考虑误判损失的Logistic违约预测模型构建

目前企业违约预测模型和现实情况存在一定差距,表现在:1)违约公司与正常公司样本数比例与实际情况严重不符;2)已有的研究极少考虑误判损失;3)鲜少提及信用等级,行业,规模,地区等定性指标对违约预测的影响.针对以上问题,建立了一个考虑误判损失的违约预测Logistic模型,摒弃以往配对原则,采用全样本分析,将地区、规模、行业作为定性指标和29个财务比率指标代入Logistic逐步回归后,最后得到一个违约判别模型.

Building up Default Predicting Model based on Logistic Model and Misclassification Loss

nowadays,there're gaps between default predicting model and true-life,those are: a) the ratio of the number of default enterprises to the number of normal enterprises in the enterprise's short-term-loan default predicting model differs badly from the practical ratio;b) there're rarely studies considering the misclassification loss;c) there're rarely studies considering the qualitative indexing,such as scale,region,industry.To solve aforementioned problems,applying stepwise Logistic regression model,we build up a default predicting model considering the misclassification loss,abandoning the pairwise pattern and using all samples,and introducing those qualitative indexes.The model is significant through testing in statistics.It is more practical and the classification rates are also better.