回歸預(yù)測分析-SVM神經(jīng)網(wǎng)絡(luò)

%%SVM神經(jīng)網(wǎng)絡(luò)的回歸預(yù)測分析---上證指數(shù)開盤指數(shù)預(yù)測%%清空環(huán)境變量functionchapter14tic;closeall;clear;clc;formatcompact;%%數(shù)據(jù)的提取和預(yù)處理%載入測試數(shù)據(jù)上證指數(shù)(1990.12.19-2009.08.19)%數(shù)據(jù)是一個4579*6的double型的矩陣,每一行表示每一天的上證指數(shù)%6列分別表示當(dāng)天上證指數(shù)的開盤指數(shù),指數(shù)最高值,指數(shù)最低值,收盤指數(shù),當(dāng)日交易量,當(dāng)日交易額.loadchapter14_sh.mat;%提取數(shù)據(jù)[m,n]=size(sh);ts=sh(2:m,1);tsx=sh(1:m-1,:);%畫出原始上證指數(shù)的每日開盤數(shù)figure;plot(ts,'LineWidth',2);title('上證指數(shù)的每日開盤數(shù)(1990.12.20-2009.08.19)','FontSize',12);xlabel('交易日天數(shù)(1990.12.19-2009.08.19)','FontSize',12);ylabel('開盤數(shù)','FontSize',12);gridon;%數(shù)據(jù)預(yù)處理,將原始數(shù)據(jù)進(jìn)行歸一化ts=ts';tsx=tsx';%mapminmax為matlab自帶的映射函數(shù)%對ts進(jìn)行歸一化[TS,TSps]=mapminmax(ts,1,2);%畫出原始上證指數(shù)的每日開盤數(shù)歸一化后的圖像figure;plot(TS,'LineWidth',2);title('原始上證指數(shù)的每日開盤數(shù)歸一化后的圖像','FontSize',12);xlabel('交易日天數(shù)(1990.12.19-2009.08.19)','FontSize',12);ylabel('歸一化后的開盤數(shù)','FontSize',12);gridon;%對TS進(jìn)行轉(zhuǎn)置,以符合libsvm工具箱的數(shù)據(jù)格式要求TS=TS';%mapminmax為matlab自帶的映射函數(shù)%對tsx進(jìn)行歸一化[TSX,TSXps]=mapminmax(tsx,1,2);%對TSX進(jìn)行轉(zhuǎn)置,以符合libsvm工具箱的數(shù)據(jù)格式要求TSX=TSX';%%選擇回歸預(yù)測分析最佳的SVM參數(shù)c&g%首先進(jìn)行粗略選擇:[bestmse,bestc,bestg]=SVMcgForRegress(TS,TSX,-8,8,-8,8);%打印粗略選擇結(jié)果disp('打印粗略選擇結(jié)果');str=sprintf('BestCrossValidationMSE=%gBestc=%gBestg=%g',bestmse,bestc,bestg);disp(str);%根據(jù)粗略選擇的結(jié)果圖再進(jìn)行精細(xì)選擇:[bestmse,bestc,bestg]=SVMcgForRegress(TS,TSX,-4,4,-4,4,3,0.5,0.5,0.05);%打印精細(xì)選擇結(jié)果disp('打印精細(xì)選擇結(jié)果');str=sprintf('BestCrossValidationMSE=%gBestc=%gBestg=%g',bestmse,bestc,bestg);disp(str);%%利用回歸預(yù)測分析最佳的參數(shù)進(jìn)行SVM網(wǎng)絡(luò)訓(xùn)練cmd=['-c',num2str(bestc),'-g',num2str(bestg),'-s3-p0.01'];model=svmtrain(TS,TSX,cmd);%%SVM網(wǎng)絡(luò)回歸預(yù)測[predict,mse]=svmpred