Predicting Chargeback Fraud Using Data Mining Techniques

Chargeback fraud, commonly known as friendly fraud, is committed intentionally or unintentionally by consumers who make online transactions using their credit card and later dispute the charge with the credit card company. Chargeback fraud can have serious implications depending on the extent of the financial losses suffered by the business, the possible damage to their reputation, and the overall burden it puts on the smooth functioning and trustworthiness of the entire payment system. This paper aims to identify the best algorithm to detect fraudulent activities in financial transactions by applying data mining techniques. The proposed work includes preprocessing, feature engineering and comparative analysis of 6 algorithms on 3 variants of an unbalanced dataset of credit card transactions. From the result it is observed that XGBoost has the best performance, with an accuracy of 98% and F1 score of 84% on the dataset.