Research on Machine Learning-based Credit Risk Prediction Models and Algorithms

In the modern financial system, credit risk assessment has become a key link in the risk management of various financial institutions. With the development of big data and artificial intelligence technology, building an efficient and accurate credit risk prediction model has become an urgent demand in the industry. Traditional methods such as linear regression and logistic regression are powerless when facing the complex and changing data environment, while machine learning technology, with its automatic data processing ability and efficient prediction ability, has become a new approach to solving the credit risk assessment problem. Based on the long-term customer credit data provided by the 2024 Sichuan Provincial Data Science and Statistical Modeling Competition, the data covers multiple dimensions of feature information, including demographic statistics and borrowing and repayment behavior, as well as risk performance data after loan application. Starting from the stages of data preprocessing, data analysis, feature selection, data set division, LightGBM model establishment, and effect evaluation, a credit risk prediction identification model is developed to identify customer risks and help financial institutions further enhance their risk identification and prevention capabilities across economic cycles. The experimental results show that the model proposed in this paper has good discriminatory ability and stability under different economic cycles. The final AUC score is 0.704523, ranking third on the private list.We've provided a new data set and source code for this work:https://github.com/YHHHSK/2025-amlds.git