SVM.rar

close all; clear all; clc; dataset=xlsread('crop.xlsx','sheet1','A2:G101') data1=dataset'; data=mapminmax(data1,0,1); dataset=data'; num=2; xlen=length(dataset) cnt=25; myrmse=zeros(4,4); myrmse1=zeros(2,100); for l=1:5 c=cvpartition(xlen,'holdout',0.3); trind=training(c); tsind=test(c); for i=1:5 for j=1:5 for k=1:5 mdl=fitrsvm(dataset(trind,1:6),dataset(trind,7),'epsilon',2^(k-7),... 'KernelFunction','rbf', 'KernelScale',i, 'BoxConstraint',j); res=predict(mdl,dataset(tsind,1:6)); res1=predict(mdl,dataset(trind,1:6)); temprmse=sqrt(mean((res-dataset(tsind,7)).^2)); myrmse(cnt,:)=[i j k temprmse]; myrmse1(cnt,:)=[tsind']; cnt=cnt+1 end end end end save cropyeild myrmse1 myrmse [a b]=min(myrmse(25:274,4)); b=b+24; i=myrmse(b,1); j=myrmse(b,2); k=myrmse(b,3); tsind=find(myrmse1(b,:)); trind=setdiff(1:xlen,tsind); mdl=fitrsvm(dataset(trind,1:6),dataset(trind,7),'epsilon',2^(k-7),... 'KernelFunction','rbf', 'KernelScale',i, 'BoxConstraint',j); res=predict(mdl,dataset(tsind,1:6)); res1=predict(mdl,dataset(trind,1:6)); myrmsefin= sqrt(mean((res-dataset(tsind,7)).^2)); plotregression(dataset(tsind,7),res)