84 Fraud Detection of Electricity Consumers: Data-Mining Techniques as a Case in Point
-
Published:2008
Download citation file:
Fraud is a problem that affects many business areas. Energy companies, for instance, have to deal with frauds perpetrated by their consumers. An electricity company located in Brazil showed that its income has been heavily affected in millions of dollars per month because of frauds. Trying to minimize such a problem for electricity companies, this paper proposes an approach based on the KDD process and data mining in order to identify consumers who are most likely to defraud an electricity delivery system. We define a measure that is the accumulated monthly differences of power consumption figures between current and previous years. This measure is calculated for each electricity consumer. Consumers who have an accumulated score above a specified threshold are considered for inspection. After prototyping the approach and adjusting a threshold value, an evaluation showed that our approach provides correct answers for 80% of commercial consumers grouped by business type.