def classify_with_tf_idf(paths):
sentences, labels, class_names = load_test_data(paths)
sentences = np.array(sentences)
labels = np.array(labels)
average_precisions = []
average_recalls = []
for train_index, test_index in sklearn.cross_validation.StratifiedKFold(labels, n_folds=3):
sentences_train, sentences_test = sentences[train_index], sentences[test_index]
labels_train, labels_test = labels[train_index], labels[test_index]
features_train, vocabulary, count_vectorizer = extract_features_and_vocabulary(sentences_train)
tfidf_features_train = transform_to_tfidf(features_train)
predicted = predict_with_svc(tfidf_features_train, labels_train, sentences_test, count_vectorizer)
print('TF-IDF')
average_precision, average_recall = evaluate_classification(predicted, labels_test, sentences, class_names)
average_precisions.append(average_precision)
average_recalls.append(average_recall)
evaluate_complete_classification(average_precisions, average_recalls)
def predict_with_svc(tfidf_features_train, labels_train, sentences_test, count_vectorizer):
clf = LinearSVC().fit(tfidf_features_train, labels_train)
features_test = extract_features_for_testing(sentences_test, count_vectorizer)
tfidf_features_test = transform_to_tfidf(features_test)
return clf.predict(tfidf_features_test)
