Enhancing Credit Card Fraud Detection Using Traditional and Deep Learning Models with Class Imbalance Mitigation

Tahani AlbalawiSamia Dardouri^*

• Shaqra University, Shaqraa, Saudi Arabia

The growing complexity of fraudulent activities poses significant challenges in detecting fraud within financial transactions. This study evaluates the performance of logistic regression, decision tree, and random forest models on a real-world credit card dataset, with a focus on addressing class imbalance and enhancing predictive accuracy. To further improve detection performance, a deep learning model incorporating focal loss was developed. The Synthetic Minority Over Sampling Technique (SMOTE) was employed to mitigate the effects of class imbalance, and hyperparameter tuning was conducted to optimize model configurations. Experimental results indicate that the random forest model achieved the best overall performance, with an accuracy of 99.95%, F1 score of 0.8256, and ROC-AUC of 0.9759. Meanwhile, the deep learning model provided a balanced output with the highest precision, demonstrating its potential in minimizing false positives. A key novelty of this work lies in the integration of focal loss within the deep learning framework, which enables the model to focus on hard-to-classify fraudulent transactions. Moreover, unlike many prior studies that evaluate models only on the Kaggle dataset, we validate our approach on both the Kaggle credit card dataset and the PaySim synthetic mobile money dataset, thereby demonstrating robustness and cross-domain generalizability. The findings highlight the effectiveness of combining data preprocessing, resampling techniques, and model optimization for robust credit card fraud detection.