Feature Selection

Installation

# To install the required package, use the following command:
# !pip install modeva

Authentication

# To get authentication, use the following command: (To get full access please replace the token to your own token)
# from modeva.utils.authenticate import authenticate
# authenticate(auth_code='eaaa4301-b140-484c-8e93-f9f633c8bacb')

Import modeva modules

from modeva import DataSet

Load data

ds = DataSet()
ds.load("BikeSharing")
ds.set_random_split()

Correlation based feature selection

results = ds.feature_select_corr(threshold=0.1, method="xicor")
results.plot()

XGB-PFI based feature selection

results = ds.feature_select_xgbpfi(threshold=0.01)
results.plot()

RCIT based feature selection

results = ds.feature_select_rcit()
results.plot()

Feature selection operations

Set selected features to be active

ds.set_active_features(features=results.value["selected"])
ds.feature_names

['season', 'yr', 'mnth', 'hr', 'weekday', 'workingday', 'weathersit', 'temp', 'atemp', 'hum', 'windspeed']

Conduct another round of feature selection

results = ds.feature_select_xgbpfi(threshold=0.1)
results.plot()

Apply another round of feature selection

ds.set_active_features(features=results.value["selected"])
ds.feature_names

['hr']

Revert all feature selection

ds.set_active_features(features=None) # by default, all features are set active
ds.feature_names

['season', 'yr', 'mnth', 'hr', 'holiday', 'weekday', 'workingday', 'weathersit', 'temp', 'atemp', 'hum', 'windspeed']