Skip to content
/ xtoy Public

Automated Machine Learning: go from 'X' to 'y' without effort.

License

Notifications You must be signed in to change notification settings

kootenpv/xtoy

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

63 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Build Status

XtoY

pip install xtoy

Go from 'X' to 'y' without effort.

from sklearn.datasets import load_diabetes
from xtoy.toys import Toy
X, y = load_diabetes(return_X_y=True)
toy = Toy()
toy.fit(X[:300], y[:300])
toy.predict(X[300:])

Tries to minimize time-to-first-model

And a reasonable one at that.

Check how important each variable is:

# names of variables are numbers - only in this example - otherwise usually strings
toy.best_features_()
[(0.02541263748358529, 4),
 (0.03964045497300279, 6),
 (0.04000655539791701, 5),
 (0.047171804294566556, 0),
 (0.05355633793403717, 1),
 (0.05598481754558562, 9),
 (0.06349342396487742, 3),
 (0.09050228976499292, 7),
 (0.28327316154993126, 2),
 (0.3009585170915041, 8)]

For further inspection, have a look at the pipeline and how important each variable is:

# toy.best_pipeline_

Guarantee

The goal will be to accept ANY data and come up with a "sensible" prediction.

If your dataset doesn't work (asymptotically not happening), post an issue.

Test driven

Quality guarantee by testing code changes, with loss measurements on lots of data problems.

Features

  • ✓ Takes care of encoding text, categorical, dates (several features), continuous
  • Considers data size (small data -> feature engineering, big data -> feature selection)
  • ✓ Takes care of missing values
  • ✓ Creates a model
  • ✓ Optimizes model parameters
  • ✓ Gives you a first prediction
  • ✓ Contains a RegexVectorizer

Roadmap

  • More customizability
  • Tree-based data (being able to exclude grouped variables quickly)

About

Automated Machine Learning: go from 'X' to 'y' without effort.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages