ANALYSIS OF CREDIT CARD FRAUD DETECTION TECHNIQUES

Abstract

The Credit Card Fraud Detection project is based on machine learning techniques. As the numbers  of  frauds  are increasing day by day, this  paper  focuses  on  finding  fraud transactions using different algorithms and measuring their accuracies. The data set used in the project consist of information of European Card- holders for two days and the model will be trained using that data with the split to train   and test the models. There are various  methods  to  perform the analysis. (Clustering, Outlier Detection, Bayesian Networks etc.) In this paper, the author has compared the capability of three machine learning algorithms(Logistic regression, Random Forest, Decision tree) to find a transaction whether it is genuine or fraud. The fraud and genuine is identified on the basis binary number system as a parameter, if the value rendered is 1 then  it is a fraud and if 0 then it is  a  genuine  transaction.