Enhanced Credit Default Prediction Performance by Combining Genetic Algorithm Feature Selection and Heterogeneous Ensemble

In recent years, artificial intelligence-based credit evaluation models have been developed to assist financial institutions in identifying borrowers who may potentially default on their loans. Existing research has shown that heterogeneous ensemble methods, which integrate multiple algorithms to produce a new model, outperform traditional credit scoring techniques. This study seeks to enhance credit default prediction performance by utilizing a heterogeneous ensemble stacking model that combines multiple classifiers with an XGBoost meta-learner and is optimized using genetic algorithm-based feature selection. Three credit datasets Australia, Germany, and PPDai were employed to facilitate the analysis. The results indicate that the German credit dataset achieves the best performance, with an accuracy of 0.9408 and the lowest error rate of 0.0553. The proposed model also consistently outperforms other comparison models. Beyond performance improvements, this research provides a more stable, scalable, and interpretable credit scoring framework, offering practical value for financial institutions seeking enhanced risk management, reduced default probability, and greater transparency in automated credit decision-making. The comparative findings demonstrate that the method proposed in this study is both effective and impactful for modern financial applications.