telecommunication case modelling – call center c.chudzian, j.granat, w.traczyk decision support...
TRANSCRIPT
Telecommunication Case Modelling –Call Center
C.Chudzian, J.Granat, W.Traczyk
Decision Support Systems Laboratory
National Institute of Telecommunications
Warsaw, Poland
Data Mining in Practice, Dortmund, 18th February 2003
2
The goal
Selecting prospective clients for targeting a marketing campaign based on the existing data (call center data, billing data and others).
Data Mining in Practice, Dortmund, 18th February 2003
3
Model of the call center
List ofthe clients
Call center
Outgoingscript
Incomingscript
Data baseClients
Reports
Data Mining in Practice, Dortmund, 18th February 2003
4
The data is distributed
Call center
Switch
Target group of clients
Analysis
Data Mining in Practice, Dortmund, 18th February 2003
5
Process of building a table for data mining
Billing
Servicedata
Clientdata
SS7
CallCenter
Newattributes
Dataaggregation
Synchronizationin time
Data miningtable
Newattributes
Dataaggregation
Newattributes
Dataaggregation
Data Mining in Practice, Dortmund, 18th February 2003
6
Billing
Call Center
Classification of clients
Switch data
No
List of clients
YES
Data Mining in Practice, Dortmund, 18th February 2003
7
SAS Enterprise Miner
Data Mining in Practice, Dortmund, 18th February 2003
8
MiningMart requirements
Datasource
Concept1
Concept..
ConceptOperator Operator
data preparedfor mining
Column set
Table View
Relationaldata model
Column set
Table View
Relationaldata model
Column set
Table View
Relationaldata model
Datapreparedformining
Data Mining in Practice, Dortmund, 18th February 2003
9
The set of operators
Client Callclass
Time .........CDR
A1 A2 .........Client
Data Mining in Practice, Dortmund, 18th February 2003
10
Client Features IIIClient Features II
Preprocessing (part I)
Process details for all clients
UnSegment
Process details fora client
SpecifiedStatistics
Segmentation
Features (I,II,III) for all clients
Client Features I
RowSelectionByQuerySpecifiedStatistics
TimeIntervalManualDiscretizationRowSelectionByQuery
SpecifiedStatistics
UnSegment UnSegment
JoinByKey
Features I for all clients Features II for all clients Features III for all clients
Data Mining in Practice, Dortmund, 18th February 2003
11
Preprocessing (part II)
Features(I,II,III) for all clients
GenericFeatureConstructionAll features for all
clients
Call Center Data
Service users
UnionByKey Service info
GenericFeatureConstruction
Service info (switch prefered)
UnionByKey
Mining Concept
Data Mining in Practice, Dortmund, 18th February 2003
12
NIT case - HCI view
Data Mining in Practice, Dortmund, 18th February 2003
13
Business data
Data Mining in Practice, Dortmund, 18th February 2003
14
NIT case – preprocessing steps
Data Mining in Practice, Dortmund, 18th February 2003
15
Conclusions
Conceptual modeling improves: • the understanding of the data preparation
process
• maintenance of the data preparation process
• Knowledge transfer for other peopleWe do not need to use programming language
