def main():
parser = argparse.ArgumentParser(
description='Ensembling HyperSpace models')
parser.add_argument('--results_dir',
type=str,
help='Path to results directory.')
args = parser.parse_args()
results = load_results(args.results_dir)
clfs = []
for _ in results:
# Use default random forest for each of the ensemble models
model = RandomForestClassifier()
clfs.append(model)
# Using the training and test sets from `optimize.py`
X_train, _, X_test, y_train, _, y_test = load_data(0.25, 0.25)
y_preds, y_proba = ensemble(clfs, X_train, y_train, X_test, y_test)
acc_test = accuracy_score(y_test, y_preds)
ll_test = log_loss(y_test, y_proba)
print(
f'\nEnsemble Test Accuracy: {acc_test:.4f}, Test Log Loss: {ll_test:.4f}'
)
def main():
parser = argparse.ArgumentParser(
description='Ensembling HyperSpace models')
parser.add_argument('--results_dir',
type=str,
help='Path to results directory.')
args = parser.parse_args()
results = load_results(args.results_dir)
clfs = []
for result in results:
# Get domain from best model in each hyperspace
hyperparams = result.x
n_estimators = hyperparams[0]
max_depth = hyperparams[1]
model = RandomForestClassifier(n_estimators=n_estimators,
max_depth=max_depth)
clfs.append(model)
# Using the training and test sets from `optimize.py`
X_train, _, X_test, y_train, _, y_test = load_data(0.25, 0.25)
y_preds, y_proba = ensemble(clfs, X_train, y_train, X_test, y_test)
acc_test = accuracy_score(y_test, y_preds)
ll_test = log_loss(y_test, y_proba)
print(
f'\nEnsemble Test Accuracy: {acc_test:.4f}, Test Log Loss: {ll_test:.4f}'
)