alvarouc / polyssifier Goto Github PK

just like the tests for classification, implement test for 'polyr'

We need to integrate imputation as part of the pipeline where the parameters of the imputation method can be searched in the grid search approach

Upgrading from Pandas 0.18 to 0.2 caused a shape error in predict_proba.
The temporary fix in commit b17d13e limits this implementation to binary classification only

xgboost is consistently the best classifier in reported exercises

Current exclusive support of binary classification is limiting the use of the tool. Multi-class problems arise frequently and it would be nice to be able to use polyssifier for multiclass data.

• a convenient feature would be to limit classifiers to inherently multiclass
• the default may be one vs all for all classifiers that support this
• it would be nice to be able to automatically reweigh classes for imbalanced datasets

There is an issue with feature plotting when feature selection is turned on.

dask-searchcv has optimized sklearn's default gridsearch in both local and cluster computing

Currently we do not support regression problems. We need to report a regression report when the "label" is a continuous variable.