Research on the bank loan access model based on the improved logistic regression score card

In recent years, financial fraud has escalated alongside the development of personal credit business, making the optimization of customer classification and pre-loan credit risk assessment crucial. Although there are already risk control methods based on machine learning and deep learning, a single model is insufficient to address complex fraud scenarios. This paper proposes a personal credit scoring model based on an improved logistic regression algorithm, using the Kaggle credit card dataset to establish a scoring card model through the analysis of user behavior data. The research encompasses data cleaning, decision tree binning, WOE encoding, IV value-based variable selection, correlation analysis, logistic regression modeling, variable significance analysis, and the evaluation of the model using KS value, ROC curve, and AUC value. The final model achieves a KS value of 0.328, an AUC value of 0.73, and a prediction accuracy of 93.44%, effectively preventing defaulting customers and enhancing risk control effectiveness.