ml-feature-selector 0.1.2

This package provides code for optimal feature selection using forward and backward wrapper-based methods. It also generates an Excel report that captures each step and result of the feature selection process, offering clear insights and explanations of the selected features.

Feature Selection

This repository contains code to perform wrapper-based feature selection (i.e. forward and backward).

Install requirements:

pip install -r requirements.txt

How to use?

Load model selection class from the package

from src.featureSelector import FeatureModelSelection

FeatureModelSelection(model=model,
                      scorer=r2_score,
                      X_train=X_train,
                      y_train=y_train,
                      higher_good=False,
                      )

model:   sklearn BaseEstimator(Machine Learning Models)(Eg: KNeighborsRegressor())
scorer:  Scorer like r2 square, % error or Custom Scorer
X_train: np.ndarray (Training input data (X,d), 
y_train: np.ndarray (Target Labels)
higher_good: Boolean (True when the performance is better for higher values and False when the performance is better for lower value)

Find the best features using the forward or backward feature selection method using feature_selection_type=('forward' or 'backward')

best_features = model_selection.find_best_features(feature_selection_type='forward')

Save each step of the feature selection method in an Excel file.

model_selection.feature_selection_tabularize(best_features, './feature_selected.xlsx')

Find the example at main.py