Stock Prediction Based On The Principal Components-BP Neural Network

With the economic growth and the conversion of people's investment consciousness, stock has become an important part of people's life in modern time. The Stock market is an extremely complex non-linear dynamics system, participating in stock investment; Stock forecast has greatly been one of focuses of public topic. The proceeds of stock investment always equal the risk. So establishing a stock forecasting model, with has higher operation rate and precision, has theoretical significance and applicable value.The inputs of forecasting affect the accuracy of the results and computational speed. The traditional method of subjective selection is inefficient in choosing the inputs. The Principal Component Analysis(PCA) method is applied to preprocessing the data sets, eliminating the correlation among the inputs, and simplifying the structure.The paper mainly study the following four sections:First chapter describes stock research background and significance of the topic. Summaries all kinds of analysis about stock, discuss the usage of neural network on the stock analysis and finally propose the problem of current stock analysis and solving solution.Chapter II main discussion on evaluation index of stock forecasting, pointed out that the main problems in stock forecasting, analysis and comparison of prediction method of stock market of several typical, simple introduced the principle of artificial neural networks. Deride the technical analysis.Chapter III describes the main components of--BP neural network stock prediction based on feasibility, how analysis of principal components analysis and BP neural network theory, the establishment of a forecast based on principal components of--BP neural network model. How to build the model based on PCA——BP neural network.The fourth chapter of Shanghai composite index, for example in Shanghai Stock Exchange respectively on single BP neural network to predict and forecast based on PAC--BP neural network simulation, simulation results show that the effectiveness of using principal component analysis of selecting input variables, it significantly reduces the prediction time, improve forecast accuracy. Using principal component analysis on the samples processed to form a new set of training samples, reduced artificial neural network modeling for network input, while eliminating correlation of input factors and simplify network structure, can greatly increase the network learning rate. Artificial neural network model to be able to achieve high accuracy, is influencing factors and the mechanism is not yet clear of artificial neural network prediction of stock market provides an effective way.This innovative idea is as follows:Using principal component analysis on the samples processed to form a new set of training samples, reduced artificial neural network modeling for network input, while eliminating correlation of input factors and simplify network structure, can greatly increase the network learning rate. Artificial neural network model to be able to achieve high accuracy, is influencing factors and the mechanism is not yet clear of artificial neural network prediction of stock market provides an effective way.Shortcomings of this article is as follows:Training small amount of sample selection and the other has a material effect on the stock market fluctuation factors not be taken into account, in addition, select the number of hidden layer node not found reliable theory based, so the work away from the practical application of this article is still a considerable distance, to be further improved in future research.