Credit Default Prediction Based on Adaptive Strategy for BWOA-CatBoost

As one of the main operations of banks, credit assessment of credit loans is a crucial step in mitigating the risk of default. In this paper, we present a novel credit default prediction method that combines an adaptive strategy with the Bayesian Whale optimization Algorithm (BWOA) and the CatBoost classifier (BWOA-CatBoost). Our primary objective is to improve the predictive performance and computational efficiency of credit default prediction. We begin by conducting exploratory data analysis on the dataset, followed by data preprocessing. For feature selection, we employ random forest and gradient boosting decision tree (GBDT) methods based on feature importance. Next, we introduce an adaptive strategy that incorporates Bayesian optimization and whale optimization algorithms to optimize the parameters of the CatBoost classifier, which improves the accuracy of credit default assessment. This adaptive strategy enables automatic parameter adjustment and enhanced algorithm efficiency based on the current problem and environment. The experimental results suggest that BWOA-CatBoost outperforms Logistic Regression, Random Forest, XGBoost, LightGBM, and CatBoost in terms of predictive accuracy, demonstrating its potential for practical applications in the financial industry.