Financial early warning Model of Listed Company Based on BP Neural Network

Financial situation can indicate company performance, by analyzing which development trend of companies can be predicted. This paper predicts future financial situation of listed American companies by setting a BP neural network model, based on its nonlinear mapping, strong learning ability and characteristic of easily coded. All the data is derived from monthly data of listed American industrial and financial companies from 1991 to 2011. The model was trained according to different prediction horizon, the paper suggested that the average prediction accuracy of each model can reach more than 80% when prediction horizon is less than two years. Furthermore, the shorter the prediction horizon is, the higher prediction accuracy we can get. When the prediction horizon is within six months, the average prediction accuracy of the model can reach more than 90%. In addition, the model is applicable during the financial crisis. Therefore, it's feasible to forecast the finance risks of corporations by neural network.