Process Improvement Study of BP Neural Networks Model for Financial Crisis Prediction Based on Chinese Real Estate Listed Enterprises

摘要

This paper tries to put BP neural networks of artificial intelligence into financial crisis prediction. When a company is degraded from "Normal" to "ST" or "*ST" by China Securities Regulatory Commission, financial crisis happened. We find out 51 real estate developments listed companies which were degraded from 2003 to 2012 in China. By analyzing the data of these companies three years ago, we calculate the probability of financial crisis. Besides, we choose 669 normal state companies as matching samples in the same period, same industry, and similar size. We find that it is better to apply one specific calculation method of financial ratio into BP neural networks model for prediction than to use all of them. And accuracy of prediction about setting one hidden layer is better than two layers.