Stock Trading Strategy Based On Svm With LDA And Wavelet Analysis

With the issuance of Dutch East India Company's stock in Amsterdam,stock,as a subject of investment,has come out for more than 400 years.Countless investors have been tossing and turning restlessly for the volatility of stock prices.But there is still no solution to predict the price of stock accurately.Because the stock price fluctuations are highly nonlinear,the accuracy is very low if we establish mathematical prediction model through traditional methods.With the development of information technology,machine learning,as an important part of artificial intelligence,has also developed greatly.SVM proposed in the 1990 s is a machine learning theory based on statistical theory and structural risk minimization principle.SVM shows excellent performance in dealing with high dimensional problems and nonlinear complex problems,by using kernel function to avoid the computation of inner product in low dimensional space,which makes the nonlinear problem skillfully transformed into the linear problem of higher dimension space,thus simplifying the computation.The core idea of the final model is as follows: firstly,data is imported into model.Then data cleaning is performed.Furthermore,wavelet analysis is used to denoise the data set.Secondly,the data is normalized or standardized.Thirdly,in order to reduce dimension,use LDA to process data.Fourthly,the data is respectively imported into the SVM model to train the model.In the parameter tuning of the model,using cross-validation to narrow the grid search and perform full traversal to find the optimal parameter combination in SVM model.Fifthly,this paper use the constructed forecasting model and new data to verify the next period's stock price fluctuations.Finally,this paper evaluates the performance of the model.The classification value predicted by the model indicates the stock price of the next period is expected to rise or fall.If the model predicts the stock price in the next period will rise,stock trading strategy will hold SSE.If the model predicts that the stock price in the next period will not rise,stock trading strategy will sell SSE.By comparing Sharpe Ratio of investing in SSE Strategy,Sharpe Ratio of SVM based on PCA,Sharpe Ratio of SVM based on LDA,Sharpe Ratio of SVM based on LDA and wavelet analysis,it is found that SVM based on LDA and wavelet analysis is most valuable to guide investment.