def run_model(classifier_name, data, epoch, window_len, stride, binary, i):
output_directory = data[7] + classifier_name + "_" + str(i) + "/"
create_directory(output_directory)
if classifier_name == "rocket":
metrics, conf_mat = run_rocket(data, epoch, window_len, stride, binary)
else:
metrics, conf_mat = run_deep_learning_models(classifier_name, data,
epoch, output_directory)
metrics.to_csv(output_directory + 'classification_metrics.csv')
np.savetxt(output_directory + 'confusion_matrix.csv',
conf_mat,
delimiter=",")
return metrics
metrics = pd.concat([metrics_train, metrics_val,
metrics_test]).reset_index(drop=True)
print(metrics.head())
metrics.to_csv(output_directory + 'classification_metrics.csv')
np.savetxt(output_directory + 'confusion_matrix.csv',
conf_mat,
delimiter=",")
if len(sys.argv) >= 6:
data_path = sys.argv[1]
output_directory = sys.argv[2]
problem = sys.argv[3]
classifier_name = sys.argv[4]
itr = sys.argv[5]
else:
cwd = os.getcwd()
data_path = cwd + "/TS_Segmentation/"
output_directory = cwd + "/output/"
problem = "Emotiv266"
classifier_name = "MHA"
itr = "itr_0"
output_directory = output_directory + classifier_name + '/' + problem + '/' + itr + '/'
create_directory(output_directory)
run_experiments()
TIME_ZONE = 'Europe/Moscow'
USE_I18N = True
USE_L10N = True
USE_TZ = True
STATIC_URL = '/static/'
CELERY_BROKER = 'redis://localhost:6379'
FILE_LOGGER = '/tmp/django_telegram/logs/debug.log'
FILE_HTTP_CLIENT_LOGGER = '/tmp/django_telegram/logs/http_client.log'
create_directory(FILE_LOGGER); create_directory(FILE_HTTP_CLIENT_LOGGER)
# CELERYD_HIJACK_ROOT_LOGGER = False # TODO True
LOGGING = {
'version': 1,
'formatters': {
'verbose': {
'format': '%(levelname)s %(asctime)s %(module)s %(process)d %(thread)d %(message)s'
},
'verbose_with_extra': {
'format': '%(levelname)s %(asctime)s %(module)s %(process)d %(thread)d '
'%(extra)s %(message)s'
},
'simple': {
'format': '%(levelname)s %(message)s'
},
def prepare_data_cnn_lstm(problem, window_len, stride, binary, data_version,
upsampled, norm, segmentation_norm):
# Set up output location
cwd = os.getcwd()
data_path = cwd + "/TS_Segmentation/"
output_directory = cwd + "/output/"
output_directory = output_directory + "compare_models" + '/' + problem + '/'
create_directory(output_directory)
print(
"#########################################################################"
)
print("[Compare_Models] Run Compare_Models")
print(
"#########################################################################"
)
print("[Compare_Models] Data path: {}".format(data_path))
print("[Compare_Models] Output Dir: {}".format(output_directory))
print("[Compare_Models] Problem: {}".format(problem))
print("[Compare_Models] Window Len: {}".format(window_len))
print("[Compare_Models] Stride: {}".format(stride))
print(
"#########################################################################"
)
data_folder = data_path + problem + "/"
train_file = data_folder + problem + "_TRAIN.csv"
test_file = data_folder + problem + "_TEST.csv"
train_data = data_loader.load_segmentation_data(train_file,
norm=segmentation_norm)
test_data = data_loader.load_segmentation_data(test_file,
norm=segmentation_norm)
print("[Compare_Models] {} train series".format(len(train_data)))
print("[Compare_Models] {} test series".format(len(test_data)))
dataset1, dataset2 = prepare_inputs_combined(train_inputs=train_data,
test_inputs=test_data,
window_len=window_len,
stride=stride,
binary=binary,
data_version=data_version,
upsampled=upsampled,
norm=norm)
X_train, y_train, X_val, y_val, X_test, y_test = dataset1[0], dataset1[
1], dataset1[2], dataset1[3], dataset1[4], dataset1[5]
X2_train, y2_train, X2_val, y2_val, X2_test, y2_test = dataset2[
0], dataset2[1], dataset2[2], dataset2[3], dataset2[4], dataset2[5]
if y_val is not None:
all_labels = np.concatenate((y_train, y_val, y_test), axis=0)
else:
all_labels = np.concatenate((y_train, y_test), axis=0)
print("[Compare_Models] All labels: {}".format(np.unique(all_labels)))
tmp = pd.get_dummies(all_labels).values
y_train = tmp[:len(y_train)]
y_val = tmp[len(y_train):len(y_train) + len(y_val)]
y_test = tmp[len(y_train) + len(y_val):]
data_deep_learning = [
all_labels, X_train, y_train, X_val, y_val, X_test, y_test,
output_directory
]
if y2_val is not None:
all_labels2 = np.concatenate((y2_train, y2_val, y2_test), axis=0)
else:
all_labels2 = np.concatenate((y2_train, y2_test), axis=0)
print("[Compare_Models] All labels 2: {}".format(np.unique(all_labels2)))
tmp2 = pd.get_dummies(all_labels2).values
y2_train = tmp2[:len(y2_train)]
y2_val = tmp2[len(y2_train):len(y2_train) + len(y2_val)]
y2_test = tmp2[len(y2_train) + len(y2_val):]
data_cnn_lstm = [
all_labels2, X2_train, y2_train, X2_val, y2_val, X2_test, y2_test,
output_directory
]
return data_deep_learning, data_cnn_lstm
