CODEWITHSUNDEEP
machine-learningweka
gstar2002
gstar2002

Reputation: 495

How to experiment with different feature set with Weka

I have a classification problem, to solve this problem I am planing to use different feature sets, say unigramms and bigramms, maybe more late on. And I want to experiment with different combination of those feature sets. What is the best way to do it? Do I have to prepair different file for different combination of feature sets?

Upvotes: 1

Views: 429

Answers (1)

etov
etov

Reputation: 3032

An easy way would be to use the "RemoveByName" filter, to filter out attributes based on a regular expression. So if, for instance, unigram feature names start with "uni_" and bigrams start with "bi_", you can use RemoveByName -E ^uni_ to leave only bigrams. You can perform this via the GUI either as a preprocessing step, or using a meta "FilteredClassifier".

However, I can't think of an automated way to try different combinations from within the WEKA.

Upvotes: 1

Related Questions