基于LVQ神经网络的财务舞弊识别模型实证研究

为了保护广大投资者和规范国内资本市场,对财务舞弊识别的研究具有重要的意义。在参考前人研究的基础上,选择能识别财务舞弊的指标,利用主成分分析法约减指标,得到9个综合变量。在此基础上,利用学习矢量量化(Learning Vector Quantization,LVQ)神经网络建立财务舞弊识别模型;此模型对测试样本的判断准确率高达90.9%,验证了模型的有效性。最后把此模型与用其他方法建立的财务舞弊识别模型进行比较,发现LVQ神经网络建立的财务舞弊识别模型,能更有效地识别测试样本有没有财务舞弊。

The Empirical Studying on Detecting the Fraudulent Financial Statements Based on LVQ Neural Network

In order to protect investors and regulate the capital market, the research of the fraudulent financial statements detection is very important. Then principal component method was used to reduce the numbers of indicators. Through it, 9 comprehensive indicators which can detect the fraudulent financial statements were acquired .Then the model of identifying the fraudulent financial statements was built by the LVQ neural network. Then an empirical research about the model has been done, the accuracy of predicting the testing samples was up to 90.9%.At last by comparing with other models, this paper find that the model built by LVQ neural network can detect the fraudulent financial statements effectively.