Abstract: With the rapid development of the Internet economy, the scale of personal credit has exploded in recent years. Credit risk control has always been a hot issue for financial institutions. This paper studies the application of ensemble learning algorithm XGBoost to the prediction of personal credit default. By analyzing the existing data, and using the XGBoost algorithm to build a personal credit default prediction model. Experimental results show that XGBoost performs better than logistic regression and random forest algorithms. By using the XGBoost algorithm to measure the importance of features, it helps to quickly and effectively judge personal credit risk.