Sentiment analysis with NLTK and Scikit-learn sklearn

sklearn.py

# train test split, we are not using test data set
from sklearn.model_selection import train_test_split

X_train, X_test, y_train, y_test = train_test_split(X_resampled, y_resampled, test_size=0.33, random_state=39)

print('X_train shape: {}'.format(X_train.shape))
print('X_test shape: {}'.format(X_test.shape))
print('y_train shape: {}'.format(y_train.shape))
print('y_test shape: {}'.format(y_test.shape))

# pick three algorithms Multinomial Naive Bayes, RandomForest and GradientBoosting
from sklearn.naive_bayes import MultinomialNB
from sklearn.ensemble import RandomForestClassifier, GradientBoostingClassifier

clf_mnb = MultinomialNB()
clf_rfc = RandomForestClassifier(random_state=39)
clf_gbc = GradientBoostingClassifier(random_state=39)
clf_names = ['MultinomialNB', 'RandomForest', 'GradientBoosting']
clf_types = [clf_mnb, clf_rfc, clf_gbc]

for (i, clf) in enumerate(clf_types):
    clf.fit(X_train, y_train.values.ravel())
    print('Result of {}\n'.format(clf_names[i]))
    predictions = clf.predict(X_test)
    print(classification_report(predictions, y_test))
    print('\n')
    print('Confusion matrix: \n', confusion_matrix(predictions, y_test))
    print('\n')
    print('Accuracy score: ', accuracy_score(predictions, y_test))
    print('\n\n\n')

# results were given
Result of MultinomialNB

              precision    recall  f1-score   support

           0       0.92      0.95      0.93      9546
           1       0.95      0.92      0.93     10070

    accuracy                           0.93     19616
   macro avg       0.93      0.93      0.93     19616
weighted avg       0.93      0.93      0.93     19616



Confusion matrix: 
 [[9044  502]
 [ 820 9250]]


Accuracy score:  0.9326060358890701




Result of RandomForest

              precision    recall  f1-score   support

           0       0.97      0.99      0.98      9632
           1       0.99      0.97      0.98      9984

    accuracy                           0.98     19616
   macro avg       0.98      0.98      0.98     19616
weighted avg       0.98      0.98      0.98     19616



Confusion matrix: 
 [[9525  107]
 [ 339 9645]]


Accuracy score:  0.977263458401305




Result of GradientBoosting

              precision    recall  f1-score   support

           0       0.95      0.73      0.82     12933
           1       0.64      0.93      0.76      6683

    accuracy                           0.80     19616
   macro avg       0.80      0.83      0.79     19616
weighted avg       0.85      0.80      0.80     19616



Confusion matrix: 
 [[9397 3536]
 [ 467 6216]]


Accuracy score:  0.7959318923327896