上市公司财务危机预测模型比较研究

A Comparative Study on the Model of Financial Crisis Prediction for Listed Companies

财务危机预测一直是公司金融领域的研究热点,危机预测模型随着研究的深入在不断地发展变化。多元线性模型、Logistic模型、COX模型、神经模型是当前主流的四种财务危机预测模型。以2003年—2008年制造业上市公司为样本,采用财务比率作为预测变量,通过样本公司财务数据建立了四种财务危机预测模型,利用判别精度和ROC曲线比较不同模型在财务危机发生提前3年的预测效果。研究发现神经网络模型在判别精度和稳定性方面均优于其他模型。多元线性模型、COX模型、Logistic模型预测虽然均具有应用价值,但三类模型在预测精度和稳定性上各有优劣。

Financial crisis prediction has been the focus in the study of corporate finance. The prediction model changes with the further study in this field. The multivariate line model, logistic model, COX model, and NN model are currently the main prediction models. This paper takes the listed companies in manufacture industry during the year from 2003 to 2008 as the sample and uses financial ratio as a prediction variable and builds up four financial crisis prediction models based on the above corporate financial data. We use judgment accuracy and ROC curve to compare the prediction effect under different models three years before the financial crisis takes place. We find that NN model has better prediction effect than that of the others, and the other models also have their advantages and disadvantages in terms of accuracy and stability though they have some application value.