Stock Price Prediction Based On Support Vector Regression And Differential Evolution

The stock market has abundant data resources,and a lot of hidden information is easy to be excavated.Stock price prediction is very difficult,because the stock price is not only related to the stock itself,but also related to the stock market,financial market,public opinion and so on.It is difficult to qualitatively describe the impact of these social dynamics on the stock.This paper only uses the stock exchange data and price as the original data to design a set of methods that can be used to explore the effective data hidden in the stock exchange data and to predict the stock price.It helps the investors to make reasonable scientific investment and reduce the risk of stock investment.Support Vector Regression(SVR)is a mature technology at present.It shows many unique advantages and good performance in solving small sample,nonlinear and high dimensional pattern recognition.It can solve common classification and regression problems effectively.It is suitable as the basic method of stock price prediction model.The trading data of several stocks in NASDAQ stock market are selected as the data set,and the SVM model is trained with the opening price,the highest price,the volume,the lowest price and the technical indexes generated as the input,and the closing price as the output.The key and difficulty of training SVM model is to solve the problem of parameter optimization.Differential evolution algorithm is suitable for parameter optimization of support vector machine model because of its intelligence and high degree of approximation.The experiment fully proves that the support vector regression optimized by the improved differential evolution algorithm in the stock price prediction can be better than the support vector regression optimized by other intelligent algorithms.