上市公司财务危机预警分析——基于数据挖掘的研究

本文以我国上市公司为研究对象,选取了1999-2001年被ST的公司和正常公司各73家作为训练样本,2002年被ST的公司和正常公司各43家作为检验样本,分析了财务危机出现前2年内各年两类公司15个财务指标。在进行数据挖掘中,我们运用了三种独立的方法,分别为判别分析、Logistic回归和神经网络,结果发现神经网络预测的效果要优于其它两种方法。最后,结合了这些方法的优点,建立了一种混合模型,研究表明预测的正确性要高于每种单独方法,从而提高了模型的预警效果。

A Prediction analysis of financial distress for listed companies——based on data mining approach

This paper has the Chinese listed company as its research object.Seventy-three companies in special trade and that much in normal trade from 1999 to 2000 are selected as the training sample.Forty-three companies in special trade and in normal trade of 2002 are selected as the test sample.Fifteen financial ratios from two years before the finance distress have been analyzed.Three single classifiers—discriminant analysis,logistic regression and neural network have been used in data mining process.It has been proved that the neural network is better than others.Combing the merits of the above methods,a hybrid method that increases prediction performance has been put forward.The empirical tests show that it can produce higher prediction accuracy than individual classifiers.