Hyper-Tuned Ensemble Machine Learning Model for Credit Card Fraud Detection
Bholeshwar Prasad Verma, Vijayant Verma, Abhishek Badholia · 2022 International Conference on Inventive Computation Technologies (ICICT) · 2022
Credit card fraud detection is a matter of concern in online financial transactions. This can be done with the help of the Machine Learning (ML) techniques trained by the huge data. Sometimes it is possible that data may suffer from the class imbalance which itself is an issue. To handle this class imbalance issue, this research study proposes a Synthetic Minority Oversampling Technique (SMOTE) with different ML models. After class balancing, an ensemble model is proposed which is hyper-tuned with a metaheuristic approach named Particle Swarm Optimization (PSO). Here, PSO is used to obtain the global optimum solution. The proposed model is evaluated with the performance matrices Precision (P), Recall (R), F1-Measure, and Accuracy (ACC). The model is evaluated and compared with baseline ML models utilized for ensemble considering with and without S MOTE. The results indicate that the proposed model is robust and good enough for the detection of fraudulent transactions.