def train_SVC(filePath):
'''TRAINING'''
train_df = pandas.read_csv(filePath, sep='\t')
# train_df = model_utils.oversample_neutral_class(train_df)
train_class = train_df[' class'].as_matrix()
train_data = model_utils.apply_aspdep_weight(train_df, 0.7)
print train_data[0]
text_clf = SVC(C=0.2,
cache_size=200,
class_weight=None,
coef0=0.0,
decision_function_shape='ovr',
degree=3,
gamma=0.5,
kernel='poly',
max_iter=-1,
probability=False,
random_state=None,
shrinking=True,
tol=0.001,
verbose=False).fit(train_data, train_class)
print(set(text_clf.predict(train_data)))
print(set(train_class))
joblib.dump(text_clf, 'model_dumps/SVC_model.pkl')
"""PERFORMANCE EVALUATION"""
accuracy, clf_report = model_utils.get_cv_metrics(text_clf,
train_data,
train_class,
k_split=10)
print("Accuracy: ", accuracy)
print(clf_report)
def train_StackedGeneralizer(filePath):
"""TRAINING"""
train_df = pandas.read_csv(filePath, sep='\t')
# train_df = model_utils.oversample_neutral_class(train_df)
train_class = train_df[' class'].as_matrix()
train_data_1 = model_utils.apply_aspdep_weight(train_df, 1.7)
# base_models = [MultinomialNB(alpha=0.6, fit_prior=True, class_prior=None), BernoulliNB(alpha=1.2, fit_prior=True, class_prior=None),
# linear_model.SGDClassifier(loss='squared_loss', penalty='l2', alpha=1e-3, random_state=607,
# max_iter=1000000, tol=1e-2)]
base_models = [
joblib.load('Multinomial_nb_model.pkl'),
joblib.load('Bernoulli_nb_model.pkl'),
joblib.load('SGD_model.pkl'),
joblib.load('RF_model.pkl')
]
# define blending model
blending_model = LogisticRegression(random_state=607)
# initialize multi-stage model
sg = StackedGeneralizer(base_models,
blending_model,
n_folds=10,
verbose=False)
sg.fit(train_data_1, train_class)
# joblib.dump(sg, 'Stacked_model.pkl')
"""PERFORMANCE EVALUATION"""
accuracy, clf_report = model_utils.get_cv_metrics(sg,
train_data_1,
train_class,
k_split=10)
print("Accuracy: ", accuracy) #Accuracy: 0.7497418660799471 Weights: 1.7
def train_SGD(filePath):
'''TRAINING'''
train_df = pandas.read_csv(filePath, sep='\t')
# train_df = model_utils.oversample_neutral_class(train_df)
train_class = train_df[' class'].as_matrix()
train_data_1 = model_utils.apply_aspdep_weight(train_df, 0.5)
text_clf = linear_model.SGDClassifier(
loss='squared_loss',
penalty='l2',
alpha=1e-3,
random_state=607,
max_iter=1000000,
tol=1e-2).fit(train_data_1,
train_class) #Accuracy: 0.7797260574839455 @ 2.1 weight
# Accuracy: 0.7428003692958715 Weights: 0.5
joblib.dump(text_clf, 'SGD_model.pkl')
"""PERFORMANCE EVALUATION"""
accuracy, clf_report = model_utils.get_cv_metrics(text_clf,
train_data_1,
train_class,
k_split=10)
print("Accuracy: ", accuracy)
print(clf_report)
def train_polarity_clf(filePath):
train_df = pandas.read_csv(filePath, sep='\t')
train_df = model_utils.oversample_neutral_class(train_df)
train_class = train_df[' class'].as_matrix()
train_data = train_df['opin_polarity'].as_matrix()
print train_data
# text_clf = BernoulliNB(alpha=1.0, fit_prior=True, class_prior=None).fit(train_data, train_class)
text_clf = LogisticRegression(random_state=0).fit(train_data)
"""PERFORMANCE EVALUATION"""
accuracy, clf_report = model_utils.get_cv_metrics(text_clf,
train_data,
train_class,
k_split=10)
print("Accuracy: ", accuracy)
print(clf_report)
def train_ET(filePath):
'''TRAINING'''
train_df = pandas.read_csv(filePath, sep='\t')
# train_df = model_utils.oversample_neutral_class(train_df)
train_class = train_df[' class'].as_matrix()
train_data_1 = model_utils.apply_aspdep_weight(train_df, 0.3)
text_clf = ExtraTreesClassifier(n_estimators=10,
max_depth=2,
random_state=0,
n_jobs=-1).fit(train_data_1, train_class)
"""PERFORMANCE EVALUATION"""
accuracy, clf_report = model_utils.get_cv_metrics(text_clf,
train_data_1,
train_class,
k_split=10)
print("Accuracy: ", accuracy)
def train_BernoulliNB(filePath):
'''TRAINING'''
train_df = pandas.read_csv(filePath, sep='\t')
# train_df = model_utils.oversample_neutral_class(train_df)
train_class = train_df[' class'].as_matrix()
train_data_1 = model_utils.apply_aspdep_weight(train_df, 0.0)
text_clf = BernoulliNB(alpha=0.6, fit_prior=True, class_prior=None).fit(
train_data_1, train_class
) # 1.2 Accuracy: 0.7047700662655685 Weights: 0.0 Alpha 0.6
joblib.dump(text_clf, 'Bernoulli_nb_model.pkl')
"""PERFORMANCE EVALUATION"""
accuracy, clf_report = model_utils.get_cv_metrics(text_clf,
train_data_1,
train_class,
k_split=10)
print("Accuracy: ", accuracy)
print(clf_report)
def train_MultinomialNB(filePath):
'''TRAINING'''
train_df = pandas.read_csv(filePath, sep='\t')
# train_df = model_utils.oversample_neutral_class(train_df)
train_class = train_df[' class'].as_matrix()
train_data_1 = model_utils.apply_aspdep_weight(train_df, 0.9)
text_clf = MultinomialNB(alpha=0.3, fit_prior=True, class_prior=None).fit(
train_data_1,
train_class) # 1.2, 0.01 Accuracy: 0.7407742366772097 Weights: 1.2
joblib.dump(text_clf, 'Multinomial_nb_model.pkl')
"""PERFORMANCE EVALUATION"""
accuracy, clf_report = model_utils.get_cv_metrics(text_clf,
train_data_1,
train_class,
k_split=10)
print("Accuracy: ", accuracy)
print(clf_report)
def train_RF(filePath):
'''TRAINING'''
train_df = pandas.read_csv(filePath, sep='\t')
# train_df = model_utils.oversample_neutral_class(train_df)
train_class = train_df[' class'].as_matrix()
train_data_1 = model_utils.apply_aspdep_weight(train_df, 1.1)
text_clf = RandomForestClassifier(n_estimators=300,
max_depth=180,
random_state=607,
n_jobs=-1).fit(train_data_1, train_class)
joblib.dump(text_clf, 'RF_model.pkl')
"""PERFORMANCE EVALUATION"""
accuracy, clf_report = model_utils.get_cv_metrics(text_clf,
train_data_1,
train_class,
k_split=10)
print("Accuracy: ", accuracy)
print(clf_report)