How to use loss() function in SVM trained model

I have 40*5 matrix 'tsetALLstd.CSV', file 1 to 4 columns represent as a predictor and 5th column represents response variables. I have attached file in here, I used machine learning toolbox for train SVM model using that 'tsetALLstd.CSV' file and I export it to my Matlab workspace.I need to calculate classification error in SVM how should I do it?

Answer 1

[Stephan]

Hi

here you find an example with step by step how to for your problem:

https://de.mathworks.com/help/stats/compactclassificationsvm.loss.html?s_tid=doc_ta

Best regards

Stephan

[Asanka Dewappriya]

thank you I followed that but I got an error.

L = loss(SVMModel,TBL,ResponseVarName)

SVMModel = I put my trained SVM model

TBL = I put 40*5 matrix 'tsetALLstd.CSV'

ResponseVarName = what should I put ResponceVarName?

[Stephan]

In your case the name of the 5th column

[Asanka Dewappriya]

how should I name 5th column?

I try to below ways I got an error

[Stephan]

Please send your .mat file containing O and trained model. Also please send your code you have tried.

Best regards

Stephan

[Stephan]

Hi,

try with attached file...-

Load this file into Workspace and then type:

>> SVMModel = fitcecoc(tsetALLstd, 'VarName5')

SVMModel = 

    ClassificationECOC
             PredictorNames: {'VarName1'  'VarName2'  'VarName3'  'VarName4'}
               ResponseName: 'VarName5'
      CategoricalPredictors: []
                 ClassNames: [0.0    1.0    2.0    3.0]
             ScoreTransform: 'none'
             BinaryLearners: {6×1 cell}
                 CodingName: 'onevsone'

    Properties, Methods

>> CV = crossval(SVMModel)

CV = 

    classreg.learning.partition.ClassificationPartitionedECOC
      CrossValidatedModel: 'ECOC'
           PredictorNames: {'VarName1'  'VarName2'  'VarName3'  'VarName4'}
             ResponseName: 'VarName5'
          NumObservations: 40
                    KFold: 10
                Partition: [1×1 cvpartition]
               ClassNames: [0.0    1.0    2.0    3.0]
           ScoreTransform: 'none'

    Properties, Methods

>> Model_loss = kfoldLoss(CV)

Model_loss =

       0

Best regards

Stephan