def create_model_error_data_test(load_data, model, test_data_start_date,
test_data_end_date, highlight_variable,
highlight_value):
#generates model forecast dependent on the model type
if (model['model_type'] == 'displacement'):
predicted_load_data = displacement_model.predict(
load_data, test_data_start_date, test_data_end_date, model)
y_pred = predicted_load_data['Load']
elif (model['model_type'] == 'regression'):
y_pred = regression_model.predict(load_data, test_data_start_date,
test_data_end_date, model)
test_data = load_data[(load_data.Date >= test_data_start_date)
& (load_data.Date <= test_data_end_date)]
y_true = test_data['Load']
highlighted_load_data = pd.DataFrame(columns=['Date'])
if (highlight_variable != 'None'
and highlight_value != 'None'): #there is highlighted data chosen
highlighted_load_data['Date'], highlighted_load_data[
'y_axis'] = zip( #get highlighted data entries
*test_data.apply(lambda row: get_highlighted_entries_dataframe(
row, highlight_variable, highlight_value, 'Load'),
axis=1))
model_error_df = pd.DataFrame()
model_error_df = create_model_error_data_visualised(
test_data['Date'], y_pred, y_true, highlighted_load_data['Date'].
tolist()) #generate model error data traces
return model_error_df
def get_test_data_metrics(load_data, models, test_data_start_date, test_data_end_date):
test_data_metrics = []
if (test_data_start_date == None or test_data_end_date == None): #if there is no test data selected, return no metrics
return test_data_metrics
test_data = load_data[(load_data.Date >= test_data_start_date)
& (load_data.Date <= test_data_end_date)]
y_true = test_data['Load']
for model in models: #iterate through all load forecasting models
model_type = model['model_type']
#predict logic for each type of load forecasting model
if (model_type == 'displacement'):
predicted_load_data = displacement_model.predict(
load_data, test_data_start_date, test_data_end_date, model)
y_pred = predicted_load_data['Load']
elif (model_type == 'regression'):
y_pred = regression_model.predict(
load_data, test_data_start_date, test_data_end_date, model)
model_error_df = pd.DataFrame() #model structure containing the error data used to calculate error distribution metrics
model_error_df = create_model_error_data(
model_error_df, y_true, y_pred)
test_data_metrics.append(model_metrics.calculate_metrics(
y_true, y_pred, model_error_df['APE'], model_error_df['Percentage']))
return test_data_metrics
def test_predict_load(self):
model = dict(coefficients=[5.0],
intercept=20,
x_columns=['Temperature'],
corrected_column=None)
start_date = '2018-01-01'
end_date = '2018-01-02'
expected_output = [25, 30, 35]
output = predict(self.load_data, start_date, end_date, model)
self.assertEqual(expected_output, output)
def test_predict_load_with_corrected_column(self):
model = dict(coefficients=[5.0],
intercept=20,
x_columns=['Temperature'],
corrected_column='Load Last Week')
start_date = '2018-01-01'
end_date = '2018-01-02'
expected_output = [197, 414, 631]
output = predict(self.load_data, start_date, end_date, model)
self.assertEqual(expected_output, output)
def test_predict_load_multiple_coefficients(self):
model = dict(coefficients=[5.0, 1.0],
intercept=100,
x_columns=['Temperature', 'Load Last Week'],
corrected_column=None)
start_date = '2018-01-02'
end_date = '2018-01-03'
expected_output = [781, 1008, 1235]
output = predict(self.load_data, start_date, end_date, model)
self.assertEqual(expected_output, output)
def get_regression_model_traces(load_data, visualised_load_data, traces, model, model_name, model_color, y_axis, start_date, end_date):
try:
predicted_load_data = regression_model.predict(
load_data, start_date, end_date, model, y_axis)
except Exception as predict_error:
raise predict_error
predicted_load_traces = generate_traces(
visualised_load_data['Date'], predicted_load_data, model_name, color=model_color)
traces.extend(predicted_load_traces)
return traces