Research on the Construction of Personal Credit Score Card Model based on Logistic Regression and Random Forest Fusion Model

Abstract

This paper discusses the indicators and models of personal credit score cards, aiming to improve the accuracy and efficiency of evaluation. Through the literature review, we compare the traditional and machine-learning, deep learning-based scoring models, and focus on the combined prediction models to cope with the complex economic situations. Data processing was performed using random forest predicting missing data and XGBoost feature weights assisted feature selection. The model constructed a combined model of random forest and logistic regression and a credit score card assisted by the XGBoost model, which proved that the prediction effect of the combined model was better than that of a single model.