Support Vectors are wrong in my SVR model

Question

hu yang 2016-8-28

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/301158-support-vectors-are-wrong-in-my-svr-model

回答： Jo 2023-10-26

target.csv

I trained an SVR model by fitrsvm. Since I need to get all support vector I have checked the model ouput. Surprisingly, the mdl.SupportVectors have nothing in common with my input mdl.X. Can anyone tell me why.

Below is my code. sorry the csv files are too larged to be uploaded.

    input = csvread('sample.csv');
  target = csvread('target.csv');
  Y = log(target(2,:)+1)';
  %Y = log(target(2,:)+1)';
  %turbidity = log(target(1,:)+1);
  turbidity = target(1,:);
  mainComp = 15;
  s = load('eigenVector.mat','s');
  s = s.s;
  v = load('singularValue.mat','v');
  v = v.v;
a = s'*input;
b = inv(v(1:mainComp,1:mainComp))*a(1:mainComp,:); 
X = [b;log(turbidity+1)];
mdl = fitrsvm(X',Y,'Standardize',true,'KernelFunction','polynomial','KernelScale','auto','OutlierFraction',0.05);
save mySVMMDL mdl;
yfit = exp(predict(mdl,X'))-1;

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Jo 2023-10-26

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/301158-support-vectors-are-wrong-in-my-svr-model#answer_1341011

Hello,

I know this post is old but maybe my answer can help someone that may have the same question ?

The data stored in "mdl.SupportVectors" are preprocessed data. As you have set the option 'standardized','true', these data are standardized. So you need to standardized manually "Mdl.X" to be able to compare both matrices.