Credit scoring is essential in the financial sector, where accurate predictions help mitigate financial risks for lending institutions and ensure fair access to credit for applicants. The traditional credit scoring models often struggle with balancing predictive accuracy and interpretability, two crucial factors in real-world applications. In this paper, we present a novel model that integrates genetic algorithms for feature selection with ensemble learning techniques to enhance the performance. The model addresses data imbalance using ADASYN oversampling and employs stacking strategies to benefit from the strengths of multiple base classifiers. Interpretability is provided through SHAP, which evaluates the contributions of base learners within the ensemble to ensure transparency. The proposed model was evaluated on three credit scoring datasets, achieving 86.00% accuracy on the german dataset, 92.03% on the australian dataset, and 92.39% on the polish dataset, outperforming benchmark classifiers. While deep learning models showed strong performance on the larger polish dataset, our approach delivers competitive results with superior interpretability and transparency. This study highlights the importance of combining feature selection, balancing techniques, ensemble methods, and interpretability for a trustworthy credit scoring model.