def test_to_data_frame(self):
# Given
time_series_row1 = TimeSeriesRow(datetime(1960, 8, 1, 11, 0, 0), 112.0)
time_series_row2 = TimeSeriesRow(datetime(1960, 8, 1, 12, 0, 0), 118.0)
time_series_row3 = TimeSeriesRow(datetime(1960, 8, 1, 13, 0, 0), 132.0)
time_series = TimeSeries(
[time_series_row1, time_series_row2, time_series_row3], "Date",
"Passengers", "yyyy-MM-dd HH:mm:ss")
# When
actual_data_frame = time_series.to_data_frame()
# Then
initial_data = {
"Date": [
datetime.strptime("1960-08-01 11:00:00", "%Y-%m-%d %H:%M:%S") +
timedelta(hours=i) for i in range(0, 3)
],
"Passengers": [112.0, 118.0, 132.0]
}
expected_data_frame = pd.DataFrame.from_dict(initial_data)
assert_frame_equal(expected_data_frame, actual_data_frame)
def test_from_json(self):
# Given
json = dict(dateColumnName='Date',
valueColumnName='Passengers',
dateFormat='yyyy-MM-dd HH:mm:ss',
rows=[
dict(date='1949-01-01 11:00:01', value=112.0),
dict(date='1949-01-01 11:00:02', value=118.0),
dict(date='1949-01-01 11:00:03', value=132.0)
])
# When
actual_time_series = TimeSeries.from_json(json)
# Then
time_series_row1 = TimeSeriesRow(datetime(1949, 1, 1, 11, 0, 1), 112.0)
time_series_row2 = TimeSeriesRow(datetime(1949, 1, 1, 11, 0, 2), 118.0)
time_series_row3 = TimeSeriesRow(datetime(1949, 1, 1, 11, 0, 3), 132.0)
expected_time_series = TimeSeries(
[time_series_row1, time_series_row2, time_series_row3], "Date",
"Passengers", "yyyy-MM-dd HH:mm:ss")
assert expected_time_series == actual_time_series
def test_to_json(self):
# Given
time_series_row1 = TimeSeriesRow(datetime(1949, 1, 1), 112.0)
time_series_row2 = TimeSeriesRow(datetime(1949, 2, 1), 118.0)
time_series_row3 = TimeSeriesRow(datetime(1949, 3, 1), 132.0)
time_series = TimeSeries(
[time_series_row1, time_series_row2, time_series_row3], "Date",
"Passengers", "yyyy-MM")
# When
actual_json = time_series.to_json()
# Then
expected_json = dict(dateColumnName='Date',
valueColumnName='Passengers',
dateFormat='yyyy-MM',
rows=[
dict(date='1949-01', value=112.0),
dict(date='1949-02', value=118.0),
dict(date='1949-03', value=132.0)
])
assert expected_json == actual_json
def test_predict(self):
# Given
body = dict(timeSeries=dict(rows=[
dict(date="1949-01", value=112.0),
dict(date="1949-02", value=118.0),
dict(date="1949-03", value=132.0),
dict(date="1949-04", value=129.0),
dict(date="1949-05", value=121.0),
dict(date="1949-06", value=135.0),
dict(date="1949-07", value=148.0),
dict(date="1949-08", value=148.0),
dict(date="1949-09", value=136.0),
dict(date="1949-10", value=119.0),
dict(date="1949-11", value=104.0),
dict(date="1949-12", value=118.0),
dict(date="1950-01", value=115.0),
dict(date="1950-02", value=126.0)
],
dateColumnName='Date',
valueColumnName='Passengers',
dateFormat='yyyy-MM'),
numberOfValues=2)
body_as_string = json.dumps(body)
# When
response = test_app.post('/time-series-analysis/predict',
data=body_as_string,
content_type='application/json')
actual_time_series = TimeSeries.from_json(json.loads(response.data))
# Then
time_series_row = TimeSeriesRow(datetime(1950, 3, 1), 125.0)
time_series_row1 = TimeSeriesRow(datetime(1950, 4, 1), 124.0)
expected_time_series = TimeSeries([time_series_row, time_series_row1],
"Date", "Passengers", "yyyy-MM")
assert expected_time_series == actual_time_series
def test_from_json(self):
# Given
json = dict(timeSeries=dict(dateColumnName='Date',
valueColumnName='Passengers',
dateFormat='yyyy-MM',
rows=[dict(date='1949-01', value=112.0),
dict(date='1949-02', value=118.0),
dict(date='1949-03', value=132.0)]),
numberOfValues=5)
# When
actual_time_series_analysis_request = TimeSeriesAnalysisRequest.from_json(json)
# Then
time_series_row1 = TimeSeriesRow(datetime(1949, 1, 1), 112.0)
time_series_row2 = TimeSeriesRow(datetime(1949, 2, 1), 118.0)
time_series_row3 = TimeSeriesRow(datetime(1949, 3, 1), 132.0)
time_series = TimeSeries([time_series_row1, time_series_row2, time_series_row3],
"Date", "Passengers", "yyyy-MM")
expected_time_series_analysis_request = TimeSeriesAnalysisRequest(time_series, 5)
assert expected_time_series_analysis_request == actual_time_series_analysis_request
def predict() -> str:
"""
Predict exact values.
Arguments
- time_series_analysis_request_json (str) : json corresponding to time_series_analysis_request_json
file path of the csv input file
Returns
time_series -> time_series corresponding to the predicted values and dates.
"""
try:
get_logger().info("[Start] predict request")
time_series_analysis_request = TimeSeriesAnalysisRequest.from_json(request.json)
time_series = time_series_analysis_request.get_time_series()
data = time_series.to_data_frame()
date_column_name = time_series.get_date_column_name()
value_column_name = time_series.get_value_column_name()
number_of_values = time_series_analysis_request.get_number_of_values()
date_format = time_series.get_date_format()
time_series_analysis_service = TimeSeriesAnalysisService(data, date_column_name, value_column_name,
number_of_values)
predicted_data_frame = time_series_analysis_service.predict()
predicted_time_series = TimeSeries.from_data_frame(predicted_data_frame, date_column_name,
value_column_name, date_format)
return json.dumps(predicted_time_series, cls=JsonComplexEncoder)
except Exception as exception:
error_message = "Exception %s raised while predicting: %s" % (type(exception).__name__, exception)
get_logger().error(error_message)
get_logger().exception(exception)
raise EngineComputationException(error_message)
finally:
get_logger().info("[End] predict request")
def from_json(cls, data: dict):
time_series = TimeSeries.from_json(data["timeSeries"])
return cls(time_series, int(data["numberOfValues"]))