Saving and loading a Multiclass SVM Model

I want to save a trained multiclass SVM and then load it to make predictions. How to do this?

Answer 1

[Stephan]

Hi,

save the workspace with the trained model in it. If you want to make predictions in later matlab sessions just load the corresponding .mat-file with your model and add the new data to workspace for that you want to make predictions.

The prediction itself is done with the

label = predict(SVMModel,X)

command, where X is your data set on which you want to make a new prediction. See the documentation for predict command for examples how to do.

Best regards

Stephan

[NC]

Thank you very much. But I can't save the model with "saveCompactModel()" function. How to save it ?

[Stephan]

Hi,

this should work like this:

saveCompactModel(SVMModel, 'your_filename');

Where SVMModel is the name of your model and your_filename means the wanted name of the .mat-file containing your model. If you dont want it to be saved in the actual working folder, specify the full path like

'C:\Projects\SVM\my_svm_model'

Do you have full rights fo write into the matlab working folder?

See saveCompactModel documentation for more information.

[Ravi Singh]

Hi stephen,

The code you provided to save is not working. Do we have something else to save.