The Study And Application Of V-SVR Based Brain Storming Algorithm

Intelligent prediction is used to predict stock market more because of its advantages in dealing with the nonlinear system problems. AS one of the most mature theory in the intelligent predictions, Support vector machine is acclaimed. Now there are a lot of research about the standard SVM, but for v-SVR is little. In this paper, the CSI300Index and shenzhen Composite Index are predicted. The Index from April20,2011to9December,2011altogether160days’data are selected for the study.Select20pairs Index influential factors including opening price, closing price, highest price, lowest price, trading volume, transaction amount, price change, change amount, The moving average line MA (5), the moving average line MA (10), the moving average line MA(20), Similarities and differences between the average MACD, plus or minus the difference DIF, stochastics KDJ.K, stochastics KDJ.D, psychological line indicator PSYMA (6), Relative Strength Index RSI (6), Relative Strength Index RSI (12), the deviation rate indicators BIAS (6), deviation rate index BIAS (24). These factors are generally thought to have a bigger impact on the stock Index. Try to establish support vector regression model between the next day’s opening price and these factors. First, these date are preprocessed, the pretreatment method are mainly the standardization of statistical processing, correlation analysis and principal component analysis. The processed date are better for the modeling effect. Then the data are analyzed by v-SVR, according to the relevant literature, parameters selection of SVM is always a problem. In the choice of three related parameters, this paper puts forward with brain storming optimization (BSO) algorithm to select parameters. Compared with the default parameters of LIBSVM package, parameters selected by grid algorithm and parameters selected by particle swarm optimization (PSO) algorithm, brain storming method selecting parameters is the best of the four methods.