Comparative Study on Models of Financial Distress Prediction Based on Neural Networks

It is helpful for a number of company stakeholders to find potential crises in time and take measures to reduce losses by constructing effective models of financial distress prediction. Firstly, the paper statistically disposes the financial indexes of Chinese listed companies in virtue of T-test. Then it constructs three models of financial distress prediction based on back propagation neural network (BPNN), probabilistic neural network (PNN), and Elman neural network respectively. Finally, the study compares the performances of three models in the training and testing sample set of Chinese A-shares listed companies. Results show that all three models can validly predict company financial distress. However, the PNN model is superior to the BPNN model and the Elman neural network model in the accuracy of prediction in the testing sample set. Results also show that the PNN model's accuracy of classification is lower than those of the BPNN model and the Elman neural network model in the training sample set.