x_train, y_train_c, y_train_r, x_dev, y_dev_c, y_dev_r, x_test = load_data_v4(
data_path='data')
print('Training Multi-Task MLP model ...')
check_pointer = ModelCheckpoint(filepath='models/multi_task_mlp.hdf5',
verbose=1,
save_best_only=True,
save_weights_only=True)
early_stopping = EarlyStopping(patience=3)
csv_logger = CSVLogger('logs/multi_task_mlp.log')
mt_model = multi_task_mlp(sample_dim=x_train.shape[1])
mt_model.fit({'mlp_input': x_train}, {
'r_output': y_train_r,
'c_output': y_train_c
},
validation_data=([x_dev], [y_dev_r, y_dev_c]),
batch_size=128,
epochs=50,
verbose=1,
callbacks=[check_pointer, early_stopping, csv_logger])
if submit_flag:
print('Generate submission ...')
mt_model.load_weights(filepath='models/multi_task_mlp.hdf5')
results = mt_model.predict([x_test])[0].reshape(-1, 1)
make_submission(result_path='submissions',
results=results,
model_name='Multi_Task_MLP')
print('***** End UAI-CUP-2017 *****')
csv_logger = CSVLogger('logs/embedding_mlp.log')
embedding_mlp_model = embedding_mlp(
cont_feature_dim=x_train_cont_features.shape[1])
embedding_mlp_model.fit(
[
x_train_day_of_week, x_train_create_hour, x_train_poi_points,
x_train_cont_features
],
y_train,
batch_size=128,
epochs=100,
verbose=1,
validation_data=([
x_dev_day_of_week, x_dev_create_hour, x_dev_poi_points,
x_dev_cont_features
], y_dev),
callbacks=[check_pointer, early_stopping, csv_logger])
if submit_flag:
print('Generate submission ...')
embedding_mlp_model.load_weights(filepath='models/embedding_mlp.hdf5')
results = embedding_mlp_model.predict([
x_test_day_of_week, x_test_create_hour, x_test_poi_points,
x_dev_cont_features
]).reshape(-1, 1)
make_submission(result_path='submissions',
results=results,
model_name='Embedding_MLP')
print('***** End UAI-CUP-2017 *****')
return model
if __name__ == '__main__':
submit_flag = True if sys.argv[1] == 'submit' else False
print('***** Start UAI-CUP-2017 *****')
print('Loading data ...')
x_train, y_train, x_dev, y_dev, x_test = load_data_v2(data_path='data')
print('Training XGBoost Regression model ...')
xgb_model = xgb()
eval_set = [(x_dev, y_dev)]
xgb_model.fit(x_train, y_train,
early_stopping_rounds=3,
eval_metric='mae',
eval_set=eval_set,
verbose=True)
plot_importance(xgb_model)
pyplot.show()
if submit_flag:
print('Generate submission ...')
results = xgb_model.predict(x_test).reshape(-1, 1)
make_submission(result_path='submissions', results=results, model_name='xgboost')
print('***** End UAI-CUP-2017 *****')
# r_check_pointer = ModelCheckpoint(filepath='models/r_phase_mlp.hdf5', verbose=1, save_best_only=True,
# save_weights_only=True)
# r_csv_logger = CSVLogger('logs/r_phase_mlp.log')
r_model = regression_mlp(sample_dim=x_train_r.shape[1])
# r_model.fit({'mlp_input': x_train_r},
# {'r_output': y_train_r},
# validation_split=0.1,
# batch_size=128, epochs=50, verbose=1,
# callbacks=[r_check_pointer, early_stopping, r_csv_logger])
if submit_flag:
print('Generate submission ...')
c_model.load_weights(filepath='models/c_phase_mlp.hdf5')
r_model.load_weights(filepath='models/r_phase_mlp.hdf5')
c_results = c_model.predict([x_test]).reshape(-1, 1)
r_results = r_model.predict([x_test]).reshape(-1, 1)
results = list()
for item in zip(c_results, r_results):
if item[0][0] > 0.5:
results.append(item[1][0])
else:
results.append(1.0)
results = np.array(results).reshape(-1, 1)
make_submission(result_path='submissions',
results=results,
model_name='Two_Phase_MLP')
print('***** End UAI-CUP-2017 *****')
if __name__ == '__main__':
submit_flag = True if sys.argv[1] == 'submit' else False
print('***** Start UAI-CUP-2017 *****')
print('Loading data ...')
x_train, y_train, x_dev, y_dev, x_test = load_data_v2(data_path='data')
print('Training Extra Trees Regression model ...')
etr = extra_trees()
etr.fit(x_train, y_train)
etr_y_train = etr.predict(x_train).reshape(-1, 1)
etr_y_dev = etr.predict(x_dev).reshape(-1, 1)
train_mae = mean_absolute_error(y_train, etr_y_train)
print('Train MAE:', train_mae)
dev_mae = mean_absolute_error(y_dev, etr_y_dev)
print('Dev MAE:', dev_mae)
if submit_flag:
print('Generate submission ...')
results = etr.predict(x_test).reshape(-1, 1)
make_submission(result_path='submissions',
results=results,
model_name='Extra_Trees')
print('***** End UAI-CUP-2017 *****')
if __name__ == '__main__':
submit_flag = True if sys.argv[1] == 'submit' else False
print('***** Start UAI-CUP-2017 *****')
print('Loading data ...')
x_train, y_train, x_dev, y_dev, x_test = load_data_v2(data_path='data')
print('Training KNN Regression model ...')
knn_reg = KNeighborsRegressor()
knn_reg.fit(x_train, y_train)
knn_y_train = knn_reg.predict(x_train).reshape(-1, 1)
knn_y_dev = knn_reg.predict(x_dev).reshape(-1, 1)
train_mae = mean_absolute_error(y_train, knn_y_train)
print('Train MAE:', train_mae)
dev_mae = mean_absolute_error(y_dev, knn_y_dev)
print('Dev MAE:', dev_mae)
if submit_flag:
print('Generate submission ...')
results = knn_reg.predict(x_test).reshape(-1, 1)
make_submission(result_path='submissions',
results=results,
model_name='KNN')
print('***** End UAI-CUP-2017 *****')