def test_model_prediction_differential(
*,
input_data_file: str = 'previous_input_data.csv',
output_data_file: str = 'previous_output_data.csv'
):
previous_data_df = load_dataset(filename=output_data_file)
previous_output_data = previous_data_df.predictions.values
current_data_df = load_dataset(filename=input_data_file)
current_data_json = current_data_df.to_json(orient='records')
current_output_data_json = make_prediction(input_json=current_data_json)
current_output_data = current_output_data_json.get('predictions')
assert len(previous_output_data) == len(current_output_data)
for previous_value, current_value in zip(previous_output_data, current_output_data):
previous_value = previous_value.item()
current_value = current_value.item()
assert math.isclose(
previous_value,
current_value,
rel_tol=config.ACCEPTABLE_MODEL_DIFFERENCE
)
def test_make_single_prediction():
test_data = np.random.randn(1, 15)
df = pd.DataFrame(test_data)
single_test_json = df.to_json(orient='records')
subject = make_prediction(input_json=single_test_json)
assert subject is not None
assert isinstance(subject.get('predictions')[0], np.integer)
assert subject.get('predictions')[0] >= 0
assert subject.get('predictions')[0] < 7
def test_make_multiple_prediction():
test_data = np.random.randn(50, 15)
original_data_length = len(test_data)
df = pd.DataFrame(test_data)
multiple_test_json = df.to_json(orient='records')
subject = make_prediction(input_json=multiple_test_json)
assert subject is not None
assert len(subject.get('predictions')) == 50
assert len(subject.get('predictions')) == original_data_length
def capture_previous_data(*,
input_data_file: str = 'previous_input_data.csv',
output_data_file: str = 'previous_output_data.csv'):
input_data = np.random.randn(50, 15)
input_data_df = pd.DataFrame(input_data)
input_data_df.to_csv(
f'{config.PACKAGE_ROOT.parent}/ml_model/ml_model/datasets/{input_data_file}'
)
input_data_json = input_data_df.to_json(orient='records')
output_data = make_prediction(input_json=input_data_json)
output_data_df = pd.DataFrame(output_data)
output_data_df.to_csv(
f'{config.PACKAGE_ROOT.parent}/ml_model/ml_model/datasets/{output_data_file}'
)
def svc():
if request.method == 'POST':
_logger.info('End-point SVC')
import json
request_json = request.get_json()
json_data = json.dumps(request.get_json())
result = make_prediction(input_json=json_data)
_logger.info(f'Result: {result}')
predictions = result.get('predictions')
version = result.get('version')
response = jsonify({
'predictions': predictions.tolist(),
'version': version
})
return response