Adaptive Fraud Detection

  • Authors:
  • Tom Fawcett;Foster Provost

  • Affiliations:
  • Nynex Science and Technology, 400 Westchester Avenue, White Plains, New York 10604.;Nynex Science and Technology, 400 Westchester Avenue, White Plains, New York 10604.

  • Venue:
  • Data Mining and Knowledge Discovery
  • Year:
  • 1997

Quantified Score

Hi-index 0.01

Visualization

Abstract

One method for detecting fraud is to check for suspicious changes inuser behavior. This paper describes the automatic design of userprofiling methods for the purpose of fraud detection, using a seriesof data mining techniques. Specifically, we use a rule-learningprogram to uncover indicators of fraudulent behavior from a largedatabase of customer transactions. Then the indicators are used tocreate a set of monitors, which profile legitimate customer behaviorand indicate anomalies. Finally, the outputs of the monitors are usedas features in a system that learns to combine evidence to generatehigh-confidence alarms. The system has been applied to the problem ofdetecting cellular cloning fraud based on a database of callrecords. Experiments indicate that this automatic approach performsbetter than hand-crafted methods for detecting fraud. Furthermore,this approach can adapt to the changing conditions typical of frauddetection environments.