An XGboost Algorithm Based Model for Financial Risk Prediction

Abstract

This study presents a novel financial risk prediction model utilizing the XGboost algorithm, analyzing macroeconomic data from the Jorda-Schularic-Taylor database. Our method achieves an 84.77% accuracy rate in predicting systemic financial risks. Unlike traditional models, this model combines the anomaly detection algorithm with the XGboost model, solving the possible "gray sample" problem and improving predictive accuracy. The model's feature importance analysis reveals key indicators, providing insights into the dynamics of financial risk occurrence. Finally, the systemic financial risk score is used to comprehensively evaluate a country's systemic financial risk level, offering a robust risk assessment and monitoring tool. This research enhances the application of machine learning in financial risk prediction, offering a reference for improving risk identification and prevention.

Keywords

• machine learning
• prediction model
• systemic financial risk
• XGBoost