Improving Genetic Algorithm Operators for Analyzing Anomalous Behavior of Online Customers

Morteza Talebi

Nowadays, banks and financial institutes have experienced a transition toward electronic banking leading to larger number of customers and huge amount of transactions. As a result of these changes new issues and challenges emerge which necessitate deep investigation of data. One of these challenges is examining customers’ behavior and identifying anomalies so that fraud could be detected. On the other hand, exact investigation of huge transaction data is not possible using conventional methods. Therefore, financial institutes are seeking for novel solutions enabling them to detect anomalous behaviors rapidly. In this paper a novel solution is proposed based on data mining and artificial intelligence. For this purpose, fuzzy expert systems and genetic algorithm are utilized. First off, data collected from transactions are analyzed to extract behavioral features. Afterwards, these features are converted to fuzzy rule base of the proposed fuzzy expert system. Subsequently, the resulted fuzzy rules are optimized using genetic algorithm. In the proposed genetic algorithm, crossover and mutation operators are defined in fuzzy form which significantly improves their efficiency and convergence speed.