Example Applications of Rough Sets Theory A Survey

1
Example Applications of Rough Sets Theory A
Survey

• Christopher Chretien
• Laurentian University
• Sudbury, Ontario
• Canada
• October 2002

2
Introduction

• Research on the application of Rough Sets Theory
• Discovering possible areas of application
• Further understanding of Rough Sets Theory usage

3
References

• Lixiang Shen, Francis E. H. Tay, Liangsheng Qu
  and Yudi Shen (2000), Fault Diagnosis using Rough
  Sets Theory , Computers in Industry, vol. 43,
  Issue 1, 1 August 2000, pp.61-72.,
• URLwww.geocities.com/roughset/Fault_diagnos
  is_using_rough_sets_theory.pdf
• Israel E. Chen-Jimenez, Andrew Kornecki, Janusz
  Zalewski, Software Safety Analysis Using Rough
  Sets,
• URLhttp//www-ece.engr.ucf.edu/jza/classes
  /6885/rough.ps
• Francis E. H. Tay and Lixiang Shen (2002),
  Economic and Financial Prediction using Rough
  Sets Model , European Journal of Operational
  Research 141, pp.643-661, URLhttp//www.geocities
  .com/roughset/EJOR.pdf
• Pawan Lingras (2001), Unsupervised Rough Set
  Classification Using GAs Journal of Intelligent
  Information Systems, 16, 215228, found on
  CiteSeer,
• URLhttp//citeseer.nj.nec.com/cs
• Rapp, S., Jessen, M. and Dogil, G. (1994). Using
  Rough Sets Theory to Predict German Word Stress.
  in Nebel, B. and Dreschler-Fischer, L. (Eds.)
  KI-94 Advances in Artificial Intelligence,
  Lecture Notes in Artificial Intelligence 861,
  Springer-Verlag, URLwww.ims.uni-stuttgart.de/rap
  p/ki94full.ps

4
Fault Diagnosis using Rough Sets Theory

• Diagnosis of a valve fault for a multi-cylinder
  diesel engine
• Rough Sets Theory is used to analyze the decision
  table composed of attributes extracted from the
  vibration signals

5
Fault Diagnosis using Rough Sets Theory

• 4 states are studied among the signal
  characteristics
• Normal state
• Intake valve clearance is too small
• Intake valve clearance is too large
• Exhaust valve clearance is too large

6
Fault Diagnosis using Rough Sets Theory

• 3 sampling points selected to collect vibration
  signals
• 1st cylinder head
• 2nd cylinder head
• centre of the piston stroke on the surface of the
  cylinder block

7
Fault Diagnosis using Rough Sets Theory
8
Fault Diagnosis using Rough Sets Theory
9
Fault Diagnosis using Rough Sets Theory
10
Fault Diagnosis using Rough Sets Theory

• 6 attributes
• Frequency domain attributes IF, CG
• Time domain attributes IT, s, Dx, a4
• 18 attributes for decision table
• 1 decision attribute with 4 possible values based
  on states

11
Software Safety Analysis using Rough Sets

• Investigates the safety aspects of computer
  software in safety-critical applications
• Assessment of software safety using qualitative
  evaluations

12
Software Safety Analysis using Rough Sets

• Use of checklists to collect data on software
  quality
• Waterfall model
• Project Planning
• Specification of requirements
• Design
• Implementation and integration
• Verification and validation
• Operation and maintenance

13
Software Safety Analysis using Rough Sets
14
Software Safety Analysis using Rough Sets
15
Software Safety Analysis using Rough Sets

• 8 student teams developing safety-related
  software
• Device control over the internet
• Elevator controller
• Air traffic control system
• System satellite control system

16
Software Safety Analysis using Rough Sets

• 150 questions about the first 5 phases of the
  waterfall model
• Overall safety level for 6 of the 8 projects was
  around 60

17
Economic and Financial Prediction using Rough
Sets Model

• Applications of Rough Sets model in economic and
  financial prediction
• Emphasis on main areas of business failure
  prediction, database marketing and financial
  investment

18
Economic and Financial Prediction using Rough
Sets Model

• Business failure prediction
• ETEVA
• Database Marketing
• Financial Investment
• TSE

19
Economic and Financial Prediction using Rough
Sets Model
20
Economic and Financial Prediction using Rough
Sets Model
21
Using Rough Set Theory to Predict German Word
Stress

• Prediction of German word stress by extracting
  symbolic rules from sample data
• Symbolic rules are induced with a machine
  learning approach based on Rough Sets Theory

22
Using Rough Set Theory to Predict German Word
Stress

• Variable Precision Rough Sets Model
• An elementary class belongs to RßX iff a (100 -
  ß) majority of its elements belongs to X
• An elementary class does not belong to URßX iff a
  (100 - ß) majority of its elements does not
  belong to X

23
Using Rough Set Theory to Predict German Word
Stress

• Corpus
• Monomorphemic words
• At least 2 non-schwa syllables
• Nouns
• 242 words

24
Using Rough Set Theory to Predict German Word
Stress

• Attributes Typ, Onset, Hoeche, Laenge, Spannung,
  Coda
• 36 attributes in total
• Attributes aligned from right to left
• Decision attribute with possible values of final,
  penult and antepenult

25
Using Rough Set Theory to Predict German Word
Stress

• 1st experiment
• Stress assignment operates from right to left
• 2nd experiment
• Estimate predictive accuracy
• 3rd experiment
• Remove length information

26
Unsupervised Rough Set Classification using GAs

• Rough Set classification using Genetic Algorithms
• Highway classification based on predominant usage

27
Unsupervised Rough Set Classification using GAs

• Applications of GAs
• Job shop scheduling
• Training neural nets
• Image feature extraction
• Image feature identification

28
Unsupervised Rough Set Classification using GAs
29
Unsupervised Rough Set Classification using GAs
30
Unsupervised Rough Set Classification using GAs
31
Unsupervised Rough Set Classification using GAs
32
Unsupervised Rough Set Classification using GAs

• Rough Set classification scheme
• Both uh and uk are in the same lower
  approximation A(Xi).
• Object uh is in a lower approximation and uk is
  in the corresponding upper approximation UA(Xi)
• Both uh and uk are in the same upper approximation

33
Unsupervised Rough Set Classification using GAs

• Total error of rough set classification is the
  weighted sum of these errors

34
Unsupervised Rough Set Classification using GAs

• Rough classification of highways
• PTC sites
• Roads classified on the basis of trip purposes
  and trip length characteristics
• Classes commuter, business, long distance and
  recreational highways
• Traffic patterns hourly, daily, monthly

35
Unsupervised Rough Set Classification using GAs

• Experiment
• 264 monthly traffic patterns on Alberta highways
  (1987-1991)
• Rough genome consisted of 264 genes
• Classes commuter/business, long distance,
  recreational

36
Conclusion

• Triggering a better understanding of Rough Sets
  Theory
• Opening eyes to different fields of application