how to correct " Error in pro_spec (line 219), if(Classif​icationSVM​(models,Te​stSet(j,:)​)) " and "Error using ClassificationSVM (line 249) Use fitcsvm to train an SVM model."

my matlab code:

%build models
for k=1:numClasses
    %Vectorized statement that binarizes Group
    %where 1 is the current class and 0 is all other classes
    G1vAll=(GroupTrain==u(k));
    models = fitcsvm(TrainingSet,G1vAll);
end
%classify test cases
for j=1:size(TestSet,1)
    for k=1:numClasses
        if(ClassificationSVM(models,TestSet(j,:))) 
            break;
        end
end

[Walter Roberson]

Please post the complete error message, everything in red.

[Walter Roberson]

Notice by the way that you are overwriting all of the variable models for each of your classes. And notice that your inner loop of the "for j" is doing the same thing for every k value.

[l divya]

i got following errors while executing attached file:

Error using ClassificationSVM (line 249)
Use fitcsvm to train an SVM model.

Error in pro_spec (line 219)
if(ClassificationSVM(models,TestSet(j,:)))

Answer 1

[Walter Roberson]

You should be using

models{k} = fitcsvm(TrainingSet,G1vAll);

and

          if predict(models{k}, TestSet(j,:))

ClassificationSVM is a class, not a routine that tests for matches.

You can be more efficient, by asking for the prediction about the entire TestSet .

[l divya]

when i used the above code, got error like: Error using ClassificationSVM (line 249) Use fitcsvm to train an SVM model.

Error in pro_spec (line 219) if(ClassificationSVM(models{k},TestSet(j,:)))

[Walter Roberson]

As I said, "ClassificationSVM is a class, not a routine that tests for matches." Class in an object-oriented sense. It is not a routine which makes predictions about which class (in the dataset label sense) a particular set of data belongs to: you need predict() for that, and I already posted the appropriate line,

            if predict(models{k}, TestSet(j,:))

[l divya]

thank you

Answer 2

[sambath kumar]

HI everone

my matlab code is

%This while loop is the multiclass SVM Trick
            c1=(C==u(itr));
            newClass=c1;
            %svmStruct = svmtrain(T,newClass,'kernel_function','rbf'); % I am using rbf kernel function, you must change it also
            model = fitcsvm(T,newClass); %#ok<SVMTRAIN>
            classes = ClassificationSVM(model,tst);
        
            % This is the loop for Reduction of Training Set
            for i=1:size(newClass,2)
                if newClass(1,i)==0;
                    c3(k,:)=T(i,:);
                    k=k+1;
                end
            end

When i use this above code, i am getting error like this:

Error using ClassificationSVM (line 249)

Use fitcsvm to train an SVM model.

Error in multisvm (line 29)

classes = ClassificationSVM(model,tst);

Error in Detect (line 144)

result = multisvm(Train_Feat,Train_Label,test);

could you resolve this

[Walter Roberson]

I already described the reason, https://www.mathworks.com/matlabcentral/answers/424242-how-to-correct-error-in-pro_spec-line-219-if-classificationsvm-models-testset-j-and-err#comment_624975

which is to say that ClassificaitonSVM is not a classification routine and you need to use predict()

I notice that a few different people have made exactly the same mistake, which suggests they are all copying the same source code, but I have not been able to locate which code is being copied.