前馈神经网络和Fisher判别分析对上市公司财务状况异常的预测研究

The research on predicting financial abnormity in listed companies using fisher discriminant and feed forward neural networks

本文建立了基于数值优化的Levenberg-Marquardt算法的前馈神经网络预测模型和Fisher多元判别分析模型对上市公司的财务状况异常进行预测,研究结果表明前馈神经网络预测模型在预测精度上较传统的Fisher多元判别分析模型有一定优越性,可以作为证券投资者和分析人员使用的有效预测工具。

This paper builds a multi-layer feed forward neural network model based on Levenberg-Marquardt algorithm and a Fisher discriminant model to predict financial abnormity in Chinese listed companies. Research result indicates that accuracy of neutral network predicting model is better than that of traditional discriminant model and it can be applied as an effective predicting tool by securities investors and analysts.