A Comparative Study of Decision Tree Algorithms for Class Imbalanced Learning in Credit Card Fraud Detection

Credit card fraud detection along with its inherent property of class imbalance is one of the major challenges faced by the financial institutions. Many classifiers are used for the fraud detection of imbalanced data. Imbalanced data withhold the performance of classifiers by setting up the overall accuracy as a performance measure. This makes the decision to be biased towards the majority class that results in misclassifying the minority class. In today’s revolutionary era of technology most transactions are based on the credit cards that make it more vulnerable to fraud. Credit card data is naturally an imbalanced data and it has been found that most of the classifiers perform poorly on the credit card imbalanced data. Resampling is a technique that deals with the imbalanced data. The aim of this paper is to find the best International Journal of Economics, Commerce and Management, United Kingdom Licensed under Creative Common Page 87 distribution among the classifiers, to get insights of credit card data by random under sampling (RUS) along with feature selection and conclude about a useful model that can measure the credit card fraud risk more efficiently. We applied RUS with feature selection for the family of Decision Tree classifier. Results showed that the given models improved the performance for the Decision Tree classifiers used in a previous study.