1/ Using random forest to calculate feature importance?
The importance score might be biased. #machinelearning #featureimportance Thanks @Matthew_N_B for pointing it out
A thread 👇
The importance score might be biased. #machinelearning #featureimportance Thanks @Matthew_N_B for pointing it out
A thread 👇
2/ explained.ai/rf-importance/
The takeaway from this article is that the most popular RF implementation in Python (scikit) and R's RF default importance strategy do not give reliable feature importances
The takeaway from this article is that the most popular RF implementation in Python (scikit) and R's RF default importance strategy do not give reliable feature importances
3/
when “... potential predictor variables vary in their scale of measurement or their number of categories.” (Strobl et al). Rather than figuring out whether your data set conforms to one that gets accurate results, simply use permutation importance.
when “... potential predictor variables vary in their scale of measurement or their number of categories.” (Strobl et al). Rather than figuring out whether your data set conforms to one that gets accurate results, simply use permutation importance.
4/
You can either use our Python implementation (rfpimp via pip) or, if using R, make sure to use importance=T in the Random Forest constructor then type=1 in R's importance() function. mljar.com/blog/feature-i…
You can either use our Python implementation (rfpimp via pip) or, if using R, make sure to use importance=T in the Random Forest constructor then type=1 in R's importance() function. mljar.com/blog/feature-i…
• • •
Missing some Tweet in this thread? You can try to
force a refresh
