Determining an Appropriate Weight attribute in Fraud Call Rate Data Using Case Based Reasoning

Abstract

Fraud cases are significantly causing huge revenue losses in telecommunication companies around the world. Although previous cases are very important data in dealing with fraud patterns, there are variations in the dataset of different fraud case scenarios which in turns need specific detection system without necessarily involving the domain expert directly. This paper investigates the appropriate weight values for attributes using fraud Call Rate Data that is based on Artificial Intelligence technique (Case Based Reasoning) with a meaningful confidence in telecommunication data. The experimental result on the fraud data reports that the weight for all attribute used in this study needs to be set at 0.9 in order to get the best performance of 98.33%.