The Research On BP Neural Network Used On Market Forcasting

The thesis pointed at the stock market, which is a nonlinear system, doing the forecasting. Employed BP neural network, studied the problem on stock market forecasting based on time sequence.Forecasting based on time sequence is a multi-subject domain. Under the guidance of time sequence and neural network theories, the thesis combine the time sequences and neural network together, went further on studying the theory, method and model.The thesis brief introduced some basic knowledge on stock, neural network, and the basic principle of time sequence, all kinds of model, analyzed the time sequence forecasting method based on neural network.The thesis studied the BP neural network's model and structure, its studying rule, built up the time sequence based on BP neural network. Also studied the problem like network size, generalization. And built up a BP network model, using single hidden layer and multiple output system to forecast the trend of the closing index of the coming 2 weeks. Used Levenberg-Marquardt arithmetic train the network, refined the numbers of nerves of hidden layer, effectively solved the problem.The thesis brief introduced how to realize BP neural network in MATLAB, introduced how to creat, initialize, train and simulate a network. Also introduced some functions usually used in MATLAB. And wrote a programme to realize the BP network.The thesis did a experiment on the ShangHai stock Index (close index) , proved that the model and research method is effective. Its simplized the network structure, enhanced the precision. The result came out to be not bad, demonstrate that the time sequence forecasting model based on BP network has a good ability to forecast and generalize.The thesis summarized the previous studying fruit, pointed at its disadvantages, brought author's opinions, put them into practice, hopping could help to do something on the research of market forecasting based on BP neural network.