The Stock Market Based On Dynamic Fuzzy Neural Network Prediction

Stock is the product of the market economy, companies can gather funds through it, investors will invest in the stock as a means. With economic development, people gradually deepening understanding of the stock market, more and more investors into the stock market, stocks have become common everyday things. Invest in stocks with high returns, but also its accompanying high-risk. In stock investors, there are many outstanding benefits achieved, but more unpredictable is the face of the market, investors don't know what to do at their mercy. However, the powerful appeal of the stock market, people has never given up its pursuit, individual investors and institutional investors have a strong desire to find an effective way to predict stock movements. More and more researchers have also invested in the tireless research of stock prediction method. Many ways to forecast the stock market, for example, through analysis of financial data, trying to find the factors which determine market trends, to forecast future trends.However, the stock market is a complex nonlinear dynamic systems, the use of traditional time series forecasting techniques, like it is almost impossible to predict. With the knowledge of the stock features of recent years, the Artificial Neural Network to predict the stock market has become mainstream. As the neural network a powerful ability to deal with nonlinear problems, a good self-learning ability and adaptive ability, it can approximate any function. Forecast the stock has achieved a certain effect. Fuzzy logic provides a mechanism for the generation of fuzzy rules, can simulate the human reasoning process and make use of existing expertise. As the fuzzy logic and neural networks has many complementary strengths and weaknesses, combining the two complement each other has become a natural thing, that is, fuzzy neural networks. Fuzzy neural networks for stock prediction there are many different forms, the author has carefully studied the results, learn from their successes, shortcomings were also analyzed, and based on the improved fuzzy neural network-dynamic fuzzy neural network(D-FNN) forecast for the stock market.Neural network prediction for the stock market, is the use of the composition of the stock of historical data time series, by neural network self-learning to discover the laws of input and output analog the function between them and use this function with in the prediction of future stock price. D-FNN for stock prediction using the same principles, but compared with the normal neural networks, learning algorithms there is a big difference. D-FNN is improved fuzzy neural network, it did not need as ordinary as the prior artificial neural network to determine the network structure, but uses a learning algorithm for adaptive fuzzy rules can be generated according to the needs of the specific problems of different fuzzy rules, and according to fuzzy rules to determine the network structure. In determining the network structure, as it is not randomly generated, and need not trial, so more scientific. In stock forecast, with the choice of different stocks, the resulting network structure is also different, so that the trained network is more focused, be more accurate forecast for the stock.The future trend of the stock relative to the specific value of the stock is easier to be predicted, and forecasts the trend of the stock is more meaningful, therefore, the paper also forecast the trend of stock. Experimental results show that the D-FNN forecasts the trend is more stable than price predicted, this finding for future research with some reference.This paper improves algorithm for fuzzy neural networks, and applied it to the stock forecast, intended to find a more suitable stock forecasting means. D-FNN for stock prediction and experiment, through experimental analysis, try to provide new ideas and practical methods for the investors and researchers.