Neural Network Modeling And Application In The Stock Market Prediction

The artificial neural networks is a newly developed cross subject. It is a nonlinear information processing system imitating the structure and the function of human brain. It possesses learning ability, parallelism, robustness and easiness for hardware implementation. It mainly applies to pattern classification, function approximation and data compression. Neural networks have developed rapidly in the last twenty years and have got a wide application in many fields.Stock market is a highly complicate nonlinear system, it is variation has its own regulation, but also is influenced by many other factors such as politics, economy and psychology. While traditional prediction techniques are based on statistics and meet difficulties in stock market analysis. Neural network enjoys the virtue of self-organization and adaptability and can learn the economical knowledge from historical data. So it is suitable to solve problems in stock market prediction. By now, many researches have established prediction models for different stock market, and have obtained good results. But Chinese stock market has only a history of ten odd years and is not yet quite perfect. Hence the theories and methods used successfully for abroad markets are not necessarily suitable to Chinese market.The following works have been done in this thesis:1. Several applications of the neural networks in the stock market prediction have been summarized.2. We introduce some technical indexes into our model of stock marketprediction; hence improve the correct rate of prediction.3. Analysis and comparison between predicting model based on the BP neural network and the one based on fuzzy neural network has shown that the latter has better performance.4. Based on consideration of the ups and downs of Shanghai stock index and Changhong stock, we find out a class of simple black-horse stock pattern.This article focuses on the empirical research about the stock prediction model based on BP neural network and fuzzy neural network. The experimental result shows that it is applicable and feasible to use neural network to predict the domestic stock market. However, the amount of training samples used in our model is relatively small, and the dimension of each sample is low. We do not consider the influence of the government policy and some other factors, which usually play an important role in the fluctuation of the stock market. So there is still a long way to go before applying it to real applications, and need to be improved further.