model_stacking_package

Python package for model stacking

Overview

Repository contains Python package with implementation of stacking for sklean classification models based on this Kaggle kernel.

Package Contents

setup.py - package setup file.
stacking - folder, which contains package source code.
init.py - init file, which contains shortcuts for SklearnHelper class and get_oof function.
stacking_implementation.py - package source code.

Package Installation

Download package.
Navigate to folder, containing package (specifically setup.py file).
Type

pip install .

Package Usage Example

Example of usage of package SklearnHelper class and get_oof function

#import package
import stacking

#import sklearn model
from sklearn.ensemble import RandomForestClassifier

#define parameters for sklearn model
rf_params = {
    'n_jobs': -1,
    'n_estimators': 100,
    'max_depth': 6,
    'min_samples_leaf': 2,
    'max_features' : 'sqrt',
    'verbose': 0
}

#wrap model with SklearnHelper class from package
rf = SklearnHelper(clf=RandomForestClassifier, seed=42, params=rf_params)

#perform out of fold training of the model
rf_oof_train, rf_oof_test = get_oof(rf, X_train, y_train, X_test)

References