Credit Debt Default Risk Assessment Based on the XGBoost Algorithm: An Empirical Study from China. (19th March 2022)

• Record Type: Journal Article
• Authors: Wang, Jun; Rong, Wei; Zhang, Zhuo; Mei, Dong
• Other Names: Qureshi Nawab Muhammad Faseeh Academic Editor.
• Is Part Of: Wireless communications and mobile computing. Volume 2022(2022)
• Publication Date: 2022-03-19
• Subjects: Wireless communication systems -- Periodicals; Mobile communication systems -- Periodicals; 621.38205
• DOI: 10.1155/2022/8005493
• Languages: English
• ISSNs: 1530-8669

Abstract

Abstract : The bond market is an important part of China's capital market. However, defaults have become frequent in the bond market in recent years, and consequently, the default risk of Chinese credit bonds has become increasingly prominent. Therefore, the assessment of default risk is particularly important. In this paper, we utilize 31 indicators at the macroeconomic level and the corporate microlevel for the prediction of bond defaults, and we conduct principal component analysis to extract 10 principal components from them. We use the XGBoost algorithm to analyze the importance of variables and assess the credit debt default risk based on the XGBoost prediction model through the calculation of evaluation indicators such as the area under the ROC curve (AUC), accuracy, precision, recall, and F1-score, in order to evaluate the classification prediction effect of the model. Finally, the grid search algorithm and k -fold cross-validation are used to optimize the parameters of the XGBoost model and determine the final classification prediction model. Existing research has focused on the selection of bond default risk prediction indicators and the application of XGBoost algorithm in default risk prediction. After optimization of the parameters, the optimized XGBoost algorithm is found to be more accurate than the original algorithm. The grid search and k -fold cross-validation algorithms are used to optimize the XGBoost model for predicting the default risk of credit bonds, resulting in higher accuracy of the proposed model. Our research results demonstrate that the optimized XGBoost model has a significantly improved prediction accuracy, compared to the original model, which is beneficial to improving the prediction effect for practical applications.