def classify_docs(self, classifier):
english_text_preprocessor = EnglishTextPreProcessor()
training_data = pnd.read_csv(Config.ENGLISH_TRAINING_DATA_DIR)
testing_data = pnd.read_csv(Config.ENGLISH_TEST_DATA_DIR)
tfidf = TfIdf("English", training_data, english_text_preprocessor)
if classifier == 'KNN':
clf = KNNClassifier(training_data, testing_data, tfidf)
elif classifier == 'Naive Bayes':
clf = NaiveBayesClassifier(training_data, testing_data, tfidf)
elif classifier == 'SVM':
clf = SVMClassifier(training_data, testing_data, tfidf)
else:
clf = RandomForestClassifier(training_data, testing_data, tfidf)
predicted_classes = np.zeros(shape=(self.n_documents, ), dtype=int)
for i, doc_words in enumerate(
tqdm(self.document_words, position=0, leave=True)):
str = ""
for word in doc_words:
str += word + " "
if classifier == 'KNN':
predicted_classes[i] = clf.predict(str[:-1], 5)
else:
predicted_classes[i] = clf.predict(str[:-1])
return predicted_classes
np.set_printoptions(precision=3)
logger.info("On " + set_name + " Data")
logger.info("K = {}:\n"
"\tAccuracy = {:.3f}\n"
"\tRecall_per class = {}\n"
"\tPrecision_per class = {}\n"
"\tmicro_F1 = {:.3f}\n"
"\tmacro_F1 = {:.3f}".format(k, accuracy, per_class_recall, per_class_precision,
micro_f1, macro_f1))
return micro_f1
def evaluate_validation_for_ks(self, ks):
micro_f1_list = []
for k in ks:
micro_f1_list.append(self.evaluate(self.validation_docs_matrix, self.validation_labels, k,
set_name="Validation"))
return ks[np.argmax(micro_f1_list)]
if __name__ == '__main__':
english_text_preprocessor = EnglishTextPreProcessor()
training_data = pnd.read_csv(Config.ENGLISH_TRAINING_DATA_DIR)
testing_data = pnd.read_csv(Config.ENGLISH_TEST_DATA_DIR)
tfidf = TfIdf("English", training_data, english_text_preprocessor)
knn_clf = KNNClassifier(training_data, testing_data, tfidf)
# a = knn_clf.predict("Italy", 5)
# print(a)