BP Neural Network Investment Analysis Application In Stock

Stocks, as a barometer of the financial markets, have to be cared by investors since their appearance. Considering the high-risk, high-yield and variability characteristics of stocks, the analyses and forecasting of stocks become the focus for investors. Stocks are affected by political, economic, and investor psychology, and many other factors, so the factors’ complexity decide the complexity and uncertainty of stock analysis. There are many methods and improved technologies proposed to improve the prediction accuracy of the stock forecasting. These improved methods are based on the traditional forecasting methods, and introducing many new techniques including artificial intelligence algorithms.The basic theory, structure, learning rules, and the building process of BP neural network are introduced in detail in this paper. A number of technical indicators are combined as input nodes when the built model is used to forecasting, two hidden layers are chosen when the network is designed and the hidden layers nodes are chosen variably as30,35,41,45,50. Basing on a number of works which have done before, this paper has controlled the number of hidden layer nodes, in order to find the best forecasting results. This paper has used the built BP model to forecast the amount of increase of Wanke stock, Wuliangye stock, Pufa bank stock, Baogang stock, China Petrochemical stock, Shanghaijiahua stock, Guizhou Maotai stock, Zhongchu stock, Yili stock and China renshou stock from July15to July26,2013. The forecasting results show that the forecasting accuracies are varying with the change of hidden layer nodes. Totally, the forecasting results are the best when the number of hidden layer nodes is chosen as35and41.