Empirical Mode Decomposition-Support Vector Regression Model And Its Application In Stock Price Prediction

Stock has attracted people's attention since it emerged because of its high risk and high return.How to balance risk and return is a problem for people.As we hope to get more benefits from the stock market,predicting stock price came into being.With the development of China's market-oriented economy,the stock is becoming increasingly popular in China.Not only is there more and more civilians involved in stock trading,the stock market is also affecting the country's economy more and more.Therefore,whether it is for the country or the individual,doing research on the stock price forecast is very necessary.There is some correlation between stock prices and their historical data,so under some conditions,we can use the historical data to make some predictions about the future.At present,there are many methods for stock prediction,such as GARCH model,neural network model,support vector regression model,and so on.The support vector regression model is based on the principle of VC dimension theory and structural risk minimization theory.Starting from some sample data,it seeks a proper balance between the accuracy and simplicity of the model to find the global optimal solution.As a result,the prediction of high-dimensional nonlinear problems is finally achieved.The support vector regression model has many advantages,such as accuracy,high efficiency,ease of operation and so on.However,because the stock price sequence is often non-linear and non-stationary,the complexity of the sequence will still cause much trouble for the prediction.In order to improve the accuracy of stock price prediction,we hope to reduce the complexity of the stock price sequence through certain methods,and the empirical mode decomposition method is one way to solve this problem.The empirical mode decomposition method is a method specially used to analyze nonlinear and non-stationary signals.It has a wide range of applications in signal analysis and has the advantages of direct,intuitive,posterior,and adaptive.This method can decompose the signal without causing any signal loss.The intrinsic mode function and residue obtained by the decomposition can be reconstructed by simple addition.The empirical mode decomposition is used to optimize the support vector regression stock price forecasting model.First,the non-stationary and non-linear stock price sequence is decomposed into several simple intrinsic mode functions,and one residue.Then through the mean reconstruction algorithm,intrinsic mode functions are classified to obtain a high-frequency mode,a low-frequency mode,and a residue.The support vector regression method is used to analyze and predict these three sequences,and three prediction results are finally performed.Finally,add and reconstruct the three prediction results to get the final prediction result.After the complicated stock price sequence is decomposed by the empirical mode decomposition algorithm,the stability will enhance,and the structure of intrinsic mode function and the residue are relatively simple,so the simulation will be easier and the prediction results will be better.This paper describes in detail the support vector regression,empirical mode decomposition and combination of the two.Then,this paper uses the stock market volatility series of the Shanghai Composite Index from July 1,2016 to December 29,2017 as an example to conduct empirical tests,comparing the prediction of the support vector regression model and the improved support vector regression modal.The results confirmed our view.