search

TeamHG-Memex / eli5

A library for debugging/inspecting machine learning classifiers and explaining their predictions
http://eli5.readthedocs.io
MIT License
2.76k stars 332 forks source link

unhashing: sign of a feature can be confusing in case of collisions #19

Closed kmike closed 7 years ago

kmike commented 7 years ago

A follow-up to https://github.com/TeamHG-Memex/eli5/issues/10 and https://github.com/TeamHG-Memex/eli5/issues/18: when deciding if a feature should be in top positive or in top negative features we should take in account sign of the most popular term, e.g. instead of

(-)people | considered | approximately +1.739 (as it is now)

it should be better to show

people | (-)considered | (-)approximately -1.739

lopuhin commented 7 years ago

Yeah, that's a great idea! Would make feature report much easier to read.

lopuhin commented 7 years ago

See #20 , but now I'm not sure it's sufficient

lopuhin commented 7 years ago

Ah, no, it's enough - I was confused because I was getting feature_names myself and forgot about always_signed.

So should be fixed in #20.

kmike commented 7 years ago

20 works great, thanks!