def compare_classes_amount(reviews_data, rates_data):
less_classes_reviews, less_classes_rates = split_by_rates(
reviews_data, rates_data)
print('two classes:')
print(
'f1 score:',
train_and_test(less_classes_reviews,
less_classes_rates,
min_df=default_min_df,
max_df=default_max_df,
ngram_range=default_n_gram,
alpha=default_alpha,
clf=classifier_name,
max_features=default_max_features))
print('four classes:')
print(
'f1 score:',
train_and_test(reviews_data,
rates_data,
min_df=default_min_df,
max_df=default_max_df,
ngram_range=default_n_gram,
alpha=default_alpha,
clf=classifier_name,
max_features=default_max_features))
def learn_model():
test_data_set = get_representation_data(AVAILABLE_DATA_DIRECTORIES[3])
prediction_data_set = get_representation_data(
AVAILABLE_DATA_DIRECTORIES[3])
train_reviews = get_reviews_from_data_set(test_data_set)
train_ratings = get_ratings_from_data_set(test_data_set)
prediction_reviews = get_reviews_from_data_set(prediction_data_set)
prediction_ratings = get_ratings_from_data_set(prediction_data_set)
# optimize_parameters(train_reviews, train_ratings)
train_and_test(train_reviews, train_ratings, prediction_reviews,
prediction_ratings)
def test_classifiers(reviews_data, rates_data):
# MultinomialNB
print(classifier_name)
print(
f'\nmin_df: {default_min_df} | max_df: {default_max_df} | ngram: {default_n_gram} | alpha: {default_alpha} '
f'| max_features: {default_max_features}')
print(
'f1 score:',
train_and_test(reviews_data,
rates_data,
min_df=default_min_df,
max_df=default_max_df,
ngram_range=default_n_gram,
alpha=default_alpha,
clf=classifier_name,
max_features=default_max_features))
# SVC_linear
classifier = 'SVC_linear'
print('\n', classifier)
print(
f'min_df: {default_min_df} | max_df: {default_max_df} | ngram: {default_n_gram} | alpha: {default_alpha} '
f'| max_features: {default_max_features}')
print(
'f1 score:',
train_and_test(reviews_data,
rates_data,
min_df=default_min_df,
max_df=default_max_df,
ngram_range=default_n_gram,
alpha=default_alpha,
clf=classifier,
max_features=default_max_features))
# SVC_rbf
classifier = 'SVC_rbf'
print('\n', classifier)
print(
f'min_df: {default_min_df} | max_df: {default_max_df} | ngram: {default_n_gram} | alpha: {default_alpha} '
f'| max_features: {default_max_features}')
print(
'f1 score:',
train_and_test(reviews_data,
rates_data,
min_df=default_min_df,
max_df=default_max_df,
ngram_range=default_n_gram,
alpha=default_alpha,
clf=classifier,
max_features=default_max_features))
def test_tuning(reviews_data, rates_data):
print('Before:')
print(
'f1 score:',
train_and_test(reviews_data,
rates_data,
min_df=default_min_df,
max_df=default_max_df,
ngram_range=default_n_gram,
alpha=default_alpha,
clf=classifier_name,
max_features=default_max_features))
print('After:')
print(
'f1 score:',
train_and_test(reviews_data,
rates_data,
min_df=2,
max_df=0.6,
ngram_range=(1, 2),
alpha=0.01,
clf=classifier_name,
max_features=default_max_features))
def run_parameters_test(reviews, rates, classifier, min_df, max_df, n_gram,
alpha, max_features, test_type):
print('\n', classifier, f'{test_type} test')
print(
f'min_df: {min_df} | max_df: {max_df} | ngram: {n_gram} | alpha: {alpha} '
f'| max_features: {max_features} ')
print(
'f1 score:',
train_and_test(reviews,
rates,
min_df=min_df,
max_df=max_df,
ngram_range=n_gram,
alpha=alpha,
clf=classifier,
max_features=max_features))
def test_size_and_random_state(reviews_data, rates_data):
default_test_size = 0.2
default_train_size = 0.25
default_random_state = 40
print('test_size')
print(
'f1 score:',
train_and_test(reviews_data,
rates_data,
test_size=0.1,
train_size=default_train_size,
random_state=default_random_state))
print('test_size')
print(
'f1 score:',
train_and_test(reviews_data,
rates_data,
test_size=0.2,
train_size=default_train_size,
random_state=default_random_state))
print('test_size')
print(
'f1 score:',
train_and_test(reviews_data,
rates_data,
test_size=0.3,
train_size=default_train_size,
random_state=default_random_state))
print('train_size')
print(
'f1 score:',
train_and_test(reviews_data,
rates_data,
test_size=default_test_size,
train_size=0.1,
random_state=default_random_state))
print('train_size')
print(
'f1 score:',
train_and_test(reviews_data,
rates_data,
test_size=default_test_size,
train_size=0.2,
random_state=default_random_state))
print('train_size')
print(
'f1 score:',
train_and_test(reviews_data,
rates_data,
test_size=default_test_size,
train_size=0.3,
random_state=default_random_state))
print('random_state')
print(
'f1 score:',
train_and_test(reviews_data,
rates_data,
test_size=default_test_size,
train_size=default_train_size,
random_state=5))
print('random_state')
print(
'f1 score:',
train_and_test(reviews_data,
rates_data,
test_size=default_test_size,
train_size=default_train_size,
random_state=15))
print('random_state')
print(
'f1 score:',
train_and_test(reviews_data,
rates_data,
test_size=default_test_size,
train_size=default_train_size,
random_state=25))