evolving classification of the behaviour of computer users...97,61% number of commands subsequences:...
TRANSCRIPT
![Page 1: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/1.jpg)
Evolving Classification ofthe Behaviour ofComputer Users
Jose Antonio IglesiasCAOS Research Group.
Carlos III University of Madrid (Spain)[email protected]
Comunication System Department. Lancaster Univeresity (UK)[email protected]
Plamen Angelov
![Page 2: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/2.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior 1
![Page 3: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/3.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior
OutlineOutline
.
•From RoboCup …
•… to UNIX User Classification
• Creating a Profile Model
• Evolving classification
• Results
• Conclusions
9
![Page 4: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/4.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior
FromFrom RoboCupRoboCup……
1
Opponent behavior Modelling / Classification(Environment: soccer simulation domain)
OpponentModeling
PatternRecognition
Off-Line Analysis
No-PatternLogFile
PatternLogFile
Base Estrategy
Pattern
RecognizedPatterns
On-Line ComparingMethod
PatternDetection
On-Line Detection
Environment Information Advices to
Players
RoboCup Soccer Server
PatternRecognized
![Page 5: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/5.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior
FromFrom RoboCupRoboCup……
1
BehaviorBehavior ClassificationClassification
Behavior as sequence of actions
Pass1to2, Dribble2, Pass2to10, Goal10, Pass1to2, Dribble2, Pass2to10, Goal10,
Pass5to6, Pass6to3, Foul5, Pass5to6, Pass6to3, Foul5, MissedShot3MissedShot3……
![Page 6: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/6.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior
…… toto UNIX UNIX UsersUsers CommandsCommands
1
BehaviorBehavior ClassificationClassification
Behavior as sequence of commands
*SS**SS*, , postnewspostnews, , enscriptenscript, , fgfg, , enscriptenscript, , cdcd, , *ES*, *SS*,*ES*, *SS*, rloginrlogin……
*SS*,*SS*, date, ls, emacs, rm, ls, date ... date, ls, emacs, rm, ls, date ...
![Page 7: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/7.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior
OutlineOutline
.
•From RoboCup …
•… to UNIX User Classification
• Creating a Profile Model
• Evolving classification
• Results
• Conclusions
9
![Page 8: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/8.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior
SequenceSequence ofof commandscommands �� ProfileProfile ModelModel
.
ImportantImportant : : sequentialitysequentiality
TrieTrie ((reretrietrievalval)) data data structurestructure::
SpecialSpecial searchsearch treetree usedused forfor storingstoringelementselements andand itsits prefixesprefixes..
EveryEvery nodenode: : – represents a command– stores useful information (times appeared,…)
9
Behavior as sequence of commands
![Page 9: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/9.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior
AnAn exampleexample trietrie
Sequence to insert initially in the trie: {ls � date � ls � dat � cat}
pwd
vi
pwd
vi
pwd
ls
Sequence
10
SequenceSequence ofof commandscommands �� ProfileProfile ModelModel
![Page 10: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/10.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior
AnAn exampleexample trietrie
Sequence to insert initially in the trie: {ls � date � ls � date � cat}
pwd
vi
pwd
vi
pwd
ls
Sequence
10
SequenceSequence ofof commandscommands �� ProfileProfile ModelModel
SubSub--sequencesequence lengthlength: 3: 3
{ls � date � ls � date � cat}
{ls � date � ls}
Sub-sequences to insert in the trie:
and {date � ls � date } and {ls � date � cat}
![Page 11: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/11.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior
Sub-sequences to insert in the trie:
Root
11
{ls � date � ls} and {date � ls � date} and {ls � date � cat}
AnAn exampleexample trietrieSequenceSequence ofof commandscommands �� ProfileProfile ModelModel
![Page 12: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/12.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior
Root
ls [1]
date [1]
ls [1]
11
Sub-sequences to insert in the trie: {ls � date � ls} and {date � ls � date} and {ls � date � cat}
AnAn exampleexample trietrieSequenceSequence ofof commandscommands �� ProfileProfile ModelModel
![Page 13: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/13.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior
Root
ls [1]
date [1]
ls [1]
date [1]
ls [1]
11
Sub-sequences to insert in the trie: {ls � date � ls} and {date � ls � date} and {ls � date � cat}
AnAn exampleexample trietrieSequenceSequence ofof commandscommands �� ProfileProfile ModelModel
![Page 14: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/14.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior 11
Sub-sequences to insert in the trie: {ls � date � ls} and {date � ls � date} and {ls � date � cat}
Root
ls [2]
date [1]
ls [1]
date [1]
ls [1]
AnAn exampleexample trietrieSequenceSequence ofof commandscommands �� ProfileProfile ModelModel
![Page 15: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/15.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior 11
Sub-sequences to insert in the trie: {ls � date � ls} and {date � ls � date} and {ls � date � cat}
Root
ls [2]
date [1]
ls [1]
date [2]
ls [2]
date [1]
AnAn exampleexample trietrieSequenceSequence ofof commandscommands �� ProfileProfile ModelModel
![Page 16: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/16.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior
Root
ls [4]
date [3]
cat [1]
date [4]
ls [2]
date [1]ls [1]
cat [1]
11
pwd
vi
pwd
vi
pwd
ls
Sub-sequences to insert in the trie: {ls � date � ls} and {date � ls � date} and {ls � date � cat}
cat [1]
AnAn exampleexample trietrieSequenceSequence ofof commandscommands �� ProfileProfile ModelModel
![Page 17: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/17.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior
Evaluate the relation/dependence between an element and its prefix
Two approaches:– Frequency-based method.� Statistical dependence method.
Our approach: Statistical Value used: Chi-square value.This value is stored in every node of the trie
12
EvaluatingEvaluating DependencesDependencesSequenceSequence ofof commandscommands �� ProfileProfile ModelModel
![Page 18: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/18.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior 13
EvaluatingEvaluating DependencesDependencesSequenceSequence ofof commandscommands �� ProfileProfile ModelModel
Support or
Relative Frequency
(of a subseq)=
Number of times the subsequence has been inserted into the tre.
Total number of subsequencesof equal size inserted
![Page 19: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/19.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior 13
EvaluatingEvaluating DependencesDependencesSequenceSequence ofof commandscommands �� ProfileProfile ModelModel
Root
ls [4] [0.44]
date [3] [0.5]
cat [1] [0.33]
date [4] [0.44]
ls [2] [0.33]
date [1] [0.33]ls [1] [0.33]
cat [1] [0.11]
cat [1] [0.16]
![Page 20: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/20.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior
Behavior � Distribution of relevant events.
ls ls-date ls-date-ls ls-date-cat date date-ls date-cat date-ls-date cat
SequenceSequence ofof commandscommands �� ProfileProfile ModelModel
![Page 21: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/21.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior
OutlineOutline
.
•From RoboCup …
•… to UNIX User Classification
• Creating a Profile Model
• Evolving classification
• Results
• Conclusions
9
![Page 22: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/22.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior
Profile a user is a difficult task for different aspects:
1. Human behavior is usually erratic
2. Humans behave differently because of a change in their goals.
ClassificationClassification……
![Page 23: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/23.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior
How to accurately profile a user while his/herbehavior changes constantly ???
ClassificationClassification……
Using Evolving Fuzzy Systems !!
![Page 24: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/24.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior
Evolving fuzzy systems (EFS) can be defined as self-developing, self-learning fuzzy rule-based or neuro-fuzzy
systems that have both their parameters but also theirstructure self-adapting on-line.
EvolvingEvolving ClassificationClassification……
They are usually associated with streaming data andon-line (often real-time) modes of operation.
![Page 25: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/25.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior
EvolvingEvolving ClassificationClassification ofof UNIX UNIX UsersUsers
![Page 26: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/26.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior
EvolvingEvolving ClassificationClassification ofof UNIX UNIX UsersUsers
![Page 27: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/27.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior
ArchitectureArchitecture ofof thethe EvolvingEvolving ClassifierClassifierFRB Classifier:
Ri = IF (x1 is around x1i*) AND (x2 is around x2
i*) AND … (xn is around xn
i*)
THEN (Labeli)
X = [x1, x2, … xn] denotes the vector of features .
L = Li*; i*= arg min(DistCosine)i=1
N
![Page 28: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/28.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior
EvolvingEvolving ClassificationClassification ofof UNIX UNIX UsersUsers
0,9----0,5------0,3
0.80.3----0.2--0,1----0,2
0,30,80,2----0,10,6
0,10,20,750,30,5
asfdmvmailrmemacsvicatls-datedatels
![Page 29: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/29.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior
ArchitectureArchitecture ofof thethe EvolvingEvolving ClassifierClassifier
ls
date
![Page 30: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/30.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior
ArchitectureArchitecture ofof thethe EvolvingEvolving ClassifierClassifierFor each User…
1. Recursive Calculation of Potential of Data Vector (x) of a User.
2. Update Potentials of the F.P. with the new data (x)
3. If (Pot (x)) > Max Pot(F.Pi))
Insert New F.P.
1. Classify x in a Cluster (using Cosine Distance)Add x as a new “point” in the Cluster.
1. ¿Necesary Remove any F.P.?
Calculate µ for each F.P.
If µi > e-1 � Remove F.P.
![Page 31: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/31.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior
OutlineOutline
.
•From RoboCup …
•… to UNIX User Classification
• Creating a Profile Model
• Evolving classification
• Results
• Conclusions
9
![Page 32: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/32.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior
168 Users, 4 different user groups:
1. Novice – programmers 55 Users
2. Experienced-programmers 36 Users
3. Computer-Scientists 52 Users
4. Non-Programmers 25 Users
168 Users
ResultsResults -- Experimental Experimental DesignDesign
Commands per user: from 600 to 3000
![Page 33: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/33.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior
ResultsResults -- Experimental Experimental DesignDesign
After processing the users commands:
Different commands per group:1. Novice – programmers (25614)2. Experienced-programmers (43049) 3. Computer-Scientists (66490)4. Non-Programmers (10572)
Different commands in the 4 groups: 135317
![Page 34: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/34.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior
ResultsResults--UsingUsing 10Fold Cross 10Fold Cross ValidationValidationNumber of F.P. per group:
FP Nov = 2FP Exp = 2FP Sci = 1FP Non = 1
Number of F.P. using Cross Validation:
Novices: 3 / 2 / 7 / 2 / 2 / 2 / 3 / 4 / 2 / 2 /Experiences: 2 / 3 / 3 / 2 / 2 / 2 / 2 / 2 / 2 / 2 /Sciences: 2 / 2 / 1 / 2 / 1 / 1 / 1 / 1 / 3 / 3 /Non: 2 / 2 / 1 / 2 / 2 / 2 / 2 / 2 / 2 / 2 /
![Page 35: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/35.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior
ResultsResults--UsingUsing 10Fold Cross 10Fold Cross ValidationValidation
25000non
05200sci
00360exp
00055nov
nonsciexpnov
----------
10NN5NN3NN1NNC4.5
Evolving System…
ClassificationRate:
100%
Number of commands subsequences: 135317
![Page 36: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/36.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior
ResultsResults--UsingUsing 10Fold Cross 10Fold Cross ValidationValidation
24010non
05011sci
00351exp
00055nov
nonsciexpnov
----------
10NN5NN3NN1NNC4.5
Evolving System…
Similar number of F.P.
ClassificationRate:
97,61%
Number of commands subsequences: 45930Support > 0.005
![Page 37: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/37.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior
ResultsResults--UsingUsing 10Fold Cross 10Fold Cross ValidationValidation
24010non
05011sci
10350exp
10054nov
nonsciexpnov
----------
10NN5NN3NN1NNC4.5
Evolving System…
ClassificationRate:
97,02%
Number of commands subsequences: 26108Support > 0.01
![Page 38: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/38.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior
ResultsResults--UsingUsing 10Fold Cross 10Fold Cross ValidationValidation
22030non
236122sci
10314exp
10450nov
nonsciexpnov
40,47%36,90%50,00%54,16%73,80%
10NN5NN3NN1NNC4.5
Evolving System…
ClassificationRate:
82,74%
Number of commands subsequences: 6210Support > 0.05
![Page 39: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/39.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior
ResultsResults--UsingUsing 10Fold Cross 10Fold Cross ValidationValidation
16054non
63556sci
50364exp
20350nov
nonsciexpnov
52,97%53,57%53,57%44,64%73,80%
10NN5NN3NN1NNC4.5
Evolving System…
ClassificationRate:
81,54%
Number of commands subsequences: 3531Support > 0.1
![Page 40: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/40.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior
ResultsResults--UsingUsing 10Fold Cross 10Fold Cross ValidationValidation
71215non
28141sci
201221exp
311239nov
nonsciexpnov
30,95%30,95%30,35%40,47%52,97%
10NN5NN3NN1NNC4.5
Evolving System…
ClassificationRate:
39,28%
Number of commands subsequences: 714Support > 0.5
![Page 41: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/41.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior
ResultsResults--UsingUsing 10Fold Cross 10Fold Cross ValidationValidation
22030non
236122sci
10314exp
10450nov
nonsciexpnov
40,47%36,90%50,00%54,16%73,80%
10NN5NN3NN1NNC4.5
Evolving System…
ClassificationRate:
82,74%
Number of commands subsequences: 6210Support > 0.05
![Page 42: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/42.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior
ConclusionsConclusions
The proposed Evolving System Classifier…
• Works with large data stream of data.
• Good results with large data stream of data.
• Fast
• Stores data very reduced
• There is no thresholds
• Does not requiere pre-training
• Starts “from scratch”
• ….
![Page 43: Evolving Classification of the Behaviour of Computer Users...97,61% Number of commands subsequences: 45930 Support > 0.005 Jose Antonio Iglesias and Plamen Angelov Evolving Classification](https://reader036.vdocuments.site/reader036/viewer/2022070909/5f8ec72dbb9809509f442683/html5/thumbnails/43.jpg)
Jose Antonio Iglesias and Plamen Angelov
Evolving Classification of Computer Users Behavior
Thanks!