Bond market default risk early warning method based on interpretable machine learning

Abstract

The invention discloses a bond market default risk early warning method based on interpretable machine learning, and relates to the field of unbalanced data processing in machine learning. A clustering sampling technology and an XGBoost machine learning algorithm are combined, an SHAP method is used for explaining a model prediction result on the basis, related samples of the China bond market in 2014-2020 are collected through an available channel, variables passing statistical test are included in a model for training and verification of the model, and the output of the method is the default risk probability. And the prediction result can be explained globally and locally through the Shapley value, through the above technical scheme, more accurate debt default risk prediction is realized, the prediction of the model can be explained on the premise of not sacrificing the precision of the model, and the defect of weak explanation of a machine learning model is made up. The overall importance of each index can be effectively identified through global interpretation, the influence of each index on each enterprise is further quantified through local interpretation, and difference research on the microscopic level is facilitated.

Description

technical field [0001] The invention belongs to the field of default risk early warning, in particular to a bond market default risk early warning method based on explainable machine learning. Background technique [0002] In the existing life, the default risk of the bond market is a crucial part of the credit risk in the financial market. Compared with stocks, the returns of bonds are relatively stable and the risk is small. However, with the rapid growth of China's economy, various enterprises in the market In the process of development and growth, financing needs and debt scale are also gradually expanding. While China's economic development has entered a new normal, the economy is facing greater downward pressure, and defaults have begun to appear in the bond market, and the frequency of occurrences has a rapid growth trend. , showing the characteristics of an increase in the number of defaulting entities, an increase in the amount of default, and the spread from privat...

AI Technical Summary

Problems solved by technology

[0002] In the existing life, the default risk of the bond market is a crucial part of the credit risk in the financial market. Compared with stocks, the returns of bonds are relatively stable and the risk is small. However, with the rapid growth of China's economy, various enterprises in the market In the process of development and growth, financing needs and debt scale are also gradually expanding. While China's economic development has entered a new normal, the economy is facing greater downward pressure, and defaults have begun to appear in the bond market, and the frequency of occurrences has a rapid growth trend. , showing the characteristics of an increase in the number of defaulting entities, an increase in the amount of default, and the spread from private enterprises to state-owned enterprises. Bond defaults occur frequently. In addition to the reasons for the macroeconomic downturn, there are also factors such as industry recession and imperfect corporate governance. Bond defaults It is the serious deterioration of the credit status of enterprises, which will have adverse effects on the external financing activities and daily production and operation of enterprises, and also cause huge losses to investors in the bond market. The development of the bond market has enriched China's multi-level financial system. It is conducive to increasing the proportion of direct financing, reducing the excessive reliance of enterprises on indirect financing from banks and other financial institutions, and alleviating the problems of difficult and expensive financing. Regulatory capabilities and investors' risk management and risk-taking capabilities have put forward higher requirements. Against the background of frequent credit bond defaults, bond default risk warning work has become more and more important practical significance

[0003] At the same time, in recent years, the application of machine learning and artificial intelligence technology in economics, management and other fields is in the ascendant. Compared with traditional statistical methods, machine learning algorithms can better fit the complex nonlinear relationship between predictors and targets, and obtain Better out-of-sample predictive performance, however, machine learning no longer provides parameter estimates linking predictors to output variables, less transparency, currently, the highest accuracy based on machine learning is usually obtained with complex models, even by experts It is difficult to explain, such as integrated learning and deep learning, which greatly reduces the acceptance of machine learning results by financial industry personnel. For debt market problems, both lenders and applicants want to know the risk assessment made by artificial intelligence models. Reason, therefore, research on interpretable prediction of debt risk based on machine learning models has practical significance

[0004] Existing studies mainly use traditional statistical methods and classic machine learning methods, such as artificial neural networks, support vector machines, decision trees, Bayesian methods, etc. However, debt default risk prediction is a typical category imbalance problem. Classification models tend to predict samples as the majority class, and cannot effectively predict minority groups. In addition, although machine learning is often higher in prediction accuracy than traditional statistical methods, due to low transparency, it is impossible to make predictions. to explain, it is difficult to really land

Images