K-fold cross-validation using Exhaustive Grid Search. Cross-validation using Scikit Learn's Grid Search.

Cross-Validation.txt

#----------- CROSS-VALIDATION WITH GRID SEARCH ------------------------#

# Naive Bayes Classifier

# combine features
features_tfidf = features_union(tfidf)

# instantiate pipeline object
nb_pipeline = Pipeline([('feats', features_tfidf),  ('clf', MultinomialNB())])

# parameter grid (3x3x2x2x3x3x2) combinations
parameters = {
    'feats__text__tfidf__max_df': (0.5, 0.75, 1.0),
    'feats__text__tfidf__ngram_range': ((1, 1), (1, 2), (2, 2)), 
    'feats__text__tfidf__use_idf': (False, True),
    'feats__text__tfidf__binary':(False, True),
    'feats__text__tfidf__binary':('l1', 'l2', None),
    'clf__alpha': (1.0, 5.0, 10.0),
    'clf__fit_prior': (True, False),     
}

# instantiate GridSearchCV object with pipeline and parameters with 3-folds cross-validation
nb_grid = GridSearchCV(nb_pipeline, parameters, cv=3) # this takes a while :/

# start time 
nb_start = time.time()

# Fit 
nb_grid.fit(X_train, y_train)

# end time 
svm_end = time.time()
print(f"Time taken to run: {round((nb_end - nb_start)/60,1)} minutes")

# Check score
nb_grid.score(X_test, y_test)