def savePredictionsStep(self):
for i, currProb in enumerate(self.predictedProb):
currSectionEntry = self.toPredictSections[i]
currSection = currSectionEntry.data
currProb = currProb
# Insert the prediction
mp = ecwm.Modeprediction()
mp.trip_id = currSection.trip_id
mp.section_id = currSectionEntry.get_id()
mp.algorithm_id = ecwm.AlgorithmTypes.SIMPLE_RULE_ENGINE
mp.predicted_mode_map = currProb
mp.start_ts = currSection.start_ts
mp.end_ts = currSection.end_ts
self.ts.insert_data(self.user_id, "inference/prediction", mp)
# There are now two predictions, but don't want to do a bunch of
# refactoring, so just create the inferred # section object right here
is_dict = copy.copy(currSectionEntry)
del is_dict["_id"]
is_dict["metadata"]["key"] = "analysis/inferred_section"
is_dict["data"]["sensed_mode"] = ecwm.PredictedModeTypes[
easf.select_inferred_mode([mp])].value
is_dict["data"]["cleaned_section"] = currSectionEntry.get_id()
ise = ecwe.Entry(is_dict)
logging.debug("Updating sensed mode for section = %s to %s" %
(currSectionEntry.get_id(), ise.data.sensed_mode))
self.ts.insert(ise)
# Set last_section_done after saving because otherwise if there is an error
# during inference, we will not save results and never re-run
self.last_section_done = self.toPredictSections[-1]
def testSelectInferredModeImbalanced(self):
# dummy_inferred_mode = dim
dim = ecwm.Modeprediction()
dim.trip_id = "test_trip_id"
dim.section_id = "test_section_id"
dim.algorithm_id = ecwm.AlgorithmTypes.SEED_RANDOM_FOREST
dim.sensed_mode = ecwma.MotionTypes.WALKING
dim.predicted_mode_map = {"WALKING": 0.9, "BUS": 0.1}
dim.start_ts = 0
dim.end_ts = 0
self.assertEqual(fc.select_inferred_mode([dim]), 'WALKING')
def testSelectInferredModeEqual(self):
# dummy_inferred_mode = dim
# select the lower value mode, typically = lower carbon intensity mode
dim = ecwm.Modeprediction()
dim.trip_id = "test_trip_id"
dim.section_id = "test_section_id"
dim.algorithm_id = ecwm.AlgorithmTypes.SEED_RANDOM_FOREST
dim.sensed_mode = ecwma.MotionTypes.WALKING
dim.predicted_mode_map = {"WALKING": 0.5, "BUS": 0.5}
dim.start_ts = 0
dim.end_ts = 0
self.assertEqual(fc.select_inferred_mode([dim]), 'WALKING')
def savePredictionsStep(self):
from emission.core.wrapper.user import User
from emission.core.wrapper.client import Client
uniqueModes = self.model.classes_
for i in range(self.predictedProb.shape[0]):
currSectionEntry = self.toPredictSections[i]
currSection = currSectionEntry.data
currProb = self.convertPredictedProbToMap(uniqueModes,
self.predictedProb[i])
# Special handling for the AIR mode
# AIR is not a mode that is sensed from the phone, but it is inferred
# through some heuristics in cleanAndResample instead of through the
# decision tree. Ideally those heurstics should be replaced by the
# inference through the decision tree, or through a separate heuristic
# step. But we are out of time for a bigger refactor here.
# so we say that if the sensed mode == AIR, we are going to use it
# directly and ignore the inferred mode
if currSection.sensed_mode == ecwma.MotionTypes.AIR_OR_HSR:
currProb = {'AIR_OR_HSR': 1.0}
# Insert the prediction
mp = ecwm.Modeprediction()
mp.trip_id = currSection.trip_id
mp.section_id = currSectionEntry.get_id()
mp.algorithm_id = ecwm.AlgorithmTypes.SEED_RANDOM_FOREST
mp.predicted_mode_map = currProb
mp.start_ts = currSection.start_ts
mp.end_ts = currSection.end_ts
self.ts.insert_data(self.user_id, "inference/prediction", mp)
# Since there is currently only one prediction, create the inferred
# section object right here
is_dict = copy.copy(currSectionEntry)
del is_dict["_id"]
is_dict["metadata"]["key"] = "analysis/inferred_section"
is_dict["data"]["sensed_mode"] = ecwm.PredictedModeTypes[
easf.select_inferred_mode([mp])].value
is_dict["data"]["cleaned_section"] = currSectionEntry.get_id()
ise = ecwe.Entry(is_dict)
logging.debug("Updating sensed mode for section = %s to %s" %
(currSectionEntry.get_id(), ise.data.sensed_mode))
self.ts.insert(ise)
# Set last_section_done after saving because otherwise if there is an error
# during inference, we will not save results and never re-run
self.last_section_done = self.toPredictSections[-1]
def savePredictionsStep(self):
from emission.core.wrapper.user import User
from emission.core.wrapper.client import Client
uniqueModes = self.model.classes_
for i in range(self.predictedProb.shape[0]):
currSectionEntry = self.toPredictSections[i]
currSection = currSectionEntry.data
currProb = self.convertPredictedProbToMap(uniqueModes, self.predictedProb[i])
# Special handling for the AIR mode
# AIR is not a mode that is sensed from the phone, but it is inferred
# through some heuristics in cleanAndResample instead of through the
# decision tree. Ideally those heurstics should be replaced by the
# inference through the decision tree, or through a separate heuristic
# step. But we are out of time for a bigger refactor here.
# so we say that if the sensed mode == AIR, we are going to use it
# directly and ignore the inferred mode
if currSection.sensed_mode == ecwma.MotionTypes.AIR_OR_HSR:
currProb = {'AIR_OR_HSR': 1.0}
# Insert the prediction
mp = ecwm.Modeprediction()
mp.trip_id = currSection.trip_id
mp.section_id = currSectionEntry.get_id()
mp.algorithm_id = ecwm.AlgorithmTypes.SEED_RANDOM_FOREST
mp.predicted_mode_map = currProb
mp.start_ts = currSection.start_ts
mp.end_ts = currSection.end_ts
self.ts.insert_data(self.user_id, "inference/prediction", mp)
# Since there is currently only one prediction, create the inferred
# section object right here
is_dict = copy.copy(currSectionEntry)
del is_dict["_id"]
is_dict["metadata"]["key"] = "analysis/inferred_section"
is_dict["data"]["sensed_mode"] = ecwm.PredictedModeTypes[easf.select_inferred_mode([mp])].value
is_dict["data"]["cleaned_section"] = currSectionEntry.get_id()
ise = ecwe.Entry(is_dict)
logging.debug("Updating sensed mode for section = %s to %s" %
(currSectionEntry.get_id(), ise.data.sensed_mode))
self.ts.insert(ise)