机构地区: 江西财经大学金融学院
出 处: 《经济学(季刊)》 2005年第B10期173-188,共16页
摘 要: 本文将中国上市公司因财务状况异常而被特别处理(ST)作为企业陷入财务困境的标志,采用主成分分析方法确定模型变量,并利用多元判别分析、Logistic回归和改进型BP神经网络三种方法进行财务困境预测。比较其预测结果发现,BP神经网络模型的预测准确率明显优于多元判别分析和Logistic回归模型,而后两者的判别效果接近,可见改进型BP神经网络模型更适合于企业财务困境预测。但三种模型的长期预警能力均不够理想,需要建立以定量模型为主、定性分析为辅的上市公司财务困境预测方式,以提高预测的准确性。 This paper treats getting ST as a signal for a company's financial distress. We first use the principle component method to determine the variables entering the prediction models. Then the multivariate discriminatory analysis, logistic regression, and neural network method are used to in the prediction. Comparing the results of the three models, we find that improved back propagation neural network model is better than multivariate discriminatory analysis and logistic regression in terms of prediction accuracy, and the two latter models' predictions are similar, Therefore, improved back propagation neural network model is more suitable to predict firm's financial distress. However, in terms of long-term prediction, all three models show weaknesses.
关 键 词: 财务困境 多元判别分析 回归 改进型 神经网络
领 域: [经济管理—国民经济]