python - How do I delete features from feature list (TfidfVectorizer)?

Question

Currently I am working on a text classification program. I elictied some features with the TfidfVectorizer. Now, I want to delete some words form the original feature list because the do not provide usefull information.

I got two questions:

1. Where are the features stored? ( via .get_feature_names() or in X_train_union_tfidf ?)
2. How can I delete them?

I found a quite similiar question here : Ignore a column while building a model with SKLearn but i cannot link it to my problem.

Code:

X_train_union_tfidf = combined_tfidf.fit_transform(X_train) 
X_test_union_tfidf  = combined_tfidf.transform(X_test)
print(feature_union_df_tfidf)


unigram__compris             15.844468
unigram__devic               16.797861
bigram__speech recognit      17.065831
bigram__invent relat         17.527465
bigram__present invent       21.158065

Lets say I want to delete the invent relat and present invent. How do i delete it from X_train_union_tfidf (sparse matrix) before passing it to the classification alogrithm?

naive_bayes = MultinomialNB()

naive_bayes.fit(X_train_union_tfidf, y_train)                                                            
predictions_NB_tfidf = naive_bayes.predict(X_test_union_tfidf) 
predicted_prob_NB_tf = naive_bayes.predict_proba(X_test_union_tfidf)                                             

question from:https://stackoverflow.com/questions/66061621/how-do-i-delete-features-from-feature-list-tfidfvectorizer

Answer 1

Waitting for answers