Research on the Application of Big Data Technology in Bank Credit Risk Assessment
TL;DR
The results show that introducing big data variables and adopting the XGBoost model can significantly improve the accuracy and AUC value of default prediction, effectively reducing the credit default risk of commercial banks.
Abstract
With the rapid development of FinTech, commercial banks are facing an increasingly complex credit environment. Traditional credit risk assessment models struggle to meet the processing demands of massive and multi-dimensional data. Big data technology provides new perspectives and tools for bank risk control. Based on the theoretical mechanism of big data risk control, this paper constructs a multi-dimensional indicator system including demographic characteristics, asset status, and behavioral preferences. Using the personal credit dataset of a domestic commercial bank, Logistic Regression (LR) model and XGBoost machine learning model were established for empirical comparative analysis. The results show that introducing big data variables and adopting the XGBoost model can significantly improve the accuracy and AUC value of default prediction, effectively reducing the credit default risk of commercial banks. Finally, countermeasures are proposed to address problems such as data silos, weak model interpretability, and privacy protection.