scikit-learn-contrib / boruta_py

Python implementations of the Boruta all-relevant feature selection method.
BSD 3-Clause "New" or "Revised" License
1.46k stars 252 forks source link

Why does the number of total features (Confirmed + Tentative + Rejected) not equal to the input features? #126

Open abhay0210 opened 1 month ago

abhay0210 commented 1 month ago

Total number of input feature = 681. forest = RandomForestClassifier(n_jobs=-1, class_weight='balanced',max_depth=5); feat_selector = BorutaPy(forest, n_estimators='auto', verbose=2, random_state=7242, max_iter = 50, perc = 95); feat_selector.fit(X, y); Sample output of a 50 iteration run: . . . Iteration: 47 / 50 Confirmed: 567 Tentative: 79 Rejected: 35 Iteration: 48 / 50 Confirmed: 567 Tentative: 79 Rejected: 35 Iteration: 49 / 50 Confirmed: 568 Tentative: 77 Rejected: 36

BorutaPy finished running.

Iteration: 50 / 50 Confirmed: 568 Tentative: 38 Rejected: 36

Till 49th iteration, Confirmed+ Tentative+ Rejected = 681. But on the final iteration this breaks. The number of tentatives drop drastically. Why is that?