Failed to Call Classification Learner's Testing Function

I was using a Matlab R2015b's Classification Learner Toolbox. I was successful in importing file data and export it into an Export Model, and i got a structure named trainedClassifier.

Import process #1

Import process #2

Training process with PCA implemented & Multi Class SVM (One vs All validation)

trainedClassifier variable generated from ToolBox

fetureVector variable which used for testing

yfit = trainedClassifier.predictFcn(featureVector)

After it, i want to doing a test with a new data with this code (i got this code from here ) :

>> yfit = trainedClassifier.predictFcn(featureVector)

Then i got an error output as a follows :

Function 'subsindex' is not defined for values of class 'cell'.
Error in mlearnapp.internal.model.DatasetSpecification>@(t)t(:,predictorNames) (line 135)
extractPredictorsFromTableFcn = @(t) t(:,predictorNames);
Error in  mlearnapp.internal.model.DatasetSpecification>@(x)extractPredictorsFromTableFcn(splitMatricesInTableFcn(convertMatrixToTableFcn(x)))
(line 136)
extractPredictorsFcn = @(x) extractPredictorsFromTableFcn(splitMatricesInTableFcn(convertMatrixToTableFcn(x)));
Error in mlearnapp.internal.model.DatasetSpecification>@(x)exportableClassifier.predictFcn(extractPredictorsFcn(x)) (line 137)
exportableClassifier.predictFcn = @(x) exportableClassifier.predictFcn(extractPredictorsFcn(x));

What is the problem and solutions?

Thanks in advance.

[Walter Roberson]

What is the data type of featureVector4 ?

The error message is saying that something is being indexed with a value that is a cell array.

[Angga Lisdiyanto]

A 1 x 64 matrix.

[Angga Lisdiyanto]

Is this a bug?

http://www.mathworks.com/support/bugreports/1300151?s_tid=answers_rc2

[Walter Roberson]

That report does not appear to be relevant.

[Angga Lisdiyanto]

Before geting featureVector4 data from a loop, i declare that variable as featureVector4 = [].

Is this the reason?

[Walter Roberson]

No, that should be fine.

Could you show the output of

which -all table

?

[Angga Lisdiyanto]

Off course.

Import process #1

Import process #2

Training process with PCA implemented & Multi Class SVM (One vs All validation)

trainedClassifier variable generated from ToolBox

fetureVector variable which used for testing

yfit = trainedClassifier.predictFcn(featureVector)

[Angga Lisdiyanto]

Hello, is someone know the solution?

Thanks

[Angga Lisdiyanto]

Help...help...help... :(

[PAVITHRA S]

i tried the above code to test my trained network(classiification learner app). i am unable to execute the code

VarNames = arrayfun(@(N) sprintf('VarName%d',N), 1:512, 'Uniform', 0);

FV_table = array2table( featureVector, 'VariableNames', VarNames);

yfit = trainedClassifier.predictFcn(FV_table)

can u suggest me a solution to test.

[Mrutyunjaya Hiremath]

'testingData.xlsx' contains only 512 colums feature vector of tesing data or matrix of N X 512.

testingData = xlsread('testingData.xlsx');
yFit = trainedClassifier.predictFcn(testingData);

Answer 1

[Walter Roberson]

You have

yfit = trainedClassifier.predictFcn(featureVector)

and my reading just now suggest that perhaps needs to be

yfit = predict(trainedClassifier, featureVector)

The error message indicates that something is trying to be indexed using a cell array as a subscript, which is a valid indexing method for tables but not an indexing method for a double array.

[Angga Lisdiyanto]

Thanks, but there are error again.

Error using predict (line 84)
Systems of struct class cannot be used with the "predict" command. Convert the system to an identified model first, such as by
using the "idss" command.

Please help me again. :(

[Walter Roberson]

I am not certain.

I see in http://www.mathworks.com/help/stats/export-classification-model-for-use-with-new-data.html#bu4764j-1 that they use the syntax you had used before, not the one I suggested here, so my suggestion was incorrect. However, I notice the documentation there says

"Supply the data T in same data type as your training data used in the app (table or matrix).

If you supply a table, ensure it contains the same predictor names as your training data. The predictFcn ignores additional variables in tables. Variable formats (e.g. matrix or vector, data type) must match the original training data."

You imported your data from xlsx and your data had at least one non-numeric field, so I suspect that your classifier was trained against a table (which is a MATLAB datatype), and so you will now need to predict against a table . That is, instead of the numeric featurevector you would need a table with field names VarName1, VarName2, and so on.

Try

VarNames = arrayfun(@(N) sprintf('VarName%d',N), 1:512, 'Uniform', 0);

FV_table = array2table( featureVector, 'VariableNames', VarNames);

yfit = predict(trainedClassifier, FV_table)

[Angga Lisdiyanto]

Thanks for your help.

But your code is still error :

Please help me again...

[Angga Lisdiyanto]

It works!!!! Huurraaayyyy!!!!

Yess yess yes yes yes oh yes hahahahha

It works with this code :

VarNames = arrayfun(@(N) sprintf('VarName%d',N), 1:512, 'Uniform', 0);
FV_table = array2table( featureVector, 'VariableNames', VarNames);
yfit = trainedClassifier.predictFcn(FV_table)

Thank you very much sir. :)

[Walter Roberson]

Yes, sorry, I copied the wrong prediction line to edit.

[Betzalel Fialkoff]

HELP, i've coping the above code, my model has only 4 predictors, but im getting an error. this is my code

VarNames = arrayfun(@(N) sprintf('VarName%d',N), 1:4, 'Uniform', 0);
FV_table = array2table( featureVector, 'VariableNames', VarNames);
yfit = m1.predictFcn(FV_table)

" Error using ' (line 421) Undefined function 'ctranspose' for input arguments of type 'table'."

I need this for an interview in in 12 hours, please help!!!

[Anna Gerald]

Please help in resolving these errors

Error using classreg.learning.internal.table2PredictMatrix>makeXMatrix (line 96) Table variable VarName1 is not a valid predictor.

Error in classreg.learning.internal.table2PredictMatrix (line 47) Xout = makeXMatrix(X,CategoricalPredictors,vrange,pnames);

Error in classreg.learning.regr.CompactRegressionTree/predict (line 557) X = classreg.learning.internal.table2PredictMatrix(X,[],[],...

Error in mlearnapp.internal.model.coremodel.TrainedRegressionTree>@(x)predict(RegressionTree,x) (line 46) functionHandle = @(x) predict(RegressionTree, x);

Error in mlearnapp.internal.model.transformation.TrainedManualFeatureSelection>@(x)decoratedPredictFunction(featureSelectionFunction(x)) (line 61) functionHandle = @(x) decoratedPredictFunction(featureSelectionFunction(x));

Error in mlearnapp.internal.model.DatasetSpecification>@(x)exportableModel.predictFcn(predictorExtractionFcn(x)) (line 167) newExportableModel.predictFcn = @(x) exportableModel.predictFcn(predictorExtractionFcn(x));

Error in copy (line 3) yfit = trainedModel1.predictFcn(FV_table)

[Hussein Hasan Mohsen]

Please some one help me with this?

[Mrutyunjaya Hiremath]

Refer this

FV_table = array2table(featureVector, 'VariableNames',{'CarbonDioxide', 'CarbonMomoxide'. ...});

[Hussein Hasan Mohsen]

still giving the same error

VarNames = arrayfun(@(N) sprintf('CarbonDioxide%', 'CarbonMonoxide%', 'Ethane%', 'Ethylene%', 'Hydrogen%', 'Methane%', 'Nitrogen%', 'Oxygen%',N), 6:9, 'Uniform', 0);
FV_table = array2table(featureVector, 'VariableNames',{'CarbonDioxide', 'CarbonMonoxide', 'Ethane', 'Ethylene', 'Hydrogen', 'Methane', 'Nitrogen', 'Oxygen'});
yfit = trainedClassifier.predictFcn(FV_table)

[Hussein Hasan Mohsen]

its workingggggggggggg by using this fuction

yfit = trainedClassifier.predictFcn(featureVector)

[Hussein Hasan Mohsen]

can some one help me in how to show the accuarcy of testing data

need code......

[Mrutyunjaya Hiremath]

Refer This

cpVal = classperf(groundTruth,classifierOutput);
fprintf('accuarcy of testing data %f', cpVal.CorrectRate);

[Hussein Hasan Mohsen]

it show me this error

>> cpVal = classperf(groundTruth,classifierOutput);
fprintf('accuarcy of testing data %f', cpVal.CorrectRate);

Not enough input arguments.

Error in groundTruth (line 289)

this.DataSource = dataSource;

and here is line 289 to 292

  this.DataSource         = dataSource;
            this.LabelDefinitions   = vision.internal.labeler.validation.checkLabelDefinitions(labelDefs);
            this.LabelData          = vision.internal.labeler.validation.checkLabelData(labelData, this.DSource, this.LabelDefinitions);
        end

[Walter Roberson]

Notice that it is reporting a problem inside a function named groundTruth, but that classperf expects its first argument to be a variable which is interpreted as ground truth -- the suggested variable in the documentation for classperf is groundTruth but that is data, not a reference to the function groundTruth

https://www.mathworks.com/help/bioinfo/ref/classperf.html#mw_fffd22d1-500b-4386-be0c-254367a7bb58

https://www.mathworks.com/help/vision/ref/groundtruth.html

So you need to assign something to groundTruth before you make that call to classperf()

[NN]

It worked for me also , thank you so much

[NN]

how do i compare this Yfit data with trained data and find the rmse value?Kindly help

[Fatma HM]

NN Did you found the answer? Plz I need it too

[刘 相志]

Thank you, bro. it helps me a lot :)

Answer 2

[naishi feng]

it works!!thanks!!!

Answer 3

[Jingwei Too]

you may have a look on this toolbox https://www.mathworks.com/matlabcentral/fileexchange/71461-simple-machine-learning-algorithms-for-classification?s_tid=prof_contriblnk