Credit Evaluation Model Based on Adaptive Synthetic and Grey Wolf Optimization Improved XGBoost Algorithm

Effective default risk prediction is a critical guarantee for the smooth development and operation of financial institutions, with sound mathematical models serving as the foundation. This paper employs an adaptive synthesis method to balance sample sets and uses the Grey Wolf Optimizer XGBoost to establish a default risk prediction model. Consequently, the ADASYN-GWO-XGBoost credit evaluation model was generated, achieving an accuracy rate of 98% in predicting creditworthiness for Chinese family farms. The study further introduces financial data of Chinese listed companies and compares the prediction results of different models across two samples based on multiple indicators. It demonstrates the effectiveness and scientific nature of the ADASYN-GWO-XGBoost in risk prediction, as well as its improved performance in evaluation capabilities, providing verifiable insights into the method.