| The Share Price Index Futures market is a product of modern capital market, it can be used for invest or used as a risk management tool, it has a neutral characteristic, its release will have some disturbance to price fluctuations of the stock market in a short term, however, this disturbance will not change the market's long-term trend, specific causes of disturbance from two aspects: overall market trends and internal valuations,among of them internal valuation is the more important factor. Of course, the Share Price Index Futures is a kind of derivatives, it reflects the trend of stock market, therefore, the appearance bound to have a significant impact to the stock market, So understand the Share Price Index Futures price trends can avoid risks better and make decisions in the investment, and then has a great theoretical and practical significance for the price forcast of Share Price Index Futures.In this paper introduces the origin and development of Share Price Index Futures, the differences and connections of the Share Price Index Futures and stock certificate, domestic and foreign research status in the first. Secondly introduces the preparatory work before data analysis and build model: data preprocessing, in order to have a better prediction effect, introduced the principles and applications of wavelet analysis, and denoising the original time series data denoising by the technology of wavelet denoising. As a relatively important financial data, Share Price Index Futures has instability and nonlinear, affected by investor psychology, national policies and other factors. So there exist many difficulties for the analysis of such data. This paper build model by BP neural network model for Share Price Index Futures, in this paper,denoising the original data by wavelet analysis,then used BP neural network model to build model, the close price,open price,highset price,lowest price and volume per minutes of dominant contract from June 23, 2014 to December 29, 2015 as the data, a total of 100,085 samples?This paper is divided into four major parts,the first part is the introduction,the second part is the data preprocessing methods and the third part is the tools and techniques for data analysis. It describes the method and application of denoising by Wavelet, the development, definition and practical application of Neural Networks.The fourth part is the empirical part of the paper.First, use wavelet analyze the whole sample data, 5 layer decomposition were conducted on the raw data by sym8 wavelet function in order to achieve the effect of denoising,make the raw data more smooth, So the accuracy of forecasts can be guaranteed, in order to fit the data well, calculate the index of data after wavelet ananlysis and as independent variable, then selected variable through random forest,last build model by Neural Networks. Calculate the average absolute error and mean relative error for the result of forcast.The average absolute error is 0.0039 in BP neural network,and the average relative error is 8.6587%, the average absolute error and mean relative error of feedforward neural network is 0.0059 and 12.8308%. In comparison, the predictive effect of BP neural network is better than feedforward neural network. |
PDF Full Text Request