def run_basic_agent():
"""
Runs the basic agent program.
:return:
"""
# Retrieve data from CSV file and encode it.
data = None
if config.csv_file == "tickets":
data = DataProcessor(config.csv_file)
elif config.csv_file == "AND_gate":
data = DataProcessor("AND_gate")
elif config.csv_file == "OR_gate":
data = DataProcessor("OR_gate")
else:
print("CSV file {} could not be found, please check spelling.")
exit(1)
encode_data(data)
# Run Grid Search algorithm to determine optimal hyperparameters for the neural network.
if config.is_grid_search:
gs = GridSearch(data.input_data_encoded, data.target_data_encoded)
run_grid_search(gs)
# Train and test the neural network based on the chosen data.
else:
mlp = create_multi_layer_perceptron(data)
start_time = time.time() # Start measuring runtime.
run_multi_layer_perceptron(mlp)
print_runtime("Training", round(time.time() - start_time,
2)) # Record and print runtime.
def run_grid_search(self):
"""
Runs the grid search algorithm, testing every possible combination in the grid_search_params dictionnary.
Measures runtime as well.
:return: None.
"""
start_time = time.time() # Start measuring runtime.
# Run Grid Search using all possible combinations in grid_search_params.
self.grid_search.fit(self.X_train, self.y_train)
print_runtime("Grid Search", round(time.time() - start_time,
2)) # Record and print runtime.
def train_classification_models(X, y, ground_truth) -> None:
"""
:param X:
:param y:
:param ground_truth:
:return:
"""
# Start recording time.
start_time = time.time()
# Create classifier model instance.
Classifier(config.model, X, y, ground_truth)
# Print training runtime.
print_runtime(round(time.time() - start_time, 2))
def run_advanced_agent():
"""
Runs the basic agent program.
:return:
"""
# Retrieve data from CSV file and encode it.
data = DataProcessor(config.csv_file)
encode_data(data)
mlp = MultiLayerPerceptron(
classifier_model="dt",
name=config.csv_file,
input_data=data.input_data_encoded,
target_data=data.target_data_encoded,
)
start_time = time.time() # Start measuring runtime.
run_multi_layer_perceptron(mlp)
print_runtime("Training", round(time.time() - start_time,
2)) # Record and print runtime.
def main() -> None:
"""
Program entry point. Parses command line arguments to decide which agent to run.
:return: None
"""
parse_command_line_arguments()
if config.verbose_mode:
print("Verbose mode: ON\n")
# Make final predictions on X_test.csv (produce 'Y_test.csv' for practical submission).
if config.section == "test":
X_test = load_testing_data()
# Run this once to save the CSV files imported into DFs in PKL format for quicker loading times.
# save_df_to_pickle(X_test, config.dataset, "X_test")
make_final_test_predictions(X_test)
# Explore and train the classifiers.
else:
X_train, y_train = load_training_data()
# Run this once to save the CSV files imported into DFs in PKL format for quicker loading times.
# save_df_to_pickle(X_train, config.dataset, "X_train")
# save_df_to_pickle(y_train, config.dataset, "y_train")
# Visualise data.
if config.section == "data_vis":
start_time = time.time()
# Split training dataset's features in 3 distinct DFs.
X_train_HoG, X_train_normal_dist, X_train_colour_hists = split_features(
X_train)
# Visualise data.
data_overview(X_train)
visualise_hog(X_train_HoG)
visualise_rgb_hist(X_train_colour_hists)
visualise_class_distribution(y_train)
visualise_correlation(X_train, y_train)
print_runtime(round(time.time() - start_time, 2))
# Train classification models.
elif config.section == "train":
# Over sample binary dataset.
if config.dataset == "binary":
X_train, y_train = over_sample(X_train, y_train)
# Run this once to save the CSV files imported into DFs in PKL format for quicker loading times.
# save_df_to_pickle(X_train_resampled, config.dataset, "X_train_resampled")
# save_df_to_pickle(y_train_resampled, config.dataset, "Y_train_resampled")
if config.verbose_mode:
visualise_class_distribution(y_train)
X = input_preparation(X_train)
y, ground_truth = output_preparation(y_train)
if config.section == "train":
train_classification_models(X, y, ground_truth)
else:
print_error_message()