machine learning - How to feature select for algorithms without co-effs or feature importance? -
using python , scikit learn.
fairly new machine learning, i've managed built 2 layer ensemble of learners. i've used rfecv each algorithm far. doing has increased roc 8%.
but last 2 algorithms (naive bayes , knn), can't used rfecv. there other techniques able use determine feature importance based on specific algorithm? couldn't find example of using scikit feature selection them these algorithms. majority of them have been manual takes.
Comments
Post a Comment