def main():
parser = argparse.ArgumentParser()
parser.add_argument(
"-f",
"--filename",
type=str,
help="Path to file with data",
)
parser.add_argument(
"-ts",
"--train-size",
type=float,
help="Percentage of dataset used in the train set.",
)
parser.add_argument(
"-srt",
"--sort-train-set",
action="store_true",
help="Sorts train set after spliting dataset to train and test set",
)
args = parser.parse_args()
_filename = args.filename
_train_size = args.train_size
_sort_train_set = args.sort_train_set
dataset = load_data_from_file(_filename)
# uncomment to check grouping classes by two chosen features
# plot_labels_by_two_features(X1, X2, dataset)
results = pd.DataFrame()
train_set, test_set = prepare_train_and_test_set(dataset, _train_size,
_sort_train_set)
classes = get_test_set_classes(test_set)
(
mean_values,
std_variation_values,
prior_probabilities,
) = calculate_statistics_by_class(train_set)
results = results.append(
test_set_classification(test_set, mean_values, std_variation_values,
prior_probabilities))
confusion_matrix = ConfusionMatrix(results, classes)
confusion_matrix.plot_confusion_matrix()
confusion_matrices = []
for i in range(10):
train_set, test_set = prepare_train_and_test_set(
dataset, _train_size, _sort_train_set)
classes = get_test_set_classes(test_set)
(
mean_values,
std_variation_values,
prior_probabilities,
) = calculate_statistics_by_class(train_set)
results = test_set_classification(test_set, mean_values,
std_variation_values,
prior_probabilities)
cm = ConfusionMatrix(results, classes)
cm.calculate_metrics_by_classes()
confusion_matrices.append(cm)
accuracy_values = []
precision_values = {}
recall_values = {}
for matrix in confusion_matrices:
accuracy_values.append(matrix.accuracy)
for class_name in matrix.metrics:
if class_name not in precision_values:
precision_values[class_name] = []
recall_values[class_name] = []
precision_values[class_name].append(
matrix.metrics[class_name]["precision"])
recall_values[class_name].append(
matrix.metrics[class_name]["recall"])
print(
f"Classifier overall accuracy statistics after 10 runs:\nMin: {min(accuracy_values)}\nMax: {max(accuracy_values)}\n"
+
f"Mean: {np.mean(accuracy_values)}\nStd deviation: {np.std(accuracy_values)}\n"
)
for class_name in precision_values:
print(
f"{class_name} precision statistics after 10 runs:\nMin: {min(precision_values[class_name])}\nMax: {max(precision_values[class_name])}\n"
+
f"Mean: {np.mean(precision_values[class_name])}\nStd deviation: {np.std(precision_values[class_name])}\n"
)
print(
f"{class_name} recall statistics after 10 runs:\nMin: {min(recall_values[class_name])}\nMax: {max(recall_values[class_name])}\n"
+
f"Mean: {np.mean(recall_values[class_name])}\nStd deviation: {np.std(recall_values[class_name])}\n"
)
