added tests for all the different scenarios

caimira/apps/calculator/co2_model_generator.py

@@ -93,7 +93,8 @@ def find_change_points(self) -> list:
         """
         Perform change point detection using scipy library (find_peaks method) with rolling average of data.
         Incorporate existing state change candidates and adjust the result accordingly.
-        Returns a list of the detected ventilation state changes.
+        Returns a list of the detected ventilation state changes, discarding any contribution
+        of occupancy state changes.
         """
         times: list = self.CO2_data['times']
         CO2_values: list = self.CO2_data['CO2']
@@ -169,6 +170,12 @@ def population_present_changes(self, infected_presence: models.Interval, exposed
         return sorted(state_change_times)
 
     def ventilation_transition_times(self) -> typing.List[float]:
+        '''
+        Check if the last time from the input data is
+        included in the ventilation ventilations state.
+        Given that the last time is a required state change, 
+        if not included, this method adds it.
+        '''
         vent_states = self.fitting_ventilation_states
         last_time_from_input = self.CO2_data['times'][-1]
         if (vent_states and last_time_from_input != vent_states[-1]): # The last time value is always needed for the last ACH interval.

caimira/tests/models/test_co2_concentration_model.py

@@ -47,40 +47,62 @@ def test_integrated_concentration(simple_co2_conc_model):
 
 
 @pytest.mark.parametrize(
-    "scenario_data, room_volume, total_people, start, finish, state_changes", [
-        ["office_scenario_1_sensor_data", 102, 4, "14:00", "17:30", (14.78, 15.10, 15.53, 15.87, 16.52, 16.83)],
-        ["office_scenario_2_sensor_data", 60, 2, "08:38", "17:30", (10.17, 12.45, 14.50)], # Second should be 12.87
+    "scenario_data, room_volume, max_total_people, start, finish, state_changes", [
+        ["office_scenario_1_sensor_data", 102, 4, "14:00", "17:30", (14.78, 15.1, 15.53, 15.87, 16.52, 16.83)],
+        ["office_scenario_2_sensor_data", 60, 2, "08:38", "17:30", (10.17, 12.45, 14.5)], # Second should be 12.87
         ["meeting_scenario_1_sensor_data", 83, 3, "09:04", "11:45", (10.37, 11.07)],
-        ["meeting_scenario_2_sensor_data", 83, 4, "13:40", "16:40", (14.37, 14.70, 14.98, 15.33, 15.68, 16.02)]
+        ["meeting_scenario_2_sensor_data", 83, 4, "13:40", "16:40", (14.37, 14.72, 15, 15.33, 15.68, 16.03)]
     ]
 )
-def test_find_change_points(scenario_data, room_volume, total_people, start, finish, state_changes, request):
+def test_find_change_points(scenario_data, room_volume, max_total_people, start, finish, state_changes, request):
+    '''
+    Specific test of the find_change_points method using the
+    scipy find_peaks and specific smoothing techniques. Only
+    the ventilation state changes are target for detection.
+    '''
     CO2_form_model: CO2FormData = CO2FormData(
         CO2_data=request.getfixturevalue(scenario_data),
         fitting_ventilation_states=[],
         exposed_start=start,
         exposed_finish=finish,
-        total_people=total_people,
+        total_people=max_total_people,
         room_volume=room_volume,
     )
     find_points = CO2_form_model.find_change_points()
     assert np.allclose(find_points, state_changes, rtol=1e-2)
 
 
-def test_predictive_model_accuracy(data_registry, office_scenario_2_sensor_data):
+@pytest.mark.parametrize(
+    "scenario_data, room_volume, occupancy, presence_interval, all_state_changes", [
+        ["office_scenario_1_sensor_data", 102, (4,), (14, 17.5), (14, 14.25, 14.78, 15.1, 15.53, 15.87, 16.52, 16.83, 17.5)],
+        ["office_scenario_2_sensor_data", 60, (2, 0, 2), (8.62, 11.93, 12.42, 17.5), (8.62, 10.17, 12.45, 14.5, 17.5, 20.)],
+        ["meeting_scenario_1_sensor_data", 83, (2, 3, 2, 3), (9.07, 9.32, 9.75, 10.75, 11.75), (9.07, 10.37, 11.07, 11.75)],
+        ["meeting_scenario_2_sensor_data", 83, (2, 3, 4), (13.67, 13.75, 15.87, 16.67), (13.67, 14.37, 14.72, 15.00, 15.33, 15.68, 16.03, 16.67)]
+    ]
+)
+def test_predictive_model_accuracy(data_registry, scenario_data, room_volume, occupancy, presence_interval, all_state_changes, request):
     '''
-    Specific test corresponding to the template data from a simulation day
-    in one office in Geneva. The room volume, number of people and ventilation
-    transition times correspond to the real occurencies in the simulation day.
+    Specific test corresponding to specific data files of four
+    different occurencies (2 office and 2 meeting room scenarios). 
+    The room volume, number of people and ventilation transition times 
+    correspond to the real occurencies in the simulation days.
+
+    Note that the last time from the input file is considered as a ventilation
+    state change.
     '''
+    input_fitting_data = request.getfixturevalue(scenario_data)
+
     fitting_model: models.CO2DataModel = models.CO2DataModel(
         data_registry=data_registry,
-        room_volume=59.787,
-        number=2,
-        presence=models.SpecificInterval(((8.63, 11.95), (12.42, 17.5))),
-        ventilation_transition_times=(8.63, 10.17, 12.89, 14.5, 17.5, 22.),
-        times=office_scenario_2_sensor_data['times'],
-        CO2_concentrations=office_scenario_2_sensor_data['CO2'],
+        room_volume=room_volume,
+        number=models.IntPiecewiseConstant(
+            transition_times=presence_interval,
+            values=occupancy
+        ),
+        presence=None,
+        ventilation_transition_times=all_state_changes,
+        times=input_fitting_data['times'],
+        CO2_concentrations=input_fitting_data['CO2'],
     )
     # Get fitting results
     fitting_results: typing.Dict = fitting_model.CO2_fit_params()
@@ -90,7 +112,8 @@ def root_mean_square_error_percentage(actual, predicted) -> float:
         return np.sqrt(np.mean(((actual - predicted) / actual) ** 2)) * 100
 
     # Calculate RMSEP metric
-    rmsep = root_mean_square_error_percentage(np.array(office_scenario_2_sensor_data['CO2']), np.array(predictive_CO2))
+    rmsep = root_mean_square_error_percentage(np.array(input_fitting_data['CO2']), np.array(predictive_CO2))
+
     acceptable_rmsep = 10 # Threshold of 10% for the accepted error margin
     assert rmsep <= acceptable_rmsep, f"RMSEP {rmsep} exceeds acceptable threshold {acceptable_rmsep}"