Analysis of Oversampling and Ensemble Learning Methods for Credit Card Fraud Detection
Ankit Kumar Singh, Rohit Bhaskar Rao Gurijala, Utkarsh Kumar Rai, Anupam Kumar, Basant Agarwal, Ashish Sharma · 2023
The frequent and fast-growing online transactions have resulted in an increase in security measures taken to protect the data of transactions and customers. Companies have been consulting various data scientists in order to provide a better framework for fraud detection. Fraudsters are in continuous pursuit of valuable data of customers and committing illegal actions. Hence, it is important for data scientists and machine learning experts to find newer frameworks and models for fraud detection. This chapter analyses the ensemble models of prominent machine learning algorithms. The ensembles were created using various algorithms such as Logistic Regression, Random Forest, KNN (K Nearest Neighbors), Naive Bayes, and SVM (Support Vector Machine). The mentioned algorithms were observed in various groups using ensemble learning. The proposed approach of the ensemble of Logistic Regression, Random Forest, and KNN performed the best among the other ensembles with a Recall of 93% and F1 score of 94%. Various oversampling techniques such as ADASYN, SMOTE, and Random Over-Sampler were applied as there were only 492 fraudulent transactions out of 284,807 transactions. The Random Over-Sampler performed the best among the three oversampling algorithms used.