lr = 1
# Construct model
clf = AdaBoostClassifier(random_state=RANDOM_STATE,
algorithm='SAMME.R',
learning_rate=lr,
n_estimators=NUM_ESTIMATORS)
# Perform training
clf.fit(train_data[predictors_str], train_data[target].values)
# Make predictions for later classification report
predictions_train = make_predictions(clf, predictors_str, train_data)
# Trace estimator weight and estimator error
is_show_model_details = False
if is_show_model_details:
print('Estimator Weight: ', clf.estimator_weights_)
print('\n')
print('Estimator type: ', clf.estimators_[0])
print('\n')
print('Estimator ErrorL ', clf.estimator_errors_)
######## Classification Performance on Validation Data with Undersampling######
# Load in validation data
for l in lr_arr:
clf = AdaBoostClassifier(random_state=RANDOM_STATE,
algorithm='SAMME.R',
learning_rate=l,
n_estimators=n)
# Perform training
print('Learing rate: ', l, ' Num of esimators: ', n)
print('\n')
clf.fit(train_data[predictors_str], train_data[target].values)
# Make predictors based on training and testing data
preds_train = make_predictions(clf, predictors_str, train_data)
preds_valid = make_predictions(clf, predictors_str, validation_data)
# Perform grid search accordinly
if is_build_model_for_under:
roc_auc_score_from_tra = roc_auc_score(train_data[target].values,
preds_train)
roc_auc_score_from_valid = roc_auc_score(
validation_data[target].values, preds_valid)
roc_auc_score_list_tra.append(roc_auc_score_from_tra)
roc_auc_score_list_val.append(roc_auc_score_from_valid)
if not is_build_model_for_under: