Text Classification

kaalgota

I trained classification model using SVM. For testing I passed data to my trained model. If keywords of testing dataset will not match with trained classifier then what will be the prediction of my classifier? In other sense, what will be prediction for my testing dataset.

MariusHelf

If you do text classification you probably used Process Documents for both your training data and your test data. Here it is important that both operators are setup completely equal. Then to make sure that you have the same keywords in both datasets you have to connect the wordlist output of the training operator to the respective input of the testing operator. That will remove all keywords from the test data that are not part of your training data. The SVM cannot make any use of the anyway since it has never seen those terms.

Best regards,
Marius