Journal of Information Security Research Xiaojun Li, Xiyan Han 15 4 2024 https://doi.org/10.6025/jisr/2024/15/4/155-162 https://www.dline.info/jisr/fulltext/v15n4/jisrv15n4_4.pdf Integrating the Internet and financial companies has expanded the market and avenues for personal loans. While the scale of personal loans has rapidly expanded, it has also brought higher default rates. This paper constructs a personal loan default risk prediction model based on an improved LightGBM model to control default rates and reduce financial company risks. The model’s accuracy in predicting default risks is enhanced by optimizing model parameters and supplementing the evaluation system witha particle swarm optimization algorithm. Experimental results show that compared to four other risk prediction models, this model performs better in predicting default risks. The introduced indicators effectively reduce prediction errors, resulting in higher model accuracy and a better fit to real-world scenarios.