def train_models(self, kernel_type):
if kernel_type == 'linear_svc':
print('Training Linear SVC...')
pipeline = get_pipeline(svm.SVC(kernel='linear', decision_function_shape='ovr', C=1, max_iter = -1, random_state = 42), 'tfidf')
self.linear_model = get_gdsearch(pipeline, 'SVC')
self.linear_model.fit(self.X, self.y)
print('Best estimator params: ', self.linear_model.best_params_,'\n')
return self.linear_model
if kernel_type == 'rbf_svc':
print('Training RBF SVC...')
pipeline = get_pipeline(svm.SVC(kernel='rbf', decision_function_shape='ovr', C=1, gamma=1, max_iter = -1, random_state = 42), 'tfidf')
self.rbf_model = get_gdsearch(pipeline, 'SVC')
self.rbf_model.fit(self.X, self.y)
print('Best estimator params: ', self.rbf_model.best_params_,'\n')
return self.rbf_model
if kernel_type == 'sigmoid_svc':
print('Training Sigmoid SVC...')
pipeline = get_pipeline(svm.SVC(kernel='sigmoid', decision_function_shape='ovr', C=1, gamma=1, max_iter = -1, random_state = 42), 'tfidf')
self.sigmoid_model = get_gdsearch(pipeline, 'SVC')
self.sigmoid_model.fit(self.X, self.y)
print('Best estimator params: ', self.sigmoid_model.best_params_,'\n')
return self.sigmoid_model
if kernel_type == 'poly_svc':
print('Training Polynomial SVC...')
pipeline = get_pipeline(svm.SVC(kernel='poly', decision_function_shape='ovr', C=1, gamma=1, max_iter = -1, random_state = 42), 'tfidf')
self.poly_model = get_gdsearch(pipeline, 'SVC')
self.poly_model.fit(self.X, self.y)
print('Best estimator params: ', self.poly_model.best_params_,'\n')
return self.poly_model
def train_model(self, skip_pipeline=False):
if not (skip_pipeline):
pipeline = get_pipeline(
XGBClassifier(random_state=42,
seed=2,
objective='multi:softmax',
eval_metric='merror',
use_label_encoder=False,
learning_rate=0.1,
n_jobs=-1,
colsample_bytree=1.0), 'tfidf')
self.model = get_gdsearch(pipeline, 'XGBoost')
self.model = self.model.fit(self.X, self.y)
print('Best model params:', self.model.best_params_)
print('All model params: ', self.model.get_params(True), '\n')
else:
self.model = XGBClassifier(random_state=42,
seed=2,
objective='multi:softmax',
eval_metric='merror',
use_label_encoder=False,
n_estimators=30,
colsample_bytree=0.8,
subsample=1.0)
self.model.fit(self.X, self.y)
def train_model(self, skip_vectorizer):
self.get_model()
pipeline = get_pipeline(self.model, 'count', True)
if skip_vectorizer:
pipeline = get_pipeline(self.model)
self.model = get_gdsearch(pipeline, 'Logistic', skip_vectorizer).fit(self.X, self.y)
print('Best estimator params:',self.model.best_params_)
def train_model(self):
pipeline = get_pipeline(
DecisionTreeClassifier(random_state=42,
criterion='gini',
splitter='best',
class_weight='balanced'))
self.model = get_gdsearch(pipeline,
model_type='DTree').fit(self.X, self.y)
def train_models(self):
pipeline = get_pipeline(
LinearSVC(C=1,
multi_class='ovr',
class_weight='balanced',
max_iter=1000000,
dual=True,
tol=1e-5), 'tfidf')
self.model = get_gdsearch(pipeline, 'LSVM').fit(self.X, self.y)
print('Best estimator params: ', self.model.best_params_)
print('All model params: ', self.model.get_params(True), '\n')
def train_model(self, model_type='Gaussian'):
if type == 'Gaussian':
vectorizer = Vectorizer(self.X, self.y)
train_features, test_features = vectorizer.get_vectorized_features(
type='count')
self.test_features = test_features
self.model = GaussianNB().fit(train_features.toarray(), self.y)
else:
self.model = get_gdsearch(get_pipeline(MultinomialNB(), True),
model_type).fit(self.X, self.y)
print('Best parameters selected', self.model.best_params_, '\n')
def train_model(self):
pipeline = get_pipeline(
RandomForestClassifier(random_state=42, n_jobs=-1))
self.model = get_gdsearch(pipeline, 'RandomForest').fit(self.X, self.y)