shapfs

A python package makes feature selection by calculating the Shapley value

The example of shapfs:

Qucik usage guide

Installation

You can install shapfs with pip:

pip install shapfs

Build the class of fs (features selection)

fs=feature_selection(type)

type=1 if the machine learning task is classification prediction. type=0 if it is regression prediction

Fit fs

fs.fit(x,y,alpha)

x is the input feature, y is the label, and alpha is the ratio of the shaply value of the selected variable to the sum of all feature shapley values

Result

After you fit the fs, you can view the Shapely values of all variables through fs.feature_ranking. fs.result gets the desired variable

fs.feature_ranking

fs.result

Plot

You can also visualize the order of importance of features.

fs.feature_importance_plot()

fs.feature_beeswarm()