There are some feature selection methods in WEKA such as "CFSSubsetEval" and "ConsistencySubsetEval" that evaluate subsets based on a predefined search algorithm resulting in a final subset regardless of the user's preferred number of features. Can we consider them as instances of the filter method?
Farideh Bagherzadeh Khiabani
Definitely yes. They are both filter feature selection methods since they do not use any prediction model while evaluating the subset of features. Instead they both take advantage of intrinsic characteristics of data. In general, CFS uses a correlation measure such as symmetrical uncertainty and Consistency uses a consistency measure such as Liu`s measure.
1 votes
0 thanks
- Badges
- Science topic
More Pooia Lalbakhsh's questions See All
Similar questions and discussions