This paper introduces a new machine learning model for credit risk evaluation in financial systems based on explainable artificial intelligence (XAI) methods combined with sophisticated hyperparameter tuning. We use a SHAP-augmented LightGBM model optimized by Optuna to enhance not only the explainability but also the predictive power of credit scores. By handling the class imbalance using SMOTE and picking the most relevant features through SHAP values, our approach exhibits strong performance with an AUC value of 0.9785 on a standard financial dataset. The combination of cutting-edge methods delivers insights into model decision-making procedures and feature importance, which are vital in financial organizations where accuracy is no more significant than transparency. Experimental findings such as ROC, precision-recall curves, and a rich confusion matrix substantiate the efficiency of our solution versus conventional solutions. In total, our solution not only enhances accuracy in risk assessments but also maximizes explainability of predictions by models, qualifying it as an effective instrument in real-world applications in finance.