish - Econometric analysis of wage in Poland in transition period).21. Lehmann, H., Wadsworth J., (2001), Wage Arrears and the Distribution of Earnings in Russia, Discussion Paper 410, Institute

for the Study of Labor (IZA).22. Löfström Å., 2009, Gender equality, economic growth and employment, working paper.23. Matuszewska-Janica A., Witkowska D., (2013), Application of Classification Trees to Investigate Gender Wage Gap,

Research Papers of Wrocław University of Economics, No. 279, Taksonomia 21, pp. 58-66, (in Polish).24. Morrison A., Raju D. Sinha N., (2007), Gender Equality, Poverty and Economic Growth, Policy World Bank Policy Research

Working Paper No. 4349.25. Mortensen D.T., (2012), Wage Dispersion: Why Are Similar Workers Paid Differently?, PTE, Warszawa, [Polish edition].26. Newell A., Socha M., (2005), The Distribution of Wages in Poland, 1992-2002, IZA Discussion Papers No. 1485, Institute for

the Study of Labor (IZA).27. Newell, A., Barry R., (2001), The Gender Pay Gap in the Transition from Communism: Some Empirical Evidence, IZA

Discussion Paper 268, Institute for the Study of Labor (IZA).28. Ñopo H., (2008), Matching as a Tool to Decompose Wage Gaps, The Review of Economics and Statistics, Vol. 90(2), pp. 290-

299.29. Oaxaca R. (1973), Male-Female Wage Differentials in Urban Labor Market, International Economic Review, No. 14(3), pp.

693-709.30. Oaxaca R.L., Ransom M.R. (1994), On discrimination and the decomposition of wage differentials, Journal of Econometrics,

No. 61(1), pp. 5-21.31. Oi W.Y., Idson T.L. (1999), Firm Size and Wages in: Ashenfelter O., Card D. [ed.], Handbook of Income Distribution, Vol.

3B, Elsevier, Amsterdam, pp. 2165-2214.32. Plasman, R., Rycx F., Tojerow I., (2006), Industry wage differentials, unobserved ability, and rent-sharing: evidence from

matched worker-firm data, 1995–2002, IZA Discussion Paper, No. 2387, Institute for the Study of Labor (IZA).33. Seguino S., (2000), Gender Inequality and Economic Growth: A Cross-Country Analysis, World Development, Vol. 28(7), pp.

1211-1230.34. Śliwicki D., Ryczkowski M. (2014), Gender Pay Gap in the micro level – case of Poland, Quantitative Methods In Economics,

No. XV(1), pp. 159–173.35. Słoczyński T., (2013), Population Average Gender Effects, IZA Discussion Papers No. 7315, Institute for the Study of Labor

(IZA).36. UNICEF, (1999), Women in Transition, The MONEE Project Regional Monitoring Report, No. 6, ICDC, Florence, Italy.37. Vöörmann R., (2009), Gender Segregated Labour Markets in the Baltics What are Prevailing – Similarities or Differences?,

Studies of Transition States and Societies, Vol. 1(1), pp. 66-80.

25


38. Witkowska D., (2013), Gender Disparities in the Labor Market in the EU, International Atlantic Economic Research, Vol. 19(4), pp. 331-254.

26

ANALYSIS AND COMPARISON OF DES CRYPTOGRAPHIC ALGORITHM AND AES CRYPTOGRAPHIC ALGORITHM IN

DIFFERENT CPU

Florim IDRIZI1, Agon MEMETI2, Burhan RAHMANI3

State University of Tetovo, Tetovo, Republic of Macedonia1, 2, 3

[email protected] [email protected] [email protected]

AbstractNowadays it is highly important the security while data transmission. Since everything nowadays is transmitted through the Internet, it is very likely for our data to be taken and misused. What we have conducted is mini (minor) software in the c# language, which makes encryption of the file in .txt format, and what we will conduct in this paper is the measurement of time of encryption of different size of files with DES algorithm and AES algorithm of different CPUs by encrypting files, with which we will make comparisons between two algorithms and also make comparisons between the CPUs.Keywords: DES, AES, CPU, encryption, algorithms.

1. IntroductionThe goal of cryptography is to make it possible for two people to exchange a message in such a way that other people cannot understand the message. There is no end to the number of ways this can be done, but here we will be concerned with methods of altering the text in such a way that the recipient can undo the alteration and discover the original text (Sumitra, 2013). In this paper there is provided a comparison of DES cryptographic algorithm and AES cryptographic algorithm in different CPU. The paper is composed in three sections. The first and the second part starts with the DES and AES algorithms, continuing with the last part which deals with the results and the experiments.

2. DES AlgorithmData encryption standard (DES) is a private key cryptography system that provides the security in communication system but now a days the advancement in the computational power the DES seems to be weak against the brute force attacks DES is based on two fundamental attributes of cryptography: Substitution (confusion) and transposition (Diffusion) (Singh, 2013). DES consists of some steps, each of which called as a Round. [1] In the first step, the initial 64-bit plain text block is handed over to in Initial Permutation (IP) function. [2] The Initial permutation is performed on plain text. [3] The initial permutation produce two halves of permuted block: Left Plain Text (LPT) and Right Plain Text (RPT) (Singh, 2013). In addition, it is a symmetric algorithm, means same key is used for encryption and decryption. It uses one 64-bit key. Out of 64 bits, 56 bits make up the independent key, which determine the exact cryptography transformation, 8 bits are used for error detection. The main operations are bit permutations and substitution in one round of DES. Six different permutation operations are used both in key expansion part and cipher part. Decryption of DES algorithm is similar to encryption, only the round keys are applied in reverse order. The output is a 64-bit block of cipher text. Many attacks and methods recorded the weaknesses of DES, which made it an insecure block cipher key (Mandal, 2012).

3. AES AlgorithmAES is a block cipher with a block length of 128 bits. AES allows for three different key lengths: 128, 192, or 256 bits. Encryption consists of 10 rounds of processing for 128- bit keys, 12 rounds for 192-bit keys, and 14 rounds for 256- bit keys. Each round of processing includes one single-byte based

27




substitution step, a row-wise permutation step, a column-wise mixing step, and in addition the round key. The order in which these four steps are executed is different for encryption and decryption (Rihan, 2015).AES perform following four operations:

I. ADD ROUND KEY (XORs the round key with the state), II. BYTE SUB (a substitution using an S-box), III. SHIFT ROW (a permutation), IV. MIX COLUMN (unless it is the last round) (Patil,

2014).

4. ResultsFor our experiment we will use three laptops with different CPUs, one laptop is 2.10 GHz i7, the other is i5 and other last one is 3.10 GHz i3. The experiment is conducted through the program created by us, in the language of programming C # in Visual Studio 2013, where DES and AES algorithms are used. The program measures the time of encryption of the file, while in the experiment we will use files of different sizes, from where we will see even more clearly the differences in the time of encryption of the files on different algorithms.Moreover in order for our experiment to be more precise, we have made each encryption of the file three times, with which we find the average and then we have listed the obtained value in the table. In the following we have the table with the obtained data from the conducted experiment during the encryption of files with different sizes.

Table 1: The time of encryption of files with DES algorithm in different CPU The size of the file CPU i3(sec) CPU i5 (sec) CPU i7 (sec)

100 Kb 10 9.2 7.7200 Kb 18.4 14.2 12.5500 Kb 35.2 31.3 26.21 Mb 110.3 59.8 39.52 Mb 140.5 110.6 58.65 Mb 300.1 275.3 180.110 Mb 621.8 545.9 272.220 Mb 1430.5 1112.8 773.250 Mb 4203.2 3117.5 2200.5

Fig.1. The time of encryption of files with DES algorithm in different CPU

28

Analysis and comparison of DES cryptographic algorithm and AES cryptographic algorithm in different CPU

Table 2: The time of encryption of files with DES algorithm in different CPU Size of file CPU i3(sec) CPU i5 (sec) CPU i7 (sec)

100 Kb 11.3 9.2 5.1200 Kb 14.4 12.4 7.5500 Kb 23.7 19 15.11 Mb 38.5 29.9 26.62 Mb 67.6 134.6 49.95 Mb 165.9 140.3 122.410 Mb 317.1 251.6 225.520 Mb 626.4 530.3 48750 Mb 1577.3 1405 1228

Fig.2. The time of encryption of files with DES algorithm in different CPU

Table 3: The time of encryption of files with DES algorithm and AES algorithm – i3 CPUThe size of the file DES Algorithm (sec) AES Algorithm (sec)

50 Kb 7.6 4.0100 Kb 10 7.8200 Kb 18.4 14.8500 Kb 37.2 36.71 Mb 110.3 75.32 Mb 140.5 152.13 Mb 209.1 236.84 Mb 281 304.9

Fig.3. The time of encryption of files with DES algorithm add AES algorithm

Table 4: The time of encryption of files with DES algorithm and AES algorithm – i5 CPU

29


The size of the file DES Algorithm (sec) AES Algorithm (sec)50 Kb 8.3 3.9100 Kb 9.2 7.2200 Kb 14.2 14.5500 Kb 31.3 361 Mb 59.8 73.32 Mb 110.6 144.8

Table 5: The time of encryption of files with DES algorithm and AES algorithm – i7 CPUThe size of the file DES Algorithm (sec) AES Algorithm (sec)

50 Kb 7 3.6100 Kb 7.7 7200 Kb 12.5 14.3500 Kb 26.2 361 Mb 39.5 73.22 Mb 58.6 140.6

5. ConclusionA comparative study was carried out on this paper between two different algorithms and between different CPUs by realizing encryption of files with different sizes. What was gained as a result was that the encryption with DES algorithm had significant differences in the time of the encryption comparing with AES algorithm if files are small then 1 Mb, but as the size of the file is bigger we see that the AES algorithm is more slowly comparing with DES algorithm.

ReferencesGupta, N. (2012) Implementation of Optimized DES Encryption Algorithm upto 4 Round on Spartan 3, International Journal of Computer Technology and Electronics Engineering Jain, R et al. (2014) AES Algorithm Using 512 Bit Key Implementation for secure communication, International Journal of Innovative Research in Computer and Communication Engineering.Rihan, Sh. (2015) A Performance Comparison of Encryption Algorithms AES and DES, International Journal of Engineering Research & Technology.Mandal, P. (2012) Evaluation of performance of the Symmetric Key Algorithms: DES, 3DES, AES and Blowfish, Journal of Global Research in Computer Science.Patil, S and Patil, R. (2014) Faster Transfer of AES Encrypted Data over Network, International Journal of Computer Science and Information Technology.Taqa, A et al. (2009) New Framework for High Secure Data Hidden in the MPEG Using AES Encryption Algorithm, International Journal of Computer and Electrical Engineering.Singh, S et al. (2013) Enhancing the Security of DES Algorithm using Transposition CryptographyTechniques, International Journal of Advanced Research in Computer Science and Software Engineering.Sumitra. (2013) Comparative Analysis of AES and DES security Algorithm, International Journal of Scientific and Research Publications.Sachin, M. (2010) Implementation and Analysis of AES, DES and Triple DES on GSM Network, International Journal of Computer Science and Network Security.

30

AN APPLICATION OF MULTILEVEL LATENT CLASS ANALYSIS WITH ADOLESCENT GAMBLERS

Emil FRASHERI1, Prof. Assoc. Dr. Besa SHAHINI21Department of Management, Faculty of Economy, Fan S. Noli University of Korça, Albania, Tel:

+355693590129, E mail: [email protected] Department of Applied Statistics and Informatics, Faculty of Economy, University of Tirana, Albania, Tel: +355684018309, E mail: [email protected] [email protected]

AbstractIt is believed sub typing models have implications for understanding disordered gambling, and the ability to identify different groups could be of great importance in defining prevention and treatment responses. Latent class analysis (LCA) is a statistical method used to identify subtypes of adolescent gamblers. However, to avoid the violation of the observations’ independence, a multilevel latent class analysis was performed amongst students and middle school adolescents, across126 urban and rural communities in the Korça region (southeast Albania). Parametric and nonparametric approaches were used for estimating a two-level latent class analysis with ordinal observed variables, using Mplus . In addition, both latent class and indicator-specific random effects models were explored. The resulting best model was comprised of three Level 1 latent gambling classes (problem gamblers, at-risk gamblers and non-problem gamblers). The computation time was reduced dramatically when either parametric or nonparametric random effects in the two-level latent class model were included. However, a significant improvement of the model was not observed.Keywords: disordered gambling, multilevel latent class analysis, parametric approach, non-parametric approach.

1. IntroductionDisordered gambling is being viewed increasingly as a behavioural addiction and has been re-classified as an addictive disorder in the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-V, APA, 2013). Problem gambling research, regularly involves investigating the relationship between individuals and society. Despite higher prevalence rates of problem gambling among youth, there are clear empirical and clinical findings suggesting that young problem and pathological gamblers, similar to adult gamblers, are, in fact, not a homogeneous group (Gupta and Derevensky, 1997). Consequently, it may be particularly useful to conceptualize gamblers as a discrete latent variable and use appropriate methods to model gambling behavior. In so doing, differences may be discovered that may play an important role in the etiology and prevention of the development of problem and pathological gambling. Latent class analysis (LCA) is a statistical method used to identify subtypes of individuals using a set of categorical or continuous observed variables. These subtypes are referred to as latent classes. Traditional LCA assumes that observations are independent of one another. However, in behavioral sciences, datasets are often structured hierarchically. This may lead however, to the independence of the observations being compromised and this should not be neglected when the data structure includes individuals nested in schools, families or communities, because, if these individuals were randomly selected from within population, then a traditional, fixed effect LCA would be adequate. These nested data structures require multilevel techniques, to avoid inaccuracies in parameter estimates, biased standard errors, and inflated Type I error rates for hypothesis tests regarding the parameters. In response to these needs, researchers (Vermunt, 2003, 2008; Asparouhov and Muthén, 2008; Henry and Muthen, 2010) presented a framework for assessing latent class models with nested data. This framework accounts for the nested structure of the data by allowing latent class intercepts and latent class indicator intercepts (thresholds) to vary across Level 2 units. This helps examine if and how Level 2 units influence Level 1 latent classes and Level 1 indicators which define latent class membership. These random intercepts allow the probability of membership in a particular Level 1 latent class to vary across Level 2 units, i.e. communities. For example, the probability of an individual belonging to the “problem gamblers” class is likely to vary significantly across communities. Therefore, in some communities there is a higher probability an individual will belong to the “problem gamblers” class, whereas in other communities the probability of belonging to the same class will be lower. The statistical idea of clustering emerges as appropriate for quantifying ‘‘contextual phenomena’’ which is very important

31




in social and behavior epidemiology. In this regard, people from the same community may be more similar to each other when compared to people from different communities with respect to the problem gambling outcome variable. Specifically, a portion of the gambling differences among individuals may be attributable to the environment in which they reside. A multilevel latent class model is considered in this article. In multilevel latent class analysis (MLCA), the dependent variable is latent, having so the advantage of modeling the measurement error in the observed indicators of the latent class model (Bandeen-Roche, Miglioretti, Zeger, &Rathouz, 1997; Vermunt, 2008).

2. The multilevel latent class modelLet Yijk denote the response of Level 1 unit i within cluster or Level 2 unit j on indicator k. The number of Level 2 units is denoted by J, the number of Level 1 units within Level 2 unit j is denoted by nj , and the number of items or indicators by K. A particular level of item k is denoted by sk and its number of categories by Sk. The latent class variable is denoted by Xij, a particular latent class by t, and the number of latent classes by T. Notation Yij is used to refer to the full vector of responses of case i in cluster j, and s to refer to a possible answer pattern. The probability structure defining a simple latent class model can be expressed as follows (Vermunt, 2003):

(1)The probability of observing a particular response pattern, P (Yij = s), is a weighted average of class-specific probabilities P (Yij =s/Xij = t). The weight P (Xij = t) is the probability that person i in cluster j belongs to latent class t. The indicators Yijk are assumed to be independent of each other given class membership (local independence assumption). The term P(Yijk =sk/Xij = t) is the probability of observing response sk on item k given that the person concerned belongs to latent class t. Equation (1) applies to both traditional and multilevel latent class models. In order to be able to distinguish these two models, the model probabilities have to be written in the form of logit equations. In the traditional LC model,

(2) (3)

with the identifying constraint .The fact that β and γ parameters appearing in equations (2) and (3) do not have an index j indicates that their values are assumed to be independent of the cluster to which one belongs. Taking into account the multilevel structure, we modify this assumption as follows:

(4) (5)In a model with two latent classes and with cluster-specific class-membership probabilities as defined by equation (4), with γij = 0 for identification, random coefficients are assumed to come from a normal distribution with a mean equal to γ2 and a standard deviation equal to τ2, yielding a LC model in which γ2j = γ2 + τ2 ∙ uj where the random deviation from the population average uj ∼N (0, 1). The magnitude of the U0j variance indicates the strength of the influence of the Level 2 units. That is, a larger variance indicates greater influence of the Level 2 units. For the three-class model, the random effects are:

(6)Vermunt (2003) proposed two approaches to multilevel LCA based on including either parametric or nonparametric random effects in the multilevel latent class model. The two approaches differ conceptually in how the between-cluster heterogeneity is explained. The parametric approach assumes that cluster effects originate from a certain probability distribution, whereas, the nonparametric approach assumes the existence of a discrete number of mixture components (Kaplan and Keller,

32

An Application of Multilevel Latent Class Analysis with Adolescent Gamblers

2011). The parametric approach allows for the calculation of intraclass correlations (ICC). ICC is defined as the proportion of the total variance accounted for by the Level 2 units, where the total variance equals the sum of the Level 1 and Level 2 variances. Hedeker (2003) showed how to compute ICC in random–coefficients multinomial logistic regression models. The same method can be used in multilevel latent class models as follows:

(7)This formula makes use of the fact that Level 1 variance can be set equal to the variance of the logistic distribution, which equals π2/3 ≈ 3.29. However, contrary to normally distributed continuous variables, intraclass correlation coefficients have serious drawbacks (Goldstein H, Browne W, Rasbash J, 2002) when it comes to categorical response variables because, the intraclass correlation does not convey information regarding variation between clusters and it is also not comparable with the fixed effects, which have odds ratio interpretations. For variables varying on the cluster level, the quantification is more difficult; the usual odds ratio interpretation is incorrect, because it is necessary to compare persons with different random effects, since the variable of interest does not vary between individuals within-cluster. Therefore, it seems natural to quantify variation between the random effects using odds ratios. New measures are needed in order to quantify effects and ultimately provide a better understanding of the data. Larsen and Merlo (2005) offered an alternative measure to quantify the between community variation, the median odds ratio (MOR). MOR quantifies the variation between clusters (the Level 2 variation) by comparing two persons from two randomly chosen, different clusters. Consider two persons with the same covariates, chosen randomly from two different clusters. MOR is the median odds ratio between the person of higher propensity and the person of lower propensity. MOR is very easy to calculate, because it is a simple function of the cluster variance, σ2, given by the following equation:

(8)MOR is always greater than or equal to 1. If MOR is 1, there is no variation between clusters. If there is considerable between-cluster variation, MOR will be large. The individual odds ratio (IOR) incorporates both the fixed effect and the cluster heterogeneity in an interval, allowing for a more detailed description of the covariate effect. The interval is narrow, if the between-cluster variation is small, and wide, if the between-cluster variation is large. If the interval contains 1, the cluster variability is large in comparison with the effect of the cluster-level variable. As discussed by Vermunt (2003) and Van Horn et al. (2008), the parametric approach model can be computationally heavy, particularly as Level 1 latent classes increase. Vermunt (2003) and Asparouhov and Muthén (2008) recommend the use of a common factor to model the random means and associated covariances. This model operates under the assumption that the random means are highly correlated, and these random means can be best represented by a single factor where different random means have different factor loadings. Specifying zero residual variances, this factor model reduces the dimensionality of the random means from T-1 to 1. This simplification avoids heavy computations due to numerical integration in the maximum-likelihood estimation (Henry and Muthen, 2010). In a nonparametric approach (Vermunt , 2003, 2008); Asparouhov and Muthén , 2008) a second latent class model is specified at Level 2. The T-1 random means from the Level 1 latent class solution are used as indicators of a second latent class model at Level 2. This yields a nonparametric random-coefficients LC model in which there are not only latent classes of Level 1 units, but, also, latent classes of Level 2 units, sharing the same parameter values. Such an approach has the advantages of a less strong distributional assumption and computational burden (Vermunt and Van Dijk, 2001; Muthen B. &Asparouhov, 2008) where, the normal distribution is replaced by a multinomial distribution assumption. If we denote with Wj the value of group j on the latent class variable, defining the discrete mixing distribution, in a nonparametric approach, the model for the latent class probability equals

(9)

33

Journal of Science, Innovation and New Technology Vol. 1, No. 15 – March, 2016with m denoting a particular mixture component. We can write also γtm = γt + utm where the utm come from an unspecified discrete distribution in the form of a histogram, where nonnormality is allowed. In the nonparametric approach, the specification of the random means is different than in the parametric approach. For example, having the Problem Gambling Severity Index (PGSI) nine items as latent class indicators, a nonparametric approach model with three Level1 latent classes will be as presented in Figure 1.

∙ ∙ Level 1 Within Community

Level 2 Between CommunityFigure1. Multilevel latent class model with three Level 1 latent classes. Non-parametric approach.

As described by Bijmolt, Paas, and Vermunt (2004), these random means vary across the Level 2, between-communities latent classes (labeled CB in the figure). In the within-community model the single filled circles represent the random means for the within-community latent classes (there are T-1 random means, where T equals the number of Level 1 latent classes). These random means are referred to as C#1 and C#2 in the between communities model. They are continuous latent variables that vary across communities. This variation of Level 1 parameters across Level 2 units is the most important feature of any multilevel model, and in an MLCA it is this variation that defines the between-community latent classes. Within the framework of Asparouhov and Muthén (2008), Henry and Muthen (2010) extended the conditional item probabilities of Equation 1 to have random intercepts as follows

(10)with j, k, t denoting respectively the Level 2 units, indicators and latent classes. A common factor is defined by the indicator intercepts τjkt to reduce dimensionality between Level 2 units, capturing indicator-specific cluster influence, by using different factor loadings for different random intercepts. This helps to avoid heavy computations, due to numerical integration in the maximum-likelihood estimation. This technique can be used in both the parametric and nonparametric approaches. A parametric approach with Level 2 factors on random means and random latent class indicators is presented in Figure 2.

∙ ∙ Level 1 Within Community

U01 U02

Level 2 Between Community

34

PGSI PGSI PGSI PGSI PGSI PGSI PGSI PGSI PGSI

C

C#11 C#21

CB

PGSI PGSI PGSI PGSI PGSI PGSI PGSI PGSI PGSI

C

C#11 C#21

FC FU

PGSI1 PGSI2 PGSI6 PGSI7 PGSI8 PGSI9PGSI3 PGSI4 PGSI5

W

W


Figure 2. Multilevel latent class model with three Level 1 latent classes. Parametric approach with Level 2 factor on random latent class intercepts and Level 2 factor on random latent class indicators.

3. MeasuresThe study included 1157 participants (53.5% females, 46.5% males) from the University of Korça and from middle schools in Korça district, ranging from 15 to 23 years of age. The survey was conducted between June and July 2015. Data were collected using a self-reporting questionnaire. We focused only on respondents who had gambled at least once during the previous 12 months. Nine categorical indicators were used for the latent class membership. These were the nine PGSI items (coefficient alpha = .769), including questions as follows:

1. Bet more than can really afford to lose. 2. Needed to gamble with larger amounts for same excitement. 3. Return to win back money lost. 4. Borrowed money or sold anything to get money to gamble. 5. Might have a problem with gambling. 6. Gambling has caused health problems. 7. Others critical of one’s gambling. 8. Gambling has caused financial problems. 9. Feeling guilty about one’s gambling.

In the PGSI screening tool, which is considered the most appropriate measure of disordered gambling in terms of psychometric properties (Jackson, Wynne, Dowling, Tomnay, & Thomas, 2009; McMillen & Wenzel, 2006; Svetieva & Walker, 2008), each item had four categories (never, sometimes, most of the times, and almost always). For the purposes of this study, we reduced the 4-point response categories of “never”, “sometimes”, “most of the times”, and “almost always” to dichotomous categories, collapsing both the first and last categories, transforming the items into binary indicators. For the purposes of the current study, the whole district of Korça was divided into 126 urban and rural communities. The analysis was performed using Mplus 6.12.

4. ResultsA traditional LCA using the nine PGSI-items as indicators, determining the appropriate latent class number, was first examined. Initial analyses ignored the clustering of individuals in communities. Using a robust maximum likelihood estimator (MLR) with one - through five-class models to find the best-fitting model, the three-class model was chosen, as the best fit for the data, having the lowest Bayesian information criterion (BIC; Schwarz, 1978) and the highest value of entropy (.863).The Lo-Mendell-Rubin adjusted LRT test (TECH 11) had a p-value of 0.0001, so these tests suggested that the model with three classes was indeed better than the solution with two classes. The four-class solution had nearly the same BIC value with the three-class solution. However, the entropy value was considerably lower (.828) compared with the three-class solution, and the posterior probabilities for the three-class solution were .945, .914, .950. The posterior probabilities for the four-class solution are .874, .832, .945, .835, with the second and fourth classes representing the separated classes from the three-class model. The low posterior probabilities for these two classes indicate that the model has difficulty distinguishing between people in the second and fourth class. Most important, the substantive interpretation of the three-class solution (non-problem, at-risk and problem gamblers) is theoretically meaningful, useful, and parsimonious. As such, we chose the three-class solution as the best model. In this solution, the largest class represented non-problem gamblers and comprised 73.29% of the total participants. Although these individuals had gambled at least once during the past year, none had problems with gambling or were at risk to develop such problems. The second class, described as “at-risk gamblers”, represented 24.63% of the sample. Individuals in this class displayed increased tendencies in these criteria:

Return to win back money lost Were criticized for their gambling behavior Gambling had caused financial problems

The third and smallest class, described as “problem gamblers”, constituted only 2.08% of the sample. All individuals in this class had financial problems and most of them had felt they might have a

35

Journal of Science, Innovation and New Technology Vol. 1, No. 15 – March, 2016problem with gambling, had often been criticized for their gambling behavior, and often needed to gamble with large amounts of money to get the same feeling of excitement.After considering this three-class Level 1 solution, a model utilizing the parametric approach to account for the nested structure of the data, was specified. Selecting the “non-problem gamblers” class as the reference group, two random intercepts were specified. One represented the variability in the log-odds of membership in the “problem gamblers” class across communities. The other one represented the variability in the log-odds of membership in the “at-risk gamblers” class across communities. This allows the probability that an individual will belong to a particular Level 1 latent class to vary across Level 2 units. The results of the model are presented in Table 1.

Table 1. Level 1 latent classes number best fit model and random effect parametric approaches

Info criteria

C(1) C(2) C(3) C(4) C(5)

C(2) random effect (parameter)

C(3) random effect (parameter)

C(3) random effect (parameter with Level-2 FC)

C(3) random effect (parameter with Level-2 FC+FU)

# of free param.

9 19 29 39 49 20 32 31 41

LL -3661.940

-3067.752 -3021.185 -

2985.077-

2985.077-

3063.153-

3013.787 -3014.478 -2997.705

AIC 7341.880 6173.503 6100.370 6048.154 6031.881 6167.622 6091.573 6090.955 6077.410BIC 7387.363 6269.521 6246.924 6245.244 6279.506 6268.694 6253.288 6247.617 6284.607SABIC 7358.776 6209.171 6154.810 6121.367 6123.866 6205.168 6151.646 6149.150 6154.377

Entropy 1.000 .835 .863 .828 .840 .836 .867 .864 .803

LRT .0000 .0001 .0331 .1719

With the addition of the random effects, the Bayesian information criterion BIC (Schwartz, 1978), the adjusted BIC, SABIC, (Sclove, 1987), and the entropy remained nearly the same as for the fixed effects model. The estimated mean of the random effect (random mean) for the “problem gamblers” class indicated that for communities at the average random mean, for both “problem gamblers” and “at-risk gamblers”, the average probability that an individual would be classified as a problem gambler was .01. The variance of the random mean describes the variation in the probability that an individual will belong to the class described as “problem gamblers” across communities. This variance was marginally significant (est. mean = 2.120, s.e. = 1.176) which indicates that communities might vary significantly regarding the probability that an individual would be a problem gambler. The median odds ratio (MOR) between a student in a community with the higher propensity to be a problem gambler and a student with a lower propensity to be a problem gambler was 4.01. This is a moderate odds ratio and indicates substantial community-level variability in the probability of problem gambling (low_IOR = .25; high_IOR = 49.05). ICC for the “problem gamblers” class was 39.9%. For the class of “at-risk gamblers”, the variance of the random mean was not statistically significant (est. mean = .172, s.e. = .121) and the MOR value was 1.49, indicating that, there is some variability across communities in the probability of being “at-risk gambler”, but considerable less than for problem gamblers (low_IOR = 1.65; high_IOR = 7.43). The estimated mean of the random effect (random mean) for the “at-risk gamblers” class indicates that, for communities at the average random mean for both “problem gamblers” and “at-risk gamblers”, the average probability that an individual would be classified as an “at-risk gambler” was .26. The ICC for the “at-risk gamblers” class resulted to be 5%. The neighboring two and four-class parametric random effects models were also estimated. However we conclude that the three-class model remains the best model. The BIC indicated a moderate improvement from two to three classes. In addition, entropy was maximized with three classes (.867). To extend this analysis, a common factor was included on Level 2 random means for the two-class, three-class and four-class solutions. This dramatically reduced computation time, resulting in a small increase in BIC. Utilizing a nonparametric approach, a Level 2 latent class model was added based on the random means from the Level 1 latent class solution. As presented in Table 2, the model with three Level 1 latent classes and two Level 2 latent classes performed the best, having lower BIC and higher entropy than the model with two Level 1 and Level 2 latent classes.

Table 2. Multilevel latent class model with Level 2 latent classes. Nonparametric approachInformation criteria Cw(2), cb(2) Cw(2),

cb(3)Cw(3), cb(2) Cw(3), cb(3) Cw(4), cb(2) Cw(4),

cb(3)

36


# of free parameters 21 23 32 35 43 47

Log likelihhod -3062.491 -3062.404 -3012.807 -3009.589 -2974.245 -2969.360AIC 6166.982 6170.807 6089.614 6089.179 6034.491 6032.721BIC 6273.107 6287.040 6251.329 6266.054 6251.795 6270.239SABIC 6206.405 6213.984 6149.687 6154.883 6115.213 6120.952Entropy .643 .692 .713 .688 .745 .752

With three Level 1 latent classes and two Level 2 latent classes (Figure 3), the first Level 2 latent class, with low gambling problems communities, was comprised of communities with about 82% considered to be non-problem gamblers and another 18% seen as at-risk and problem gamblers. This class represented nearly 55% of individuals. The second Level 2 latent class community, with high gambling problems was comprised of communities with about 63% considered to be non-problem gamblers and a further 37% considered “at-risk” and “problem” gamblers. This class represented about 45% of the participants.

low PG Communi-ties (55%)

high PG Communities

(45%)

020406080

100

non-problem gamblersat-risk gamblersproblem gamblers

Figure 3. Non-parametric multilevel latent class solution, cw (3), cb(2)

With three Level 2 latent classes (Figure 4), a low problem gambling community, moderate problem gambling community, and high problem gambling community emerged. Most participants lived in a low problem gambling communities (56%). In these communities, about 84% of the participants were non-problem gamblers and none of them was a problem gambler. About 23% of participants lived in moderate gambling communities and 21% of gamblers lived in problem gambling communities. In problem gambling communities none of the participants was considered a problem gambler. All the problem gamblers were found to live in moderate gambling communities.

04080

non-problem gamblersat-risk gamblersproblem gamblers

Figure 4. Non-parametric multilevel latent class solution, cw (3), cb(3)

Figure 5 demonstrates the nonparametric characterization of the random logit means for the category “at-risk gamblers”. Here, the distribution of the random means is not assumed or represented as normal, as is the case in the parametric maximum likelihood LCA model. Rather, the histogram captures the discrete distribution of the random means, which are represented by three bars showing a negative skewed distribution across the Level 2 latent classes. The bars represent the percent of students in the “at-risk gamblers” class in each between communities latent class from the Level 1 and Level 2 three-class solution (cw(3), cb(3).

37


02040

at-risk gamblers

Figure 5. The discretized, non-normal distribution of the non-parametric approach for “at-risk gamblers”

Finally, including a common Level 2 factor for the Level 1 latent class indicators, using the parametric approach (Table 1, Figure 2), the BIC did not indicate any model improvement. Moreover, the entropy showed a significant decline (.803) compared with the model with only one common factor on Level 2 random latent class intercepts (.864). This may indicate that communities did not have a substantial influence on the Level 1 indicators of individual gambling subtypes.

5. ConclusionsThe purpose of the current study was to determine the appropriate number of categories of adolescent gamblers, related gambling problems, taking into account the variability between the communities in which they reside. Parametric and nonparametric approaches, in line with previous research, were used including common factors and/or Level 2 latent classes from random intercepts and indicator intercepts. At Level 1, the three class model proved to be the most appropriate and the division of gamblers in three categories (non-problem, at-risk and problem gamblers) seems to be a relevant solution. With the addition of random effects, in the nonparametric approach, the model with three Level 1 latent classes and two Level 2 latent classes were the best performers. Adding common factors at Level 2, the computational time was dramatically reduced, but without a significant improvement of the model. This does not necessarily mean the contextual factor is not important compared with individuals’ factors. Rather, it may indicate that the geographical boundaries, we used to define the actual communities, did not correspond with the boundaries shaping the relevant environment, influencing individual gambling outcome. In the two-level model, when the response variable is categorical, MOR and IOR may be appropriate measures of the heterogeneity between clusters.

ReferencesAmerican Psychiatric Association (APA) (2013). Diagnostic and Statistical Manual of Mental Disorders (DSM-5). Arlington, VA: American Psychiatric Publishing.Bijmolt, T.H.A., Paas, L.J., & Vermunt, J.K. (2004). Country and consumer segmentation: Multi-level latent class analysis of financial product ownership. International Journal of Research in Marketing, 21(4), 323–340. DOI:10.1016/j.ijresmar.2004.06.002Goldstein, H., Browne, W. J., and Rasbash, J. (2002). Partitioning Variation in Multilevel Models, Understanding Statistics, 1, 223–232.Gupta, R., & Derevensky, J. L. (1997). Familial and social influences on juvenile gambling behavior. Journal of Gambling Studies, 13, 179-192.Hedeker, D. (2003). A mixed-effects multinomial logistic regression model. Statistics in Medicine, 22, 1433–1446.Henry, K. L., Muthen, B. (2010). Multilevel latent class analysis: An application of adolescent smoking typologies with individual and contextual predictors. Structural Equation Modeling, 17, 193–215. doi: 10.1080/10705511003659342Jackson, A. C., Wynne, H., Dowling, N. A., Tomnay, J. E., & Thomas, S. (2009). Using the CPGI to determine problem gambling prevalence in Australia: Measurement issues. International Journal of Mental Health and Addiction. doi: 10.1007/s11469-009-9238-9Kaplan, D., & Keller, B. (2011). A note on cluster effects in latent class analysis. Structural Equation Modeling: A Multidisciplinary Journal, 18, 525–536. doi: 10.1080/10705511.2011.607071Larsen, K., & Merlo, J. (2005). Appropriate assessment of neighborhood effects on individual health: Integrating random and fixed effects in multilevel logistic regression. American Journal of Epidemiology, 161(1), 81–88.McMillen, J., & Wenzel, M. (2006). Measuring problem gambling: Assessment of three prevalence screens. International Gambling Studies, 6(2), 147-174. doi: 10.1080/14459790600927845Muthén, B. O., & Asparouhov, T. (2008). Growth mixture modeling: Analysis with non-Gaussian random effects. In G. Fitzmaurice, M. Davidian, G. Verbeke, & G. Molenberghs (Eds.), Longitudinal data analysis, 143–165. Boca Raton, FL: Chapman & Hall/CRC Press, 213–239. Boston, MA: Blackwell.Schwartz, G. (1978). Estimating the dimension of a model. The Annals of Statistics, 6, 461-464.Sclove, L. S. (1987). Application of model-selection criteria to some problems in multivariate analysis. Psychometrika, 52, 333-343.

38


Van Horn, M. L., Fagan, A. A., Jaki, T., Brown, E. C., Hawkins, J. D., Arthur, M. W., et al. (2008). Using multilevel mixtures to evaluate intervention effects in group randomized trials. Multivariate Behavioral Research, 43, 289–326.Vermunt, J. K. (2003). Multilevel latent class models. In R. M. Stolzenberg (Ed.), Sociological methodology , 23, 213-239.Vermunt, J. K. (2008). Latent class and finite mixture models for multilevel data sets. Statistical Methods in Medical Research, 17(1), 33–51.Vermunt, J. K., & Van Dijk, L. (2001). A nonparametric random-coefficients approach: The latent class regression model. Multilevel Modeling Newsletter, 13, 6–13.

39

CODE AND MESSAGE FOOTPRINTS OF GENERATED CODE FOR EMBEDDED DEVICES AS COMPONENTS OF SOA

Kujtim HYSENI1, Neki FRASHERI2

1 Department of Information Engineering, Faculty of Information Technology, Polytechnic University of Tirana, Mother Teresa Square 1, Tirana, Albania

2 Center for Research and Development in IT, Faculty of Information Technology, Polytechnic University of Tirana, Mother Teresa Square 1, Tirana, Albania

[email protected], [email protected]

AbstractThis paper shows the code and message footprints of generated embedded server to be implemented in microcontroller. The code is generated by the code generator tools we constructed. Generated code represents simplified versions of CORBA and Web Services server. We give an example of temperature sensor and thermostat definition in IDL as well as its implementation, by one instance each. We show the size of server’s code after compilation, and corresponding message sizes. Besides message sizes, each byte of respective message is described in detail. Obtained results are much better than in related work when server’s size is considered and are comparable with those in related work when message sizes are considered.Keywords: code footprint, message footprint, service oriented architecture, embedded device, microcontroller, code generation

IntroductionMicrocontrollers [1, 5, 6] are heavily used to communicate sensors and actuators in configuration with a PC (Personal Computer) [7, 8]. This is due to a low price compared to other devices which play the same function with similar performance. Also PC-s are widely available on the market, and we intend to incorporate those standards because in most cases they fulfill the performance requirements. Architectures of the two project works [7, 8] use the microcontroller i.e. integrate and communicate it in a specific manner designed particularly for the project. The project work [7] is more specific because the microcontroller is connected to other devices besides sensors and actuators and has to be managed from a distance (wireless) – by contrast to standard way, through its default serial port. From our view point project [8] is more common, due to infrastructure of the connections and involved components. We don’t deal with applications (purpose) of the above mentioned projects, but only with communication and integration. What is missing to those works is the standardization of communication. We aim to build a standardized, configurable, scalable and easy maintainable way of communication with a microcontroller. In general, we see a microcontroller as a network component, which could be integrated and accessed through internet besides LAN (Local Area Network). This logically raises the question whether a new (dedicated) platform as in [9] should be developed, or common web technologies of the field such as Web Services [10], CORBA [11], .NET Remoting [12] or related should be used.Building a new platform could benefit in performance considering the limited devices (microcontrollers) that we target, but would cost in terms of time and resources. Another drawback would be in-compatibility with existing platforms and the need for training people to learn and use the newly developed platform. On the other hand, using already mentioned web technologies would make our system easy to be developed, since more people have knowledge on those technologies, and the integration into a network or the internet would become a trivial task. The bad side is that they have a large code and memory footprint, making them impossible to be implemented in limited devices such as microcontrollers 8051 [1, 6].This paper is organized as follows. Section 2 covers related works about the problem we addressed. In section 3 we propose the contribution taking in consideration related work, requirements and future perspective. Section 4 shows the structure of generated bind and embedded server code. Section 5 shows some details of the code generator we built. Section 6 shows the results of our work which are measured by microcontroller’s code size after compilation and sizes of messages exchanged between bind and embedded server.

41

2 Related workIn this section we show the most related works regarding integration of embedded limited devices in Service Oriented Architectures (SOA). Most of them are based in Web Services platform. Also, details of some most used protocols from embedded domain are given.Paper [17] deals with implementation of SOA in limited devices at which cannot be executed TCP/IP stack, and thus cannot be integrated directly in IP based networks. Solution consists of using the logic of function, instead of HTTP itself. Accessing WSN (Wireless Sensor Network) from standard network is obtained through layers, which have the primary objective “translation” of verbose HTTP requests into short µIP based.Authors in [18] propose the TinyREST protocol, which is a result of combination of messaging system integrated in TinyOS operating system over which work MicaZ limited devices, and the already known HTTP protocol REST (REpresentational State Transfer). Clients initiate REST requests which through HTTP-2-TinyRest gateway are translated in TinyREST, which are transmited to devices which control sensors/actuators wirelessly.Paper [19] presents the tool gSOAP, the code generator in C/C++ for communication based in SOAP (Simple Object Access Protocol) known as Web Services. gSOAP generates sources for stubs and skeletons which makes possible to clients and servers to communicate. Small code footprint (after compilation) and speed of communication make this tool suitable for use in embedded devices. This is due to advanced de-marshalling algorithms and pre-compilation of stubs and skeletons.Paper [20] presents a solution in integrating devices limited in resources and power in Web Services. Solution consists in two layered architecture – that based in SOAP and that in REST. Upper layer, the full functional node is accessible through SOAP protocol from clients, and at the same time represents a consumer to lower layer, to reduced functionality node implemented as REST Web Service.Interesting works dealing with integration of limited devices into SOA are shown from Käbisch, et al. [21, 22, 23]. These works are of particular interest since they represent integration of microcontrollers as very limited devices in Web Services. All the work is based on EXI [24] standard, the binary alternative of already popular XML standard. Instead of exchanging SOAP messages between clients and servers realized in Web Services technology, they exchange EXI messages which are built by encoding respective alternatives of SOAP messages.Modbus [25] is one of the longest in use protocols. Of particular interest for us is Modbus RTU [13], since it is suited for microcontrollers and the new protocol we’ll define is heavily based in Modbus RTU. Some of the characteristics are as follows. It is a strict master/slave protocol. Only one master at given time is connected to communication bus and several slaves (maximum 247). Only a master can initiate requests. Slaves cannot transmit data without receiving request from master. Slaves cannot communicate with each other. Only one slave can respond to given request.

3 The concept of bind serverWe introduce the concept of a bind server [4] which mediates between the client and the embedded server. It is technology-specific, meaning that is different for CORBA and different for Web Services. Indeed, given a technology, it is an ordinary server for that technology which translates high level requests coming from clients to simple ones based on the AMISP (Adaptive Minimal Inter-Server Protocol) protocol [2] dedicated to microcontroller. Client and bind server communicate through a technology-specific protocol, while bind and embedded server communicate through AMISP. Thus, the CORBA client and server communicate through IIOP (Internet-Inter Orb Protocol) while Web Services client and server communicate through SOAP (Simple Object Access Protocol), see figure 1. The bind server is organized in such a way that it allows accessing embedded server from different technologies. This is achieved by managing connections with the embedded server.AMISP [2] is quite simple as shown in figure 2. It is motivated by Modbus RTU [13]. Its byte (8 bit) orientation makes it even more adoptable in the microcontroller area, due to the availability of respective 8-bit instructions of data comparison and decision. Since our study is based on Service Architectures, the protocol should enable addressing of server (within microcontroller), operation within server, status of operation (in response) and bytes of data.

42

Learning Management System using REST Services in Cloud Computing

Fig. 1. Accessing limited device (8051 microcontroller) server from clients of different technologies through bind servers (from [3])

1 byte 1 byte 1 byte 1 byte n bytes m bytes …

Message length

Interface ID

Operation ID Status Prm. 1 Prm. 2 …

Fig. 2. AMISP protocol

4 Code generator and structure of generated codeTo avoid errors and required time due to manual development of bind and embedded servers, we developed code generators to generate them basing on IDL and implementation files as input. The IDL file defines the interfaces and exceptions which may be local (defined and used within interface) or global which may be used from any interface. For the code generator developed we used the Coco/R compiler framework [14]. Coco/R takes attributed grammar of new language and generates the scanner and parser for that language. Within the attributed grammar are nested semantic actions, a piece of code in the target language of Coco/R, which defines what to do upon detecting certain syntax in the input of the code generator. Semantic actions are copied in parser code where the execution takes place. Their validity (i.e. syntax correctness) is not checked until the parser is compiled in the target language of the code generator. Semantic actions build the objects list, which is constituted by definitions such as interfaces and global exceptions. Interfaces are further defined by local exceptions, standard operations, oneway operations and attributes. Standard operations may raise global or local exceptions. Only basic data types are supported excluding string and wstring data types. No complex data types are supported such structs, arrays or a combination of them.Instead of giving details of code generation we give briefly the structure of the generated bind and embedded server codes. For each interface, declared in implementation file, a respective bind server class is generated. For the JacORB [15] alternative the class extends that of POA of respective interface defined in the IDL file, while for the Java Web Services alternative it implements the interface generated also based on the interface declared in the IDL file. For each operation, being standard or one-way, is generated a method which serializes the input parameters and sends them through serial port to the microcontroller together with the other data which identifies the operation, according to the AMISP protocol (figure 2). This part is defined for both types of operations.Waiting for response from the microcontroller and de-serializing the parameters from response and assigning their values to method parameters is done only for standard operations. The standard operation might raise exceptions if is declared that might thrown them, and when the Status field of receiving stream from microcontroller is different from zero i.e. it contains an exception code. One-way operations have just input parameters and do not return a response. Attributes are treated as

43


standard operations. For a read-only attribute a single method is generated that returns attribute’s value, for standard attribute are generated two methods. The first method returns the attribute’s value while the second method is used to set the attribute’s value.A similar logic follows the embedded server’s code. It is generated in assembler [16] for 8051 microcontroller. At the beginning are initialized the serial port and serial port interrupt. The microcontroller may do other jobs when there are no requests coming from bind server. It uses interrupts to notify it when a request has been received. When the first byte of request is received, the execution of code is continued to serial port interrupt processing address, where the interrupt is disabled and other bytes are received. After the entire stream is received, it is checked whether a certain interface instance is addressed (as declared in implementation file) and then whether is addressed certain operation within interface previously checked. If the checks are successful it is continued with operation implementation code execution. This code is written by the application developer and is specific for each operation i.e. for which the operation is designated. It is worth to mention that the code generator generates all but implementations of operations which have to be appended by the server designer (application developer). After the operation code is executed the serial port interrupt is activated again for enabling the receipt of new requests and is continued with execution of code where it is left.

5 Code and message footprintsThe first step towards generating the embedded and the bind server is the IDL file writing. It describes the server through interfaces and global exceptions. The interfaces further are consisted of operations, attributes and local exceptions. Global exceptions can be thrown by standard operations of any interface, while local exceptions can be thrown only by operations of interface within which are declared. Second step is implementation file writing. Only interfaces defined in IDL file may figure in implementation file. The interfaces are implemented in microcontroller in the order as they appear in implementation file. For each of them is generated respective bind server.The code footprint is valid only for embedded server. It is the size of embedded server file after compilation i.e. the true calculated size of HEX file. It is measured in bytes. The message footprints are the sizes of messages exchanged between bind and embedded server. The messages contain fields according to AMISP protocol (see fig. 2). They are measured in bytes too. There might be several messages exchanged between bind and embedded server, since interfaces may contain several operations or attributes. Only the fields necessary to identify uniquely operation or response are generated within messages, according to the AMISP protocol. We give an example of server definition in listing 1 and explain messages and their respective sizes. The code size is obtained after server compilation. The message sizes can be obtained from bind server operation implementation where they are created, sent, received and processed.

exception DeviceNotFound {short deviceID;};

interface TempSensor {readonly attribute long minTempRead;readonly attribute long maxTempRead;float readTemp(in short devID) raises(DeviceNotFound);

};

interface Termostat {readonly attribute long minTempAllowed;readonly attribute long maxTempAllowed;exception TempNotAllowed {short deviceID; float temperature;};void setTemperature(in short devID, in float temp) raises (DeviceNotFound, TempNotAllowed);

};a) IDL definition file

implement TempSensor;implement Termostat;b) Implementation file

Listing. 1. An example of server definition in CORBA

44


Table 1 shows the embedded server code size and the details of exchanged messages with bind server. To given request message there may respond several response messages, as is the case for readTemp operation within TempSensor interface or setTemperature operation within Termostat interface. This is since the operation may terminates successfully (with status 0 - OK) or with exception (status > 0). Table 1. Code and exchanged messages details of generated embedded serverIDL file Impl. file Size in bytesListing 1.a Listing 1.b 438Interface implem. Interface ID TempSensor 0 Operation Request Response(s)(get) minTempRead 3 bytes length 7 bytes length 1-st byte: msgLen = 3 1-st byte: msgLen = 7 2-nd byte: intID = 0 2-nd byte: intID = 0 3-rd byte: opID = 0 3-rd byte: opID = 0 4-th to 7-th byte: return(get) maxTempRead 3 bytes length 7 bytes length 1-st byte: msgLen = 3 1-st byte: msgLen = 7 2-nd byte: intID = 0 2-nd byte: intID = 0 3-rd byte: opID = 1 3-rd byte: opID = 1 4-th to 7-th byte: returnreadTemp 3 bytes length 8 bytes length 1-st byte: msgLen = 3 1-st byte: msgLen = 8 2-nd byte: intID = 0 2-nd byte: intID = 0 3-rd byte: opID = 2 3-rd byte: opID = 2 4-th to 5-th byte: devID prm 4-th byte: status = 0 (OK) 5-th to 8-th byte: return 6 bytes length 1-st byte: msgLen = 6 2-nd byte: intID = 0 3-rd byte: opID = 2 4-th byte: status = 1 (DeviceNotFound) 5-th to 6-th byte: deviceID prmInterface implem. Interface ID Termostat 1 Operation Request Response(s)(get) minTempAllowed 3 bytes length 7 bytes length 1-st byte: msgLen = 3 1-st byte: msgLen = 7

45


2-nd byte: intID = 1 2-nd byte: intID = 1 3-rd byte: opID = 0 3-rd byte: opID = 0 4-th to 7-th byte: return(get) maxTempAllowed 3 bytes length 7 bytes length 1-st byte: msgLen = 3 1-st byte: msgLen = 7 2-nd byte: intID = 1 2-nd byte: intID = 1 3-rd byte: opID = 1 3-rd byte: opID = 1 4-th to 7-th byte: returnsetTemperature 9 bytes length 4 bytes length 1-st byte: msgLen = 9 1-st byte: msgLen = 4 2-nd byte: intID = 1 2-nd byte: intID = 1 3-rd byte: opID = 2 3-rd byte: opID = 2 4-th to 5-th byte: devID prm 4-th byte: status = 0 (OK) 6-th to 9-th byte: temp prm 6 bytes length 1-st byte: msgLen = 6 2-nd byte: intID = 1 3-rd byte: opID = 2 4-th byte: status = 1 (DeviceNotFound) 5-th to 6-th byte: deviceID prm 10 bytes length 1-st byte: msgLen = 10 2-nd byte: intID = 1 3-rd byte: opID = 2 4-th byte: status = 129 (TempNotAllowed) 5-th to 6-th byte: deviceID prm 7-th to 10-th byte: temperature prm

Results presented in table 1 outperform respective results in related work for code size (438 bytes) and are comparable in exchanged message sizes. For instance, [23] reports code sizes of 40 up to 70 kbytes (KB) for EXI based Web Services, while for gSOAP based Web Services [19] reports code sizes of 300 kbytes. For exchanged messages [21] reports sizes of the order of 10 bytes, for simple services similar to that presented with listing 1.

6 ConclusionsThe introduction of the bind server results in small embedded server’s code after compilation and small messages exchanged between embedded and bind server. The server definition with listing 1 is explained in detail with table 1, giving at beginning its code size following with messages with description of each byte consisting the message. As shown at the end of section 5, obtained results are satisfactory. The resulting code size is much smaller, while the messages sizes are comparable with those in related work.

46


References[1] I. S. MacKenzie and R. Chu-Wei Phan, The 8051 Microcontroller. Prentice Hall, Third Edition, 1995.[2] K. Hyseni and N. Frasheri, “Integrating embedded devices in SOA through optimized re-targetable code

generation.” In publishing process, 2016. [3] K. Hyseni and N. Frasheri, “Microcontrollers as components of service oriented architectures.” 3rd IEEE

Mediterranean Conference on Embedded Computing (MECO), Budva, Montenegro, 2014.[4] K. Hyseni and N. Frasheri, “Bind server – an adapter from high level to embedded communication in service

oriented architectures.” 9th Annual South-East European Doctoral Student Conference (9th SEE DSC), Thessaloniki, Greece, 2014.

[5] M. Verle, Architecture and Programming of 8051 Microcontrollers. Mikroelektronika, 2008. [6] Atmel Corporation. AT89S8253 Datasheet. 2008.[7] T. R. Szarek, On the use of microcontrollers for data acquisition in an introductory measurements course .

Master’s Thesis, University of Notre Dame, Indiana, 2003. [8] S. R. Saraf and R. M. Holmukhe, “Microcontroller based data acquisition system for electrical motor

vibrations using VB software.” Indian Journal of Computer Science and Engineering (IJCSE), 2011.[9] U. Brinkschulte et al. “Distributed Real-Time Computing for Microcontrollers - the OSA+ Approach.”

Proceedings of the Fifth IEEE International Symposium on Object-Oriented Real-Time Distributed Computing (ISORC '02), 2002.

[10] W3C Working Group Note. Web Services Architecture. 2004. Available: http://www.w3.org/TR/ws-arch/[11] OMG's CORBA web-site. Available: http://www.corba.org/[12] .NET Remoting Overview. Available: http://msdn.microsoft.com/en-us/library/kwdt6w2k(v=vs.71).aspx[13] Modbus Organization. MODBUS over Serial Line. Specification and Implementation Guide V1.02, 2006.

Available: http://www.modbus.org/docs/Modbus_over_serial_line_V1_02.pdf[14] H. Mössenböck, The Compiler Generator Coco/R - User Manual. Johannes Kepler University Linz,

Institite of System Software, Linz, 2010. [15] JacORB web-site. Available: www.jacorb.org/ [16] C51ASM web-site. Available: http://www.atmel.com/tools/C51ASM.aspx [17] A. Sleman and R. Moeller, “Micro SOA Model for Managing and Integrating Wireless Sensor Network

Into IP-Based Networks.” 2010 Second International Conference on Computational Intelligence, Communication Systems and Networks, 2010.

[18] T. Luckenbach et al. “TinyREST – a Protocol for Integrating Sensor Networks into the Internet.” Proceedings of REALWSN, 2005, pp. 101-105.

[19] R. Van Engelen and K. Gallivan, “The gSOAP Toolkit for Web Services and Peer-To-Peer Computing Networks.” 2nd IEEE International Symposium on Cluster Computing and the Grid (CCGrid2002), 2002.

[20] A. Bagnasco et al. “Application of web services to heterogeneous networks of small devices.” Proceedings of the 8th WSEAS International Conference on Automatic Control, Modeling and Simulation, Prague, Czech Republic, 2006.

[21] S. Käbisch et al. “XML-based Web Service Generation for Microcontroller-based Sensor Actor Networks.” 8th IEEE International Workshop on Factory Communication Systems (WFCS), 2010.

[22] S. Käbisch et al. “Efficient and Flexible XML-based Data-Exchange in Microcontroller-based Sensor Actor Networks.” IEEE 24th International Conference on Advanced Information Networking and Applications Workshops, 2010.

[23] S. Käbisch et al. “Optimized XML-based Web Service Generation for Service Communication in Restricted Embedded Environments.” IEEE 16th Conference on Emerging Technologies & Factory Automation (ETFA), 2011.

[24] J. Schneider and T. Kamiya, Efficient XML Interchange (EXI) Format 1.0 (Second Edition). W3C Recommendation 11 February 2014, Available: http://www.w3.org/TR/exi/

[25] Modbus Organization. Modbus Protocol. Available: http://www.modbus.org/specs.php

47


CALL FOR PAPERSInternational Journal of Science, Innovation and New Technology

Department of Statistics and Applied Informatics, University of Tirana,Department of Computer Science, University of Durres

A publication of ASET

The Editors and co-editors of the International Journal of Science, Innovation and New Technology (IJSINT) would like to invite you to consider submitting a manuscript for inclusion in this scholarly journal. Authors should e-mail their manuscripts to [email protected], [email protected] or [email protected]. There is no charge for paper reviews. The normal review time for submissions is 30 days.

Authors submitting a manuscript for publication consideration must guarantee that the document contains the original work of the authors, has not been published elsewhere and is not under publication consideration elsewhere. In addition, submission of a manuscript implies that the author is prepared to pay the publication fee should the manuscript be accepted.

Topics of interest include but are not limited to:

Information Systems

Web 2.0 and Web 3.0; Enterprise 2.0; social media in the enterprise; IS Control and Innovation Systems: Groupware; Financial Systems; Business Intelligence; Global IS and Development; IT Trends and Future Studies; Information Technology and Culture

Data Management

Intelligent Systems; Data Mining; e-Publishing and Digital Libraries; Information Search and Retrieval; Knowledge Management

Platforms

High performance computing systems; Distributed intelligent systems; Mobile systems; Hardware issues; Networks and Data Transmission

e-Science

Natural sciences in digital society; Financial Mathematics and Computation; Statistical Modeling; Mathematical Modeling in Sciences; Applied Mathematics and Computational Science Education; Bioinformatics; Intellectual Property Rights; Decision Support Systems; Simulations;

Communication and Secure Technologies

Internet security; Web services and performance; Cryptography; e-Risk; Information assurance; Mobile social networks; Peer-to-peer social networks

Education and Learning

Virtual Learning Environments and Issues; e-Learning Tools and e-Education

e-Services

E- government; Democracy and the Citizen; e-Administration; Policy Issues; Virtual Communities

Ideas for special theme issues may be submitted to [email protected]: www.ijsint.org

48

http://www.ijsint.org/
