Evaluation of Data Mining Models for Credit Card Default Prediction

This study evaluates the performance of various machine learning models in predicting credit card defaults, including logistic regression, random forest, Naïve Bayes, and support vector machine (SVM). Unlike many studies that address class imbalance using resampling techniques, this research emphasizes model evaluation under real-world conditions where imbalanced datasets are prevalent. The analysis follows a structured process, starting with data collection and preprocessing, followed by model implementation and performance assessment using standard evaluation metrics. Among the tested models, the SVM with an RBF kernel demonstrated the highest accuracy at 82.909%, along with consistent performance across other metrics, establishing it as the most effective approach for predicting credit card defaults in this context. These findings provide valuable insights for financial institutions, highlighting the potential of machine learning models to enhance risk assessment strategies without relying on resampling techniques, thereby maintaining the integrity of real-world data distributions.