Add more feature selection operators

EpistasisLab / tpot

A Python Automated Machine Learning tool that optimizes machine learning pipelines using genetic programming.

http://epistasislab.github.io/tpot/

GNU Lesser General Public License v3.0

9.75k stars 1.57k forks source link

Add more feature selection operators #45

Closed rhiever closed 8 years ago

rhiever commented 9 years ago

Add more feature selection operators from sklearn:

[x] VarianceThreshold: http://scikit-learn.org/stable/modules/generated/sklearn.feature_selection.VarianceThreshold.html#sklearn.feature_selection.VarianceThreshold
[x] SelectKBest: http://scikit-learn.org/stable/modules/generated/sklearn.feature_selection.SelectKBest.html#sklearn.feature_selection.SelectKBest
[x] SelectPercentile: http://scikit-learn.org/stable/modules/generated/sklearn.feature_selection.SelectPercentile.html#sklearn.feature_selection.SelectPercentile
[x] Recursive Feature Elimination: http://scikit-learn.org/stable/modules/generated/sklearn.feature_selection.RFE.html#sklearn.feature_selection.RFE

bartleyn commented 9 years ago

Would these be options that fall under the _dt_feature_selection method?

rhiever commented 9 years ago

Nope - these will be separate functions.

On Friday, December 4, 2015, Nathan notifications@github.com wrote:

Would these be options that fall under the _dt_feature_selection method?

— Reply to this email directly or view it on GitHub https://github.com/rhiever/tpot/issues/45#issuecomment-162125985.

Randal S. Olson, Ph.D. Postdoctoral Researcher, Institute for Biomedical Informatics University of Pennsylvania

E-mail: rso@randalolson.com | Twitter: @randal_olson https://twitter.com/randal_olson http://www.randalolson.com

bartleyn commented 9 years ago

Right, duh, because dt stands for decision tree. These operators would then act as alternatives to _dt_feature_selection. I'd be interested in taking a stab at it tonight/this weekend.

rhiever commented 9 years ago

Awesome. I'll keep an eye out for the PR!

On Friday, December 4, 2015, Nathan notifications@github.com wrote:

Right, duh, because dt stands for decision tree. These operators would then act as alternatives to _dt_feature_selection. I'd be interested in taking a stab at it tonight/this weekend.

— Reply to this email directly or view it on GitHub https://github.com/rhiever/tpot/issues/45#issuecomment-162132367.

Randal S. Olson, Ph.D. Postdoctoral Researcher, Institute for Biomedical Informatics University of Pennsylvania

E-mail: rso@randalolson.com | Twitter: @randal_olson https://twitter.com/randal_olson http://www.randalolson.com

rhiever commented 8 years ago

@bartleyn, just wanted to ping you about this one to see if you're working on it. If not, I'll likely try to tackle this issue this week.

bartleyn commented 8 years ago

Was actually about to submit my PR -- I've gotten all four to work as additional operators, but could probably use some optimization (no caching, etc). Also could probably abstract a couple additional things.

rhiever commented 8 years ago

Woohoo! :+1:

In your PR, please note what operators are slow (and how they can be optimized) and how else the code can be improved so we can file them as PRs or fix them on the spot.

Cheers!