基于数据挖掘的公司债券风险预测分析

Prediction and Analysis of Corporate Bond Risk Based on Data Mining

摘要: 大数据时代的到来，网络与计算机技术的发展，给债券市场带来了风险预警的新工具。本文从公司内部经营状况的微观风险信息角度出发，利用数据挖掘技术找出影响公司债券到期偿还的关键因素，并建立预测债券违约的方法。研究以XGBoost极端梯度提升算法发现债券是否违约的主要影响因素是营业收入同比增长率和资产负债率，然后建立了债券是否违约的二元logistic回归模型，通过二元logistic回归模型可以进行债券违约的预测。

Abstract: The advent of the era of big data and the development of the Internet and computer technology have brought new tools for risk early warning to the bond market. From the perspective of micro risk information of the company’s internal operation, this paper uses data mining technology to find out the key factors affecting the maturity of corporate bonds, and establishes a method to predict bond default. Using the extreme gradient boosting algorithm, it is found that the main influencing factors of whether the bond defaults are the year-on-year growth rate of operating revenue and asset liability ratio in this paper. And then we establish a binary logistic regression model of the bond. The binary logistic regression model can predict whether the bond defaults.