CO V-MWSVR And Its Application In The Stock Market

The stock market is an important part of modern financial markets. It has an irreplaceable role in national economic, joint-stock companies and stock investors. The volatility in the stock Market has big side effects on economic construction, so the stock market trend analysis and forecasting has great significance. Due to the stock price influence factors are many and complicated, it is difficult for researchers to carry on accurate prediction on stock market. The methods of time series of stock price forecasting are good choices. Stock prices is nonlinear and has the characteristics of the heteroscedasticity and high noise, so that the traditional models of time series are not good analysis methods of the forecast stock. The main work of this paper is to establish the regression prediction model based on chaos optimization algorithm and the multi-scale wavelet kernel v-support vector machine to achieve the purpose of more accurately predicting stock prices. Compared to single-scale wavelet kernel v-support vector machine, wavelet neural network and radial basis function ε-support vector machine, this model can more accurately predict the future trend of stock prices. Stock market participants with the help of this model can reduce investment risk and gain higher return on investment.In the first part of this paper, we summarizes the research progress of the stock price time series forecasting methods, especially the hybrid model based on artificial intelligence algorithms and wavelet theory; in the second part, we expounded the theoretical basis of support vector machines and wavelet theory. Based on the methods of constructing support vector machine kernel functions, the article focuses on the rationality of the single-scale wavelet kernel and multi-scale wavelet kernel as the kernel function of the v-support vector machine. Then, this paper analyzes the advantage of v-MWSVR to deal with nonlinear, high-noise characteristics of time series. There are so many parameters in this model that the commonly used parameter selection method-support vector machine cross-validation method is not suitable. CV algorithm has some shortcomings, such as big calculating, parameters space discretization, not enough optimizing accuracy, in particular the calculation of the amount of growth with the number of parameters. To solve this problem, this paper presents the chaos optimization algorithm as the parameter selection method of v-MWSVR.The specific research of this paper is:firstly use building index published by the Shanghai Stock Exchange to test the effect of chaos optimization method for v-MWSVR. In this paper, the results of several experiments to compare the particle swarm optimization, Chaos optimization method and chaotic particle swarm optimization model optimization prove the effectiveness and stability of the chaos optimization algorithm for the model. Chaotic particle swarm optimization method optimization effect curve shows that the particle swarm optimization has no further optimization to the particles from the chaos optimization method. Then, this article compares the prediction effects of the multi-scale wavelet kernel v-support vector machines, single-scale wavelet nuclear v-support vector machine, wavelet neural network and radial basis kernel function ε-support vector machine prediction for Shanghai Stock Index. In this experiment, the Shanghai index is divided into three stages. At each stage, prediction effect based on chaos optimization algorithm and the v-MWSVR makes better than the other three models.