Ensemble Learning based Credit Card Fraud Detection System

Abstract

In this digital era cashless transactions are very much common. Simplicity and convenience of online and credit card transaction has raised the popularity and revenue of e-commerce site. Every coin has two side, as the credit card transaction growing day by day, so does the number of fraudulent transactions. Credit card fraud can be detected by evaluating client spending history from prior transaction data and detecting variation in their spending behavior. With the technological advancement rule based techniques are not so effective in detecting credit card fraud from huge datasets. Now banks and credit card firms are using various classification techniques like decision trees, logistic regression, Random forest for this purpose, but one of the biggest challenge for computational intelligence technologies in detecting credit card fraud are class imbalance and concept drift problem. To overcome these problem effectively hybrid approaches are gaining momentum. In this paper ensem-ble learning technique is employed by parallel applying Decision Tree, Logistic Regression, Naive base classifiers and then best output is selected through hard voting. The experimental results conclusively proven that accuracy of Ensemble learning with hard voting achieve better accuracy as compared to other classifiers in detecting credit card fraud.