matlab神經(jīng)網(wǎng)絡(luò)30個(gè)案例分析源代碼demochapter

byliyang[faruto]@faruto'sStudio~ .cn/farutohttp:/ 選擇回歸預(yù)測分析最佳的SVM參數(shù)functioncloseallfunctioncloseall;format%%載入測試數(shù)據(jù)上證指數(shù)(1990.12.19-%數(shù)據(jù)是一個(gè)4579*6的double型的矩陣,%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 loadchapter14_sh.mat;%[m,n]=ts=tsx=sh(1:m-畫出原始上證指數(shù)的每日開盤數(shù)gridon;%數(shù)據(jù)預(yù)處理,ts=ts';tsx=tsx';mapminmax[TS,TSps]=將映射函數(shù)的范圍參數(shù)分別置為1和2TSps.ymin=1;TSps.ymax=%對(duì)ts[TS,TSps][TS,TSps]=%gridon;%對(duì)TS進(jìn)行轉(zhuǎn)置,以符合libsvmTS=mapminmax[TSX,TSXps]=將映射函數(shù)的范圍參數(shù)分別置為1和2TSXps.ymin=1;TSXps.ymax=%對(duì)tsx[TSX,TSXps]=%對(duì)TSX進(jìn)行轉(zhuǎn)置,以符合libsvmTSX=選擇回歸預(yù)測分析最佳的SVM參數(shù)%%首先進(jìn)行粗略選擇%c的變化范圍是2^(-5),2^(-%g的變化范圍是2^(-5),2^(-[bestmse,bestc,bestg]=SVMcgForRegress(TS,TSX,-5,10,-%str=sprintf('BestCrossValidationMSE=%gBestc=%gBestg=%g',bestmse,bestc,bestg);%根據(jù)粗略選擇的結(jié)果圖再進(jìn)行精細(xì)選擇%c的變化范圍是%g的變化范圍是2^(-2),2^(-[bestmse,bestc,bestg]=SVMcgForRegress(TS,TSX,0,10,-%str=sprintf('BestCrossValidationMSE=%gBestc=%gBestg=%g',bestmse,bestc,bestg);BestCrossValidationMSE= Bestc=0.25Bestg=BestCrossValidationMSE= Bestc=1Bestg=cmdcmd=['-c',num2str(bestc),'-g',num2str(bestg),'-s3-p0.01'];model=svmtrain(TS,TSX,cmd);%model=svmtrain(TS,TSX,'-s3-c1-g2-p[predict,mse]=svmpredict(TS,TSX,model);predict[predict,mse]=svmpredict(TS,TSX,model);predict=mapminmax('reverse',predict,TSps);%str=sprintf('均方誤差MSE=%g相關(guān)系數(shù)R=%g%%',mse(2),mse(3)*100);Meansquarederror=1.95029e-005Squaredcorrelationcoefficient=0.999345均方誤差MSE=1.95029e-005相關(guān)系數(shù)R=holdon;holdoff;gridon;%webhttp:/webhttp://forum-31-1.html-子函數(shù)functionfunction[mse,bestc,bestg]=%%輸入%train_label:訓(xùn)練 %train:訓(xùn)練集.要求與libsvm工具箱中要求一致%cmin:懲罰參數(shù)c的變化范圍的最小值(取以2為底的對(duì)數(shù)后),即c_min=2^(cmin).默認(rèn)為-%cmax:懲罰參數(shù)c的變化范圍的最大值(取以2為底的對(duì)數(shù)后),即c_max=2^(cmax).默認(rèn)為%gmin:參數(shù)g的變化范圍的最小值(取以2為底的對(duì)數(shù)后),即g_min=2^(gmin).默認(rèn)為-%gmax:參數(shù)g的變化范圍的最小值(取以2為底的對(duì)數(shù)后),即g_min=2^(gmax).默認(rèn)為%v:crossvalidation的參數(shù),即給測試集分為幾部分進(jìn)行crossvalidation.默認(rèn)為%cstep:參數(shù)c步進(jìn)的大小.默認(rèn)為%gstep:參數(shù)g步進(jìn)的大小.默認(rèn)為%msestep:最后顯示MSE圖時(shí)的步進(jìn)大小.默認(rèn)為%輸出%bestacc:CrossValidation%bestc:最佳的參數(shù)%bestg:最佳的參數(shù)%abouttheparametersofSVMcgForRegressifnargin<10msestep=ifnargin<msestep=v=cstep=gstep=ifnargin<msestep=v=cstep=gstep=gmax=ifnargin<msestep=v=cstep=gstep=gmax=gmin=-ifnargin<msestep=v=cstep=gstep=gmax=gmin=-cmax=ifnargin<msestep=v=cstep=gstep=gmax=gmin=-cmax=cmin=-%X:cY:g[X,Y]=meshgrid(cmin:cstep:cmax,gmin:gstep:gmax);[m,n]=size(X);cg=%recordaccuracywithdifferentc&g,andfindthebestmsewiththesmallestcbestc=0;bestg=0;mse=10^10;basenum=2;fori=1:mforj=cmd=['-v',num2str(v),'-c',num2str(basenum^X(i,j)),'-g',num2str(basenum^Y(i,j)),'-s3'];cg(i,j)=svmtrain(train_label,train,ifcg(i,j)<msemse=cg(i,j);bestc=basenum^X(i,j)