xszpo
My aspiration is to create a set of tool that filling gaps in machine learning process when you want to:
- use (sklearn.pipeline) Pipeline and FeatureUnion from the beginning to the end,
- use Pandas DataFrame,
- explain results using ELI5 package,
My perferct Pipe (sklearn.pipeline.Pipeline):
- DataPreparation
- IN: pd.DataFrame
- OUT: pd.DataFrame
- FeatureEngineering
- IN: pd.DataFrame
- OUT: np.array
- FeatureSelection
- IN: np.array
- OUT: np.array
- Model
- IN: np.array
- OUT: np.array
Explaining predictions:
- Interpret
- IN: Pipeline
- OUT: eli5.show_weights, eli5.show_prediction
Basic - modul with additional methods
pakiet wykorzystuje niektóre metody https://github.com/pjankiewicz/mercari-solution