A Survey of Data Mining Techniques in Fraud Detection

Abstract

Data mining is extracting knowledge from huge amounts of information. There are many number of data mining techniques in regression,classification,clustering and association. Particularly the data mining techniques from classification task in current practice are implemented in order to compare the results. Millions of credit cards are issued by credit card provider.There are many cases where credit cards are issued to the bad customers which leads to financial crisis.Even though detecting the credit card fraud is common problem but still the issues are not being solved. According to the recent study, many surveys are done until now. Many papers show the results on implementation of their methods. We have studied many research works published, but the last survey paper is in 2012. This report is "two-pronged". The first part is to do the systematic survey of recent literature to clearly understand the state of art theoretically also to give a overall review of different techniques in detecting fraud in credit card processing until 2016.The second part is to the practical implementation where we apply four fraud detection methods (Support Vector Machine Algorithm,Naive Bayes,Decision tree and K-Nearest Neighbor) on German data set. The results the results are compared based on certain metrics.From many algorithms only few data mining algorithms and techniques are considered and implemented in this research. The main goal of this study is helpful for credit card providers in selecting an accurate result to their issue and for the researchers to get an compete view of literature in this field and also to get a practical knowledge on classifying the fraud transactions can give us more clearer view which is different from others.