A soft voting ensemble learning approach for credit card fraud detection

Area of Science:

• Computer Science
• Data Science
• Financial Technology

Background:

• Credit card fraud is a growing global problem, costing billions annually.
• Class imbalance in transaction data hinders accurate fraud detection.
• Existing machine learning methods struggle with accuracy on imbalanced datasets.

Purpose of the Study:

• To propose a novel soft voting ensemble learning approach for credit card fraud detection.
• To address the challenge of class imbalance in financial transaction data.
• To enhance the accuracy and reduce misclassification costs in fraud detection systems.

Main Methods:

• Developed a soft voting ensemble learning model for credit card fraud detection.
• Evaluated the model against various sampling techniques (oversampling, undersampling, hybrid).
• Compared ensemble classifiers with and without sampling methods on imbalanced data.

Main Results:

• The proposed soft voting ensemble approach significantly outperformed individual classifiers.
• Achieved high performance metrics: precision 0.9870, recall 0.9694, F1-score 0.8764, AUROC 0.9936.
• Demonstrated effectiveness in mitigating the class imbalance problem.

Conclusions:

• Soft voting ensemble learning is a robust method for credit card fraud detection on imbalanced data.
• The proposed methodology offers a reliable solution for financial institutions to combat fraud.
• This approach enhances the integrity of payment systems by improving fraud identification accuracy.