A Comparative Study of Credit Card Fraud Detection Using the Combination of Machine Learning Techniques with Data Imbalance Solution

Due to the rapid spread of fraud and cybersecurity risks in digital economy, fraud detection stands as a prime issue of modern technology. However, the analysis of fraud cases is computationally difficult because, fraud cases conjure less than 0.2% of the transactions. Thus to figure out the best classification technique to use for fraud detection, this paper has conducted a thorough experimentation of Machine Learning (ML) techniques. It has implemented six ML techniques i.e. Logistic Regression (LR), Support Vector Machine (SVM), Naive Bayes (NB), Random forest (RF), Decision Tree (DT), and K-nearest neighbour (KNN) classifiers to detect credit card fraud. The investigation used five type of datasets i.e. imbalanced data, Under Sampled (US) data, Over Sampled (OS) data, sampled data using Synthetic Minority Over Sampling Technique (SMOTE) and Adaptive Synthetic Sampling Method for Imbalanced Data (ADASYN). The best combination of these classification approaches is selected based on five performance evaluation criteria i.e. Accuracy, Area Under the Curve (AUC), Precision, Recall score and fl-score. After evaluation of the classifiers it has showed that among 30 different classification approaches, RF classifier with over sampling (OS) technique was found to be the best approach in terms of all the performance criteria. It showed 99.99 % accurate and precise results with 99.99 % AUC, fl-score and 100 % Recall rate. Our choosen approach has obtained the highest accuracy over other studies on the same dataset. The banking sector as well as other financial institutions might use this suggested machine learning based combination approach to minimize (debit/credit card) frauds.