Research And Application Of Algorithm About Stock Price Prediction Based On Support Vector Machine

With the vigorous devel.opment of the stock market,more and more people were beginning to known the stock market well,and chose to stock investment as a way of managing money.Investors expected to get a investment advice which is less risky compared with others,and they wanted to get some suggestions and assistance in making investment decisions.There are prediction researches about the stock market by some scholars,including the forecast of the whole stock market,the prediction of stock price change and so on.In this paper,the prediction models and methods of stock price for the different stocks were researched.There is significance of theory and prospect of application for this investigative job.Token stock trading in the Shanghai Stock Exchange as an object of study subjects,and the application of algorithm about stock price of future trading day according to support vector machine were researched.Firstly,two different parameter options based on the prediction models of stock price were designed,the prediction model of stock price based on support vector machine were constructed,and the closing price of future trading days were predicted.In order to discover the effects of different normalized ranges for the predicting stock price,the historical data of stock price normalized were needed before training the prediction model.Different combinations of feature showed different prediction results when we predicted the stock price,and considered the correlation between the different characteristics of the data,we used the method which is the law of permutation and combination and the characteristics of the data in order from less to more,to filer features when selecting the data characteristics to get the desired model of training.Using different kernel functions and comparing and analyzing the different forecasting prices and real prices of stock when we trained the model of prediction for the different stocks,selected the kernel function and parameters of model for the prediction model of different stocks.Finally,in order to make the difference became the smallest between prediction results and real stock prices and give investors the reference of price that can get more profit,we need to select the best combination of conditions for the prediction model of different stocks according to the selection method of combination,and used this parameters to build the model.Stock price prediction involves modeling of complex nonlinear problems.The influence will be found in the model because of the unpredictable factors when building the prediction model.This article extracted key characteristics and influencing factors from the changes of the stock market and set up the model for the question,improved accuracy of the prediction of stock price through using parameters options of the appropriate model,and provided a reference investment advice for stock investors.