Credit Rating Prediction Based on WOA-Cross-Validation-LightGBM

One of the key challenges faced by banks in the context of continuous growth in customer loan demand is how to effectively execute customer credit ratings. This study introduces SMOTE to solve the problem of data imbalance and utilizes RFE_XG-Boost for feature importance assessment to screen out the key features that have a significant impact on credit ratings. Based on these selected features, the LightGBM model is used to predict customer credit ratings. To further optimize the model performance, this paper applies the Whale Optimization Algorithm (WOA) to adjust the parameters of the LightGBM model and combines with the 5-fold cross-validation technique to enhance the model prediction accuracy. Using this method, the study significantly improves the accuracy of the model prediction. The experimental results confirm that the LightGBM model, which combines SMOTE, RFE_XGBoost feature selection and WOA, shows high accuracy and practicality in credit rating prediction. The methodology of this study provides an efficient solution for bank credit rating and lays the foundation for future research in similar areas.