classifier_tfidf.fit(training_set_tfidf, labels_array)
###########
# testing #
###########
# issue predictions
predictions_degree = classifier_degree.predict(testing_set_degree)
predictions_w_degree = classifier_w_degree.predict(testing_set_w_degree)
predictions_closeness = classifier_closeness.predict(testing_set_closeness)
predictions_w_closeness = classifier_w_closeness.predict(testing_set_w_closeness)
predictions_twicw = classifier_twicw.predict(testing_set_twicw)
predictions_tfidf = classifier_tfidf.predict(testing_set_tfidf)
print('========== accuracy for', clf ,'classifier ==========')
print("accuracy TW-IDF degree:", round(metrics.accuracy_score(truth_array,predictions_degree)*100,3))
print("accuracy TW-IDF weighted degree:", round(metrics.accuracy_score(truth_array,predictions_w_degree)*100,3))
print("accuracy TW-IDF closeness:", round(metrics.accuracy_score(truth_array,predictions_closeness)*100,3))
print("accuracy TW-IDF weighted closeness:", round(metrics.accuracy_score(truth_array,predictions_w_closeness)*100,3))
print("accuracy TW-ICW degree:", round(metrics.accuracy_score(truth_array,predictions_twicw)*100,3))
print("accuracy TF-IDF:", round(metrics.accuracy_score(truth_array,predictions_tfidf)*100,3))
print( '========== top 10 for logreg classifier_w_closeness classifier ==========' )
print_top10( all_unique_terms , classifier_w_closeness , unique_labels )
print( '========== bot 10 for logreg classifier_w_closeness classifier ==========' )
print_bot10( all_unique_terms , classifier_w_closeness , unique_labels )
# show the most and less important features for each class
### fill the gaps ### hint: pick a classifier (e.g., 'classifier_tfidf'), and pass it to the 'print_top10' and 'print_bot10' functions along with 'unique_labels' and 'all_unique_terms'
3))
print(
"accuracy TW-IDF closeness:",
round(
metrics.accuracy_score(truth_array, predictions_closeness) * 100,
3))
print(
"accuracy TW-IDF weighted closeness:",
round(
metrics.accuracy_score(truth_array, predictions_w_closeness) * 100,
3))
print(
"accuracy TW-ICW degree:",
round(metrics.accuracy_score(truth_array, predictions_twicw) * 100, 3))
print(
"accuracy TF-IDF:",
round(metrics.accuracy_score(truth_array, predictions_tfidf) * 100, 3))
# show the most and less important features for each class
classifiers = [
classifier_degree, classifier_w_degree, classifier_closeness,
classifier_w_closeness, classifier_twicw, classifier_tfidf
]
i = 1
for classifier in classifiers:
print("classifier ", i)
i += 1
print("TOP 10:")
print_top10(all_unique_terms, classifier, unique_labels)
print("BOT 10:")
print_bot10(all_unique_terms, classifier, unique_labels)
# testing #
###########
# issue predictions
predictions_degree = classifier_degree.predict(testing_set_degree)
predictions_w_degree = classifier_w_degree.predict(testing_set_w_degree)
predictions_closeness = classifier_closeness.predict(testing_set_closeness)
predictions_w_closeness = classifier_w_closeness.predict(testing_set_w_closeness)
predictions_twicw = classifier_twicw.predict(testing_set_twicw)
predictions_tfidf = classifier_tfidf.predict(testing_set_tfidf)
print('========== accuracy for', clf ,'classifier ==========')
print("accuracy TW-IDF degree:", round(metrics.accuracy_score(truth_array,predictions_degree)*100,3))
print("accuracy TW-IDF weighted degree:", round(metrics.accuracy_score(truth_array,predictions_w_degree)*100,3))
print("accuracy TW-IDF closeness:", round(metrics.accuracy_score(truth_array,predictions_closeness)*100,3))
print("accuracy TW-IDF weighted closeness:", round(metrics.accuracy_score(truth_array,predictions_w_closeness)*100,3))
print("accuracy TW-ICW degree:", round(metrics.accuracy_score(truth_array,predictions_twicw)*100,3))
print("accuracy TF-IDF:", round(metrics.accuracy_score(truth_array,predictions_tfidf)*100,3))
# In[6]:
# show the most and less important features for each class
my_clf = classifier_tfidf
print("\nMost important features for each class:")
print_top10(all_unique_terms, my_clf, unique_labels)
print("\nLess important features for each class:")
print_bot10(all_unique_terms, my_clf, unique_labels)