# Experiment on first (and only) building
b = dataset.buildings[1]
# Filtering to include only top 8 appliances
b = filter_top_k_appliances(b, 3)
# Dividing the data into train and test
train, test = train_test_split(b)
# Again subdivide data into train, test for testing on even smaller data
#train, test = train_test_split(train, test_size=.5)
# Initializing FHMM Disaggregator
disaggregator = FHMM()
train_mains = train.utility.electric.mains[
train.utility.electric.mains.keys()[0]][DISAGG_FEATURE]
# Get appliances data
app = train.utility.electric.appliances
train_appliances = pd.DataFrame({appliance: app[appliance][DISAGG_FEATURE] for appliance in app if DISAGG_FEATURE in app[appliance]})
# Train
t1 = time.time()
disaggregator.train(train, disagg_features=[DISAGG_FEATURE])
t2 = time.time()
print("Runtime to train = {:.2f} seconds".format(t2 - t1))
# In[ ]:
from nilmtk.disaggregate.fhmm_exact import create_combined_hmm
# In[ ]:
learnt_model_combined = create_combined_hmm(new_learnt_models)
# In[ ]:
from nilmtk.disaggregate.fhmm_exact import FHMM
# In[16]:
f = FHMM()
# In[17]:
f.model = learnt_model_combined
f.individual = new_learnt_models
# In[18]:
import pickle
# In[19]:
pickle.dump(f, open("../fhmm_model_all_%d.p" % mins, "wb"))
# In[8]:
print("Runtime to export to HDF5 = {:.2f}".format(t2 - t1))
# Now importing the data from HDF5 store
t1 = time.time()
dataset = pecan.Pecan_15min()
dataset.load_hdf5(EXPORT_PATH)
t2 = time.time()
print("Runtime to importing from HDF5 = {:.2f}".format(t2 - t1))
# Doing analysis on Home_10
b = dataset.buildings[9]
train, test = train_test_split(b)
# Initializing CO 1D Disaggregator
disaggregator = FHMM()
train_mains = train.utility.electric.mains[
train.utility.electric.mains.keys()[0]][DISAGG_FEATURE]
# Get appliances data
app = train.utility.electric.appliances
train_appliances = pd.DataFrame({appliance: app[appliance][DISAGG_FEATURE] for appliance in app})
# Train
disaggregator.train(train_mains, train_appliances)
# Disaggregate
disaggregator.disaggregate(test.utility.electric.mains[
test.utility.electric.mains.keys()[0]][DISAGG_FEATURE])
# Metrics
from nilmtk.disaggregate.fhmm_exact import create_combined_hmm # In[ ]: learnt_model_combined = create_combined_hmm(new_learnt_models) # In[ ]: from nilmtk.disaggregate.fhmm_exact import FHMM # In[16]: f = FHMM() # In[17]: f.model = learnt_model_combined f.individual = new_learnt_models # In[18]: import pickle # In[19]:
# Get data of Home_01
building = dataset.buildings[1]
# Preprocessing:
print(' :::::::::: Dividing data into test and train')
lenToUse = len(building.utility.electric.appliances.items()[0][1])
train, test = train_test_split(
building,
train_size=int(float(1) / 2.0 * lenToUse),
test_size=int(float(1) / 2.0 * lenToUse)
) # This splits the dataset in 1/2 for training and 1/2 for testing. Test-training ratio should be set by users
### ---------- Disaggregation with FHMM ---------- #
# --- training model
disaggregator = FHMM()
disaggregator_name = "FHMM"
print(' :::::::::: Training FHMM disaggregator')
t1 = time.time()
disaggregator.train(train, disagg_features=[DISAGG_FEATURE])
t2 = time.time()
print("Runtime to train for {} = {:.2f} seconds".format(
disaggregator_name, t2 - t1))
train_time = t2 - t1
# --- disaggregation
print(' :::::::::: Starting disaggregation')
t1 = time.time()
disaggregator.disaggregate(test)
t2 = time.time()
print("Runtime to disaggregate for {}= {:.2f} seconds".format(
