def apply_word2vec_extratrees(data_frame, classif_level, classif_type,
source_path):
data_frame['text'] = data_frame.apply(
lambda row: th.tokenize_complex_text(row['text']), axis=1)
data_frame['classification'] = data_frame.apply(
lambda row: th.tokenize_complex_text(row['classification']), axis=1)
df_single_classification = ch.get_list_each_text_a_different_classification(
data_frame)
x = df_single_classification['text']
y = df_single_classification['classification']
X_train, X_test, y_train, y_test = ch.get_train_test_from_data(x, y)
model_w2v = wmh.get_word2vec_model(X_train)
etree_w2v = Pipeline([("word2vec vectorizer",
wmh.MeanEmbeddingVectorizer(model_w2v)),
("extra trees", pmh.get_extra_tree())])
etree_w2v_tfidf = Pipeline([("word2vec vectorizer",
wmh.TfidfEmbeddingVectorizer(model_w2v)),
("extra trees", pmh.get_extra_tree())])
# NB!!!: the model does not support multi targets, so i duplicate the sources and give them different targets
y_pred = pmh.fit_predict_functions(etree_w2v_tfidf, X_train, y_train,
X_test)
classifier_name_0 = 'Word2Vec/MeanEmbeddingVectorizer'
classifier_name_1, parameters_1 = ch.get_extratree_classifier_information(
str(etree_w2v))
model_name = '[all classes predictions]' + classifier_name_0 + '/' + classifier_name_1
# this should be changed by comparing all the possibilities for specified text (i can use the original dataframe!)
list_metrics = mh.calculate_metrics(model_name, y_test, y_pred)
none_average, binary_average, micro_average, macro_average = list_metrics
ch.save_results(classifier_name_1, list_metrics, parameters_1, model_name,
classif_level, classif_type, source_path)
# NB!!!: the model does not support multi targets, so i duplicate the sources and give them different targets
y_pred = pmh.fit_predict_functions(etree_w2v, X_train, y_train, X_test)
classifier_name_2 = 'Word2Vec/TfidfEmbeddingVectorizer'
model_name = '[all classes predictions]' + classifier_name_2 + '/' + classifier_name_1
# this should be changed by comparing all the possibilities for specified text (i can use the original dataframe!)
list_metrics = mh.calculate_metrics(model_name, y_test, y_pred)
none_average, binary_average, micro_average, macro_average = list_metrics
ch.save_results(classifier_name_1, list_metrics, parameters_1, model_name,
classif_level, classif_type, source_path)
def apply_label_powerset(X_train, y_train, X_test, y_test, baseline_name,
source_path):
classifier = ch.get_label_powerset(pmh.get_logistic())
classifier_name, parameters = ch.get_complex_classifier_information(
str(classifier), 1, 1, 2, 0)
y_pred = pmh.fit_predict_functions(classifier, X_train, y_train, X_test)
model_name = '[all classes predictions]' + baseline_name + classifier_name
list_metrics = mh.calculate_metrics(model_name, y_test, y_pred)
none_average, binary_average, micro_average, macro_average = list_metrics
ch.save_results(classifier_name, list_metrics, parameters, model_name,
classif_level, classif_type, source_path)
def apply_svm(X_train_tfidf, y_train, X_test_tfidf, y_test, classif_level,
classif_type, source_path):
svm = pmh.get_SVC()
classifier_name, parameters = ch.get_classifier_information(str(svm))
y_pred = pmh.fit_predict_functions(svm, X_train_tfidf, y_train,
X_test_tfidf)
model_name = '[all classes predictions]label_encoder/tfidf/' + classifier_name
list_metrics = mh.calculate_metrics(model_name, y_test, y_pred)
none_average, binary_average, micro_average, macro_average = list_metrics
ch.save_results(classifier_name, list_metrics, parameters, model_name,
classif_level, classif_type, source_path)
def apply_onevsrest(X_train, y_train, X_test, y_test, classes, baseline_name,
source_path):
# X_train.sort_indices() # SVC needs this line in addition
custom_pipeline = Pipeline([
('clf', OneVsRestClassifier(pmh.get_logistic(), n_jobs=-1)),
# ('clf', OneVsRestClassifier(pmh.get_SVC(), n_jobs=-1)),
# ('clf', OneVsRestClassifier(pmh.get_multinomialNB(), n_jobs=-1)),
# ('clf', OneVsRestClassifier(pmh.get_decision_tree(), n_jobs=-1)),
# ('clf', OneVsRestClassifier(pmh.get_kneighbors(), n_jobs=-1)),
# ('clf', OneVsRestClassifier(pmh.get_linear_SVC(), n_jobs=-1)),
# ('clf', OneVsRestClassifier(pmh.get_random_forest_classifier(), n_jobs=-1)),
# ('clf', OneVsRestClassifier(pmh.get_SGD_classifier(), n_jobs=-1)),
])
classifier_name, parameters = ch.get_complex_classifier_information(
str(custom_pipeline), 3, 1, 4, 0)
model_name = '[each class predictions]' + baseline_name + classifier_name
accuracies = []
for _class in classes:
print('**Processing {} texts...**'.format(_class))
y_pred = pmh.fit_predict_functions(custom_pipeline, X_train,
y_train[_class], X_test)
accuracies.append(mh.get_accuracy_score(y_test[_class], y_pred))
list_metrics = mh.calculate_metrics(model_name, y_test[_class], y_pred)
none_average, binary_average, micro_average, macro_average = list_metrics
ch.save_results(classifier_name, list_metrics, parameters, model_name,
classif_level, classif_type, source_path)
print('Pipeline score on training {}'.format(
custom_pipeline.score(X_train, y_train[_class])))
true_positives, false_positives, tpfn = mh.get_predictions_distribution(
y_test[_class], y_pred)
model_name = '[all classes predictions]' + baseline_name + classifier_name
if tpfn == 0 or (true_positives + false_positives) == 0:
mh.display_directly_metrics(model_name, 0, 0, 0, -1)
else:
precision = true_positives / (true_positives + false_positives)
recall = true_positives / tpfn
mh.display_directly_metrics(
model_name, precision, recall,
2 * (precision * recall) / (precision + recall), -1)
def apply_adapted_algorithm(X_train, y_train, X_test, y_test, baseline_name,
source_path):
classifier = pmh.get_MLkNN()
classifier_name, parameters = ch.get_complex_classifier_information(
str(classifier), 0, 0, 1, 0)
X_train, y_train, X_test = th.get_lil_matrices(X_train, y_train, X_test)
y_pred = pmh.fit_predict_functions(classifier, X_train, y_train, X_test)
model_name = '[all classes predictions]' + baseline_name + classifier_name
list_metrics = mh.calculate_metrics(model_name, y_test, y_pred)
none_average, binary_average, micro_average, macro_average = list_metrics
ch.save_results(classifier_name, list_metrics, parameters, model_name,
classif_level, classif_type, source_path)
def apply_doc2vec_logistic_regression(data_frame, text_vectorizer,
class_vectorizer, classif_level,
classif_type, source_path):
baseline_name = '[all classes predictions]' + text_vectorizer + '/' + class_vectorizer
results_model_dbow = ch.apply_doc2vec_separated_train_test(
data_frame, baseline_name)
X_train, y_train, X_test, y_test, model_dbow, train_tagged, test_tagged = results_model_dbow
logreg = pmh.get_logistic()
classifier_name_0, parameters_0 = ch.get_classifier_information(
str(model_dbow))
classifier_name_1, parameters_1 = ch.get_classifier_information(
str(logreg))
y_pred = pmh.fit_predict_functions(logreg, X_train, y_train, X_test)
model_name = baseline_name + classifier_name_0 + '/' + classifier_name_1
list_metrics = mh.calculate_metrics(model_name, y_test, y_pred)
none_average, binary_average, micro_average, macro_average = list_metrics
ch.save_results(classifier_name_1, list_metrics, parameters_1, model_name,
classif_level, classif_type, source_path)
# baseline_name = '[each class predictions]'+text_vectorizer+'/'+class_vectorizer
# vectorizer_results = ch.apply_df_vectorizer(data_frame, 'doc2vec', 'multi_label', baseline_name)
# X_train, X_test, y_train, y_test, classes, n_classes, vocab_processor, len_vocabulary = vectorizer_results
# model_name = baseline_name+'/'+classifier_name_0+'/'+classifier_name_1
#
# for i in range(n_classes):
# unique, counts = np.unique(y_train[:, i], return_counts=True)
# if len(counts) > 1 and counts[1] > 1:
# y_pred = pmh.fit_predict_functions(custom_pipeline, X_train, y_train[:, i], X_test)
# print('### ', classes[i] ,' ###')
# list_metrics = mh.calculate_metrics(model_name, y_test[:, i], y_pred)
# none_average, binary_average, micro_average, macro_average = list_metrics
#
# ch.save_results(classifier_name_1, list_metrics, parameters_1, model_name, classif_level, classif_type, source_path)
# improvement
improved_logistic_regression(train_tagged, test_tagged, model_dbow, logreg,
text_vectorizer, class_vectorizer,
classif_level, classif_type, source_path)
def improved_logistic_regression(train_tagged, test_tagged, model_dbow, logreg,
text_vectorizer, class_vectorizer,
classif_level, classif_type, source_path):
classifier_name_0, parameters_0 = ch.get_classifier_information(
str(model_dbow))
classifier_name_1, parameters_1 = ch.get_classifier_information(
str(logreg))
model_dmm = wmh.train_doc2vec_with_tagged_data(train_tagged.values)
classifier_name_2, parameters_2 = ch.get_classifier_information(
str(model_dmm))
y_train, X_train = wmh.vec_for_learning(model_dmm, train_tagged)
y_test, X_test = wmh.vec_for_learning(model_dmm, test_tagged)
y_pred = pmh.fit_predict_functions(logreg, X_train, y_train, X_test)
model_dbow.delete_temporary_training_data(keep_doctags_vectors=True,
keep_inference=True)
model_dmm.delete_temporary_training_data(keep_doctags_vectors=True,
keep_inference=True)
merged_model = wmh.get_concatenated_doc2vec(model_dbow, model_dmm)
y_train, X_train = wmh.vec_for_learning(merged_model, train_tagged)
y_test, X_test = wmh.vec_for_learning(merged_model, test_tagged)
y_pred = pmh.fit_predict_functions(logreg, X_train, y_train, X_test)
model_name = '[all classes predictions]' + classifier_name_0 + '/' + classifier_name_2 + '/' + classifier_name_1
list_metrics = mh.calculate_metrics(model_name, y_test, y_pred)
none_average, binary_average, micro_average, macro_average = list_metrics
ch.save_results(classifier_name_1, list_metrics, parameters_1, model_name,
classif_level, classif_type, source_path)
def apply_multi_label_classification_without_pipeline(
data_frame, text_vectorizer, class_vectorizer, classif_level,
classif_type, source_path):
baseline_name = text_vectorizer + '/' + class_vectorizer + '/onevsrest/'
vectorizer_results = ch.apply_df_vectorizer(data_frame, text_vectorizer,
class_vectorizer,
'[both]' + baseline_name)
X_train, X_test, y_train, y_test, classes, n_classes, vocab_processor, len_vocabulary = vectorizer_results
# len_vocabulary = 57335
# len_vocabulary = 34736
print(X_train[0])
print('len_vocabulary: ', len_vocabulary, ' num_classes: ', n_classes)
# Run classifier
classifier = OneVsRestClassifier(pmh.get_logistic()) # here
# classifier = OneVsRestClassifier(pmh.get_SVC()) # here
# classifier = OneVsRestClassifier(pmh.get_multinomialNB()) # ERROR WORD2VEC/DOC2VEC (X has negative value)
# classifier = OneVsRestClassifier(pmh.get_decision_tree()) # here
# classifier = OneVsRestClassifier(pmh.get_kneighbors()) # here
# classifier = OneVsRestClassifier(pmh.get_linear_SVC())
# classifier = OneVsRestClassifier(pmh.get_random_forest_classifier()) # here
# classifier = OneVsRestClassifier(pmh.get_SGD_classifier())
train_predictions = np.ndarray(shape=(n_classes, y_train.shape[0]),
dtype=int)
predictions = np.ndarray(shape=(n_classes, y_test.shape[0]), dtype=int)
###
precision = dict()
recall = dict()
average_precision = dict()
classifier_name, parameters = ch.get_complex_classifier_information(
str(classifier), 1, 1, 2, 0)
second_training = False
just_once = True
another_try = False # single train and estimation instead of multi-train-estimation steps (it performs better with svm and logistic)
if not another_try:
for _ in range(1):
for i in range(n_classes):
if second_training:
if classifier_name in [
'DecisionTreeClassifier', 'KNeighborsClassifier',
'MultinomialNB', 'RandomForestClassifier'
]:
# do not provide the second metrics
break
elif just_once:
# fit again with the whole set of classes - should be better
classifier.fit(X_train, y_train)
y_score = classifier.decision_function(X_test)
precision[i], recall[i], average_precision[
i] = mh.calculate_recall_curve_precision_score(
y_test[:, i], y_score[:, i], None,
y_test[:, i], y_score[:, i])
just_once = False
else:
predictions[i] = pmh.fit_predict_functions(
classifier, X_train, y_train[:, i], X_test)
train_predictions[i] = classifier.predict(X_train)
print('**Processing classes {0:0.2f} % ...**'.format(
((i + 1) / n_classes) * 100))
second_training = True
break
predictions = predictions.transpose()
print('transposed')
else:
predictions = pmh.fit_predict_functions(classifier, X_train, y_train,
X_test)
train_predictions = classifier.predict(X_train)
# metrics
model_name = '[each class predictions]' + baseline_name + classifier_name
manual_metrics = mh.calculate_manual_metrics(model_name, y_test,
predictions)
none_average, binary_average, micro_average, macro_average = manual_metrics
# metrics
list_metrics = mh.calculate_metrics(model_name, y_test, predictions)
none_average, binary_average, micro_average, macro_average = list_metrics
ch.save_results(classifier_name, list_metrics, parameters, model_name,
classif_level, classif_type, source_path)
if not just_once:
print('not just once')
train_predictions = classifier.predict(X_train)
predictions = classifier.predict(X_test)
# metrics
mh.calculate_metrics_with_recall_curve(y_score, y_train, y_test,
train_predictions, predictions)
# metrics
model_name = '[all classes predictions]' + baseline_name + classifier_name
list_metrics = mh.calculate_metrics(model_name, y_test, predictions)
none_average, binary_average, micro_average, macro_average = list_metrics
ch.save_results(classifier_name, list_metrics, parameters, model_name,
classif_level, classif_type, source_path)