Detecting Fraudulent Credit Card Transactions Using Outlier Detection

摘要

Credit card frauds transactions are becoming more frequent day by day and it is becoming more difficult for humans to analyse fraudulent transactions by analysing transaction hence it has become necessary for humans to develop an intelligent system to determine fraudulent transactions. The technique we applied to determine fraudulent transactions are anomaly detection (Outlier detection). Several intelligent algorithms can be used in this context for anomaly detection (outlier detection), In this paper we implemented Decision Tree algorithm, Random Forest and Neural Networks to determine which algorithm is best fit in terms of time taken and accuracy. We were able to formulate results of 284,407 transactions over a period of two days in September 2013. We were able to identify that three models are almost equal when it comes to accuracy but random forest is more precise.