基于概率神经网络的欺诈性财务报告的识别研究

Fraudulent Financial Statements Identification Using Probabilistic Neural Networks

本文探索概率神经网络PNNs(Probabilistic Neural Networks)在构建欺诈性财务报告识别模型方面的有效性,重点探讨了PNN模型变量的选择及平滑参数的确定问题,同时将所提出模型的性能和人工神经网络(ANNs)、logit回归模型的性能进行了比较.结果证明,PNN模型具有很高的预测力,并发现该模型的性能优于ANN模型以及logit回归模型.

This paper uses Probabilistic neural networks detects firms issuing fraudulent financial statements （FFS） （PNNs） for the development of a model that Focusing on how to select the variables that can identify FFS best and specification of the smoothing parameter of the model, a comparative analysis on efficiency of PNNs with artificial neural networks （ANNs） and logistic regression were also made. The results demonstrate the high explanatory power of the PNN model in identifying fraudulent financial statements. The model is also found to outperform traditional ANN model, as well as logistic regression.