'raw_data/models/reduced_search/best_model_{0}.pkl'.format(lang))
if not my_file.exists():
continue
else:
with open(
'raw_data/models/reduced_search/best_model_{0}.pkl'.format(
lang), 'rb') as file:
best_model = joblib.load(file)
# Load best model parameters
with open(
'raw_data/experiments/reduced_search/best_model_parameters_{0}.pkl'
.format(lang), 'rb') as file:
best_params = joblib.load(file)
dataset = mlconjug3.DataSet(mlconjug3.Verbiste(language=lang).verbs)
dataset.split_data(proportion=0.95)
predicted = best_model.predict(dataset.test_input)
predicted2 = best_model.predict(dataset.verbs_list)
score = len([
(a, b) for a, b in zip(predicted, dataset.test_labels) if a == b
]) / len(predicted)
misses = len([(a, b) for a, b in zip(predicted, dataset.test_labels)
if a != b])
entries = len(predicted)
print('The score of the {0} best model is {1} on test set.'.format(
lang, score))
score2 = len([(a, b)
langs = ('ro', 'it', 'en', 'es', 'pt', 'fr')
if __name__ == "__main__":
# multiprocessing requires the fork to happen in a __main__ protected
# block
for lang in langs:
my_file = Path(
'raw_data/models/reduced_search/best_model_{0}.pkl'.format(lang))
if my_file.is_file():
continue
else:
# #############################################################################
# Initialize Data Set
dataset = mlconjug3.DataSet(
mlconjug3.Verbiste(language=lang).verbs)
# #############################################################################
# Define a pipeline.
# Transforms dataset with CountVectorizer. We pass the function extract_verb_features to the CountVectorizer.
ngrange = (2, 7)
vectorizer = mlconjug3.CountVectorizer(analyzer=partial(
mlconjug3.extract_verb_features,
lang=lang,
ngram_range=ngrange),
binary=True)
# Feature reduction
feature_reductor = mlconjug3.SelectFromModel(
mlconjug3.LinearSVC(random_state=42,