Research On Financial Early Warning Of SMEs Based On BP Neural Network

The small and medium-sized enterprises are important part of national economy, and have the important influence on the whole macro economy. While the complicated environment of small and medium-sized enterprises and the limitations of its own, make it in the process of management facing huge financial risk. In order to prevent financial risks to the financial crisis, it is necessary to set up an early warning mechanism, found a potential crisis in time, and then takes effective measures to ensure healthy and normal operation of the enterprises.The purpose of this article is to establish a set of high accuracy and practical model of the small and medium-sized enterprise financial risk research. In order to achieve this goal, it has proposed the model on the basis of the theoretical analysis, combined with related research and the characteristics of small and medium-sized enterprise itself in this paper, and proved has better accuracy and practicability. The main research work has the following several aspects.Firstly, according to the characteristics of quantity and complex of the financial risk index, this paper conducted two screening of financial indicators. The first step is to statistics of financial indicators, and then preliminary screening out the indicators that is representative and applied in the study of application of high frequency. The second step is through U statistical tests to secondary screening of primary indicators; ultimately determine the application index of each year. These indicators cover the profit ability, debt paying ability, operation ability, and cash flow and so on six aspects.Secondly, on the basis of the above warning index, this paper puts forward the financial early warning model based on BP neural network. It selected the 56 of ST companies and 28 ST companies. In order to make the result more clear, calculation easier, and to retain the original information also, this article mixes the sample data which was standardized and dimensionless by factor analysis method. It reduces the input dimension of BP neural network, and improves the stability of the network. Through training and testing of the neural network, the accuracy of the trained network is verified.Thirdly, through comparing to the three models, this article put forward the financial early warning model based on BP neural network has the highest prediction accuracy. It is 95%. Through comparing the empirical results, this paper proposes the financial early warning model can have predictive effect. And in the three years, the t-2 year the accuracy is better than the other two years.