################################################################################
### Logistic Regression ########################################################
from sklearn.linear_model import LogisticRegression
train_X = df_lr['dep_A', 'dep_B', 'dep_C', 'dep_D', 'dep_E']
train_y = df_lr['indep']
lr = LogisticRegression(C=100000,
random_state=1234,
penalty='l2',
solver='newton-cg')
model = lr.fit(train_X, train_y)
# accuracy 계산
result = model.predict_proba(test_X)
result = pd.DataFrame(result)
criteria = 0.8 # for example
result = result.assign(
estimation=result[1].apply(lambda x: 'Y' if x >= criteria else 'N'))
# column 1 is probability for Y
test_y = test_y.reset_index()
result = pd.merge(result, test_y, left_index=True, right_index=True)
result = result.assign(accuracy=(result['estimation'] == result['fact']) * 1)
Accuracy = result['accuracy'].sum() / result['accuracy'].count()
################################################################################
### mean squared error #########################################################
from sklearn.metrics import mean_squared_error
