The Prediction Research On CSI300Index Based On GSA-AVR Model

The stock market is an important part of China’s financial market s and hasclose relation to the stability and healthy development of China’s economy. As arepresentative of the stock market, the Shanghai and Shenzhen300covers thebasics of the two major markets and it has laid a good foundation for the study ofthe China’s stock market. With the modern development of economy, science andtechnology, the research methods on the stock market has risen from a meretheory to empirical analysis, relying on computer equipment for technicalsupport seems to be the main trend for research. In the full grasp of the CSI300stock index, based on a grid search algorithm for support vector regressionoptimized by computer operations, access to monthly predict of CSI300closingIndex. The accurate prediction of stock has a strong practical significance onboth the stock market regulatory authority and the stock market investors.The theme of the paper is based on the domestic and foreign related researchconditions. First of all is to determine the predictive input indicator. Thedevelopment of the stock markets and the macroeconomic environment areinextricably linked and therefore initially selected four market indicators andnine macroeconomic indicators, making an analysis with CSI300closing price.After a series of measurement statistical processing, eliminating the weakcorrelation indicators, and ultimately determining the three market indicators andfour macroeconomic indicators as input variables of the model; Second, afterdetermining the input variables, should optimize the construction of predictivemodels through a grid search algorithm. For support vector regression model, thesearch for the optimal parameters directly affect the outcome of the finalprediction accuracy, and therefore put grid search algorithm to optimize theprediction model, building the GSA-SVR predictive models on CSI300Index,using55monthly data for model training based on four different kernel functions,and analyzing and comparing the training results; Again, on the basis of fourdifferent kernel functions using seven monthly data on the training model of theindex trend forecasting, comparing different kernel function with predictionresults, and found that a support vector regression model based on the linearkernel function are superior to other kernel functions in relation to training timeand prediction accuracy, and owing the highest degree of fit between thepredicted results and the true value; Eventually, on the basis of the results of theanalyzing and forecasting, providing some constructive comments respectively from the micro and macro perspective on the development of China’s stockmarket. From the microscopic point of view, the stock market regulatoryauthorities should do well on supervisory work, stock market investors have tolearn to analyze the stock market for their own benefit; From the macro point ofview, starting with a few macroeconomic factors that affected the stock market todevelop relevant policies measures to promote the healthy development of thestock market.