Getting startedΒΆ
The key idea behind ReliefE is simplicity. The purpose of this library is to offer off-the-shelf functionality not supported elsewhere, with minimal user effort. The data used in the example is accessible at: https://github.com/SkBlaz/reliefe/tree/master/data
The minimal example is given next.
import reliefe
import scipy.io as sio
import numpy as np
# Load the data first
mat_obj = sio.loadmat("data/mcc/chess.mat")
x = mat_obj['input_space'] ## scipy csr sparse matrix (or numpy dense)
y = mat_obj['target_space'] ## scipy csr sparse matrix (or numpy dense)
# Fully fledged ReliefE (with all functionality)
reliefE_instance = reliefe.ReliefE() # Initialize default ReliefE
reliefE_instance.fit(x, y) # Compute rankings
print(reliefE_instance.feature_importances_) # rankings for features (same order as x)
Returns the following ranking:
[ 9.89309972e-003 1.20427974e-002 1.34322081e-002 9.30750360e-003
4.11818629e-003 1.60451980e-002 3.65662586e-003 4.95127759e-003
-1.21291185e+304 -1.21724878e+304 -5.00084339e+302 -1.19073976e+293
3.64385300e+226 -3.60334541e+266 -3.85113882e+293 1.38678002e-003
-1.21291185e+304 -1.21724878e+304 -1.12834143e+303 -2.91441558e+302
-3.06704357e+247 1.81048784e+303 1.65435470e+303 1.65414412e+303
-1.21291185e+304 -1.21724542e+304 0.00000000e+000 -2.91441558e+302
-1.57470695e+298 1.31441274e+302 -2.39764884e+301 -1.83316404e+302
-1.21290833e+304 -1.21724481e+304 0.00000000e+000 -2.91441558e+302
-3.15257597e+298 -2.46423553e+301]