MATLAB: Error using predict function for SVM

fitrsvmMATLABpredictStatistics and Machine Learning Toolbox

Hi there i am trying to create a SVM model to predict the data from the one that i have attached here and the code
load mec_prop_tempcooling20sec
table_mec_prop = array2table(mec_prop_tempcooling20sec);
table_mec_prop.Properties.VariableNames{4} = 'UTS';
table_mec_prop.Properties.VariableNames{5} = 'EaB';
table_mec_prop.Properties.VariableNames{6} = 'YM';
table_mec_prop.Properties.VariableNames{1} = 'Heating';
table_mec_prop.Properties.VariableNames{2} = 'Build';
table_mec_prop.Properties.VariableNames{3} = 'Cooling';
Mdl_tensile_Gau = fitrsvm(table_mec_prop,'UTS~Heating+Build+Cooling','KernelFunction','gaussian','KernelScale','auto','Standardize',true,'Leaveout','on');
Mdl_ymodulus_Gau = fitrsvm(table_mec_prop,'YM~Heating+Build+Cooling','KernelFunction','gaussian','KernelScale','auto','Standardize',true,'Leaveout','on');
Mdl_extension_Gau = fitrsvm(table_mec_prop,'EaB~Heating+Build+Cooling','KernelFunction','gaussian','KernelScale','auto','Standardize',true,'Leaveout','on');
Mdl_tensile_lin = fitrsvm(table_mec_prop,'UTS~Heating+Build+Cooling','KernelFunction','linear','KernelScale','auto','Standardize',true,'Leaveout','on');
Mdl_ymodulus_lin = fitrsvm(table_mec_prop,'YM~Heating+Build+Cooling','KernelFunction','linear','KernelScale','auto','Standardize',true,'Leaveout','on');
Mdl_extension_lin = fitrsvm(table_mec_prop,'EaB~Heating+Build+Cooling','KernelFunction','linear','KernelScale','auto','Standardize',true,'Leaveout','on');
mseGau_tensile = kfoldLoss(Mdl_tensile_Gau);
mseGau_ymodulus = kfoldLoss(Mdl_ymodulus_Gau);
mseGau_extension = kfoldLoss(Mdl_extension_Gau);
mseLin_tensile = kfoldLoss(Mdl_tensile_lin);
mseLin_ymodulus = kfoldLoss(Mdl_ymodulus_lin);
mseLin_extension = kfoldLoss(Mdl_extension_lin);
Mdl_tensile = fitrsvm(table_mec_prop,'UTS~Heating+Build+Cooling','KernelFunction','linear','KernelScale','auto','Standardize',true,'Leaveout','on');
Mdl_ymodulus = fitrsvm(table_mec_prop,'YM~Heating+Build+Cooling','KernelFunction','linear','KernelScale','auto','Standardize',true,'Leaveout','on');
Mdl_extension = fitrsvm(table_mec_prop,'EaB~Heating+Build+Cooling','KernelFunction','linear','KernelScale','auto','Standardize',true,'Leaveout','on');
X = mec_prop_tempcooling20sec(:,1:3);
YSVMnew_tensile = predict(Mdl_tensile, X);
YSVMnew_ymodulus = predict(Mdl_ymodulus, X);
YSVMnew_extension = predict(Mdl_extension, X);
After running through the code i have obtained this error
Error using predict (line 84)
No valid system or dataset was specified.
Error in SVM_temp_regression_cooling20sec
(line 31)
YSVMnew_tensile = predict(Mdl_tensile, X);
The .mat file is attached below
Thanks in advance

Best Answer

I was able to run the code without error by removing the name value pair ('Leaveout', 'on) from the function call to "fitrsvm".
I saw in the documentation for "fitrsvm" that setting KFold, Holdout, Leaveout, CrossVal, or CVPartition creates a RegressionPartitionedSVM cross-validated model. Otherwise, it creates a RegressionSVM model.
The predict function is only used with "RegressionSVM" models. In the tips section, it says "If mdl is a cross-validated RegressionPartitionedSVM model, use kfoldpredict instead of predict to predict new response values." See that information at the following link:
https://www.mathworks.com/help/stats/compactregressionsvm.predict.html
Refer to the documentation page for "fitrsvm" for more information:
https://www.mathworks.com/help/stats/fitrsvm.html#input_argument_d0e363970
Related Question