Hybrid Model Based on Genetic Algorithm, Bayesian Optimization and Machine Learning for Predicting Credit Status of Legal Clients

This article presents a novel hybrid model for credit scoring of banking customers, combining Genetic Algorithm, Bayesian Optimization, and the XGBoost machine learning model. The primary goal of this model is to enhance accuracy and efficiency in credit risk assessment and reduce the costs associated with prediction errors. In this study, real-world data from banking customers were utilized, and after preprocessing, including normalization and handling of missing data, the Genetic Algorithm was employed for optimal feature selection. Subsequently, Bayesian Optimization was applied as an advanced tool to fine-tune the hyperparameters of XGBoost. The results indicate the superior performance of the proposed model compared to conventional credit rating methods. The hybrid model achieved an accuracy of 79.3% and demonstrated excellent classification performance for both creditworthy and non-creditworthy customers, particularly in high-risk categories. Statistical analyses and performance comparisons with existing methods confirm the positive impact of feature selection and optimized hyperparameter tuning. This model can serve as a practical tool for banks and financial institutions to mitigate credit risk and improve customer management.