DATA MINING PROCESS IN TELECOMMUNICATIONS FOR FRAUD BEHAVIOUR DETECTION

1
DATA MINING PROCESS IN TELECOMMUNICATIONS FOR
FRAUD BEHAVIOUR DETECTION

• Jelena Mamcenko
• Vilnius Gediminas Technical University, Lithuania
  

2
OUTLINE

• Common conseption of fraud
• Fraud defenition
• Classification
• Data Mining technology application
• Data selection
• Requirements
• Proposed method
• Interpretation

3
WHAT IS FRAUD?

• Fraud is specifically used here in the sense of
  fraudulent use of telecommunications services
  rather than fraudulent activities by means of
  telecommunications networks (hacking etc.).

4
Classification of fraud

• Call sell operation
• Premium Rate Service fraud

5
The main ways of commiting fraud are

• Subscriber fraud
• Surfing
• Cloning of handsets
• Calling card fraud
• Misuse of hardware
• Teeing-in other subscribers line by phisicaly
  connecting to the line

6
Data Mining technology in telecoms

• Expert hot lines for computers or other technical
  equipment
• Advice for insurence or juridical questions
• Stock market tips
• Or phone sex

7
Data requirement
8
Call detailed record database
9
(No Transcript)
10
Fields used for fraud detection
11
The proposed mining technique

• Neural clustering
• Demographic clustering

12
(No Transcript)
13
(No Transcript)
14
Cluster Field Characteristics (For Numeric Fields
Only)
15
(No Transcript)
16
Deploying the mining results

• Charge the callers corresponding to these records
  are referring to weekly billing
• Limit the call with premium numbers for that
  caller
• Force these callers to open deposit account
  against the billing will be done
• Confirm the correct subscriber address and the
  employemt status

17
Conclusions

• Data Mining technology application in different
  domain within huge database.
• Analytical methods integration in Data Mining
  with some kind of knowledge could determine and
  minimize fraud events.

18

• ? ? ?