Fuzzy Theory And Neural Networks In Stock Forecasting Application

Stock price is a barometer and a indicator to judge stock market, which comprehensively reflects the development of stock market. We can invent and make policy basing on fully grasping the development of stock market through prediction and research. It helps to develop national economy orderedly and healthy. At present, artificial intelligence have used in stock prediction and got better resulte, besides mathematical methods. In the article, fuzzy theory combines with neural networks to establish the stock prediction system model.Firstly, this paper makes use of improved fuzzy degree of nearness to categorize the opening price,closing price,change places,rose drop,minimum price,maximum price,volume and turnover of stock price, and then chooses input samples of BP neural network, thus presents a new method for choosing input samples of network. As a result, the accuracy increases about 50 percentage.Secondly, according to the practical application of the stock price forecasting, this paper improves the triangular membership function,semi-trapezoidal membership function and fuzzy evaluation operator of fuzzy comprehensive evaluation system in fuzzy theory, and presents a new method basing on improved degree of nearness to calculate weight of the fuzzy comprehensive evaluation system, which raises the correct rate about 4 percentage.Finally, this paper establishs portfolio BP neural network for the opening price,closing price,change places,rose drop,minimum price,maximum price,volume and turnover of stock price. According to classification to choose input samples of subnet. Then we take advantage of the improved fuzzy comprehensive evaluation system to judge the output of network and predict stock price movement. Neural network combines with the improved fuzzy comprehensive evaluation system effectively, and builds an integrated forecasting system for stock.