def co_disag():
# TRAIN
disag = CombinatorialOptimisation()
disag.train(meters)
# TEST
appliance_powers = disag.disaggregate_chunk(mains)
for i, df in appliance_powers.iteritems():
appliance = disag.model[i]['training_metadata'].dominant_appliance()
appliance_type = appliance.identifier.type
y_pred = df.values
if appliance_type in MAINS_APPLIANCES:
np.save(join(BASE_DIRECTORY, 'CO_' + appliance_type), y_pred)
appliance_powers = disag.disaggregate_chunk(fridge_mains)
for i, df in appliance_powers.iteritems():
appliance = disag.model[i]['training_metadata'].dominant_appliance()
appliance_type = appliance.identifier.type
y_pred = df.values
if appliance_type in ['fridge freezer']:
np.save(join(BASE_DIRECTORY, 'CO_' + appliance_type), y_pred)
appliance_powers = disag.disaggregate_chunk(kettle_mains)
for i, df in appliance_powers.iteritems():
appliance = disag.model[i]['training_metadata'].dominant_appliance()
appliance_type = appliance.identifier.type
y_pred = df.values
if appliance_type in ['kettle']:
np.save(join(BASE_DIRECTORY, 'CO_' + appliance_type), y_pred)
def test_co_correctness(self):
elec = self.dataset.buildings[1].elec
co = CombinatorialOptimisation()
co.train(elec)
mains = elec.mains()
pred = co.disaggregate_chunk(mains.load(sample_period=1).next())
gt = {}
for meter in elec.submeters().meters:
gt[meter] = meter.load(sample_period=1).next().squeeze()
gt = pd.DataFrame(gt)
pred = pred[gt.columns]
self.assertTrue(gt.equals(pred))
def test_co_correctness(self):
elec = self.dataset.buildings[1].elec
co = CombinatorialOptimisation()
co.train(elec)
mains = elec.mains()
pred = co.disaggregate_chunk(next(mains.load(sample_period=1)))
gt = {}
for meter in elec.submeters().meters:
gt[meter] = next(meter.load(sample_period=1)).squeeze()
gt = pd.DataFrame(gt)
pred = pred[gt.columns]
self.assertTrue(gt.equals(pred))
if algorithm == 'fhmm':
clf = fhmm_exact.FHMM()
elif algorithm == 'combOpt':
print('here')
clf = CombinatorialOptimisation()
start = time.time()
clf.train(train_elec, sample_period=samplePeriod)
end = time.time()
print('Training runtime =', end - start, 'seconds.')
# make predicitons
pred = {}
testChunks = test_elec.mains().load(sample_period=samplePeriod)
for i, chunk in enumerate(testChunks):
chunk_drop_na = chunk.dropna()
pred[i] = clf.disaggregate_chunk(chunk_drop_na)
print('---------------------------------')
print('Testing done')
print('---------------------------------')
# If everything can fit in memory
pred_overall = pd.concat(pred)
pred_overall.index = pred_overall.index.droplevel()
# use appliance names as the labels
appliance_labels = []
for m in pred_overall.columns.values:
name = m.appliances[0].metadata['original_name']
name = name.replace('_', ' ')
name = name.capitalize()
appliance_labels.append(name)
pred_overall.columns = appliance_labels
