微型电脑应用

我国房地产企业债券自2018年起频繁违约，违约债券数量和违约金额均远超其他行业。为此，基于机器学习技术构建房地产企业债券违约预警模型，提前识别可能发生违约的债券。研究发现：在不同的违约预测时间点，用于预测的指标重要性不同，且构建的预警指标组在违约发生前半年对违约的预测性能最好；在指标体系中加入“三道红线”财务指标提高了房地产企业债券违约预测的性能；引入债券交易数据特征作为违约预测的影响因子，对违约有着较好的预测作用和解释意义。

Since 2018, China's real estate enterprise bonds have defaulted frequently, and the number of defaulted bonds and the amount of defaults far exceed those of other industries. Therefore, based on machine learning technique, an early warning model for bond default of real estate enterprises is constructed to identify bonds that may default in advance. The results show that: the importance of the indicators used for prediction is different at different default prediction time points, and the constructed early warning indicator group has the best prediction performance for default half a year before default occurs; adding the “three red lines” financial indicators to the index system improves the performance of bond default prediction of real estate enterprises; the characteristics of bond transaction data are introduced as the influencing factors of default prediction, which has a good predictive effect and explanatory significance for default.