def create_train_model(train_dset, ids, max_states, max_obs, precision):
sshmms = []
print('Creating load PMFs and finding load states...')
print('\tMax partitions per load =', max_states)
pmfs = []
for id in ids:
pmfs.append(
EmpiricalPMF(id, max_obs * precision,
list(train_dset[id].astype(int))))
pmfs[-1].quantize(max_states, ε)
#%
print()
print('Creating compressed SSHMM...')
incro = 1 / precision
sshmm = SuperStateHMM(pmfs, [i for i in frange(0, max_obs + incro, incro)])
print('\tConverting DataFrame in to obs/hidden lists...')
#%
#tempme = deepcopy(train_dset)
train_dset = train_dset.astype(int)
#len(train_dset)
obs_id = list(train_dset)[0]
obs = list(train_dset[obs_id])
hidden = [i for i in train_dset[ids].to_records(index=False)]
sshmm.build(obs, hidden)
sshmms.append(sshmm)
return sshmms
datasets_dir = './datasets/%s.csv'
logs_dir = './logs/%s.log'
models_dir = './models/%s.json'
print()
print('Loading saved model %s from JSON storage (%s)...' %
(modeldb, models_dir % modeldb))
fp = open(models_dir % modeldb, 'r')
jdata = json.load(fp)
fp.close()
folds = len(jdata)
print('\tModel set for %d-fold cross-validation.' % folds)
print('\tLoading JSON data into SSHMM objects...')
sshmms = []
for data in jdata:
sshmm = SuperStateHMM()
sshmm._fromdict(data)
sshmms.append(sshmm)
del jdata
labels = sshmms[0].labels
print('\tModel lables are: ', labels)
print()
print('Testing %s algorithm load disagg...' % algo_name)
acc = Accuracy(len(labels), folds)
test_times = []
indv_tm_sum = 0.0
indv_count = 0
y_noise = 0.0
y_total = 0.0
calc_done = [0, 0]
for (fold, priors, testing) in folds:
del testing
tm_start = time()
print()
print('Creating load PMFs and finding load states...')
print('\tMax partitions per load =', max_states)
pmfs = []
for id in ids:
pmfs.append(EmpiricalPMF(id, max_obs * precision, list(priors[id])))
pmfs[-1].quantize(max_states, ε)
print()
print('Creating compressed SSHMM...')
incro = 1 / precision
sshmm = SuperStateHMM(pmfs, [i for i in frange(0, max_obs + incro, incro)])
print('\tConverting DataFrame in to obs/hidden lists...')
obs_id = list(priors)[0]
obs = list(priors[obs_id])
hidden = [i for i in priors[ids].to_records(index=False)]
sshmm.build(obs, hidden)
sshmms.append(sshmm)
train_times.append((time() - tm_start) / 60)
print()
print('Train Time was', round(sum(train_times), 2), ' min (avg ', round(sum(train_times) / len(train_times), 2), ' min/fold).')
print()
for (fold, priors, testing) in folds:
del testing
tm_start = time()
print()
print('Creating load PMFs and finding load states...')
print('\tMax partitions per load =', max_states)
pmfs = []
for id in ids:
pmfs.append(EmpiricalPMF(id, max_obs * precision, list(priors[id])))
pmfs[-1].quantize(max_states, ε)
print()
print('Creating compressed SSHMM...')
incro = 1 / precision
sshmm = SuperStateHMM(pmfs, [i for i in frange(0, max_obs + incro, incro)])
print('\tConverting DataFrame in to obs/hidden lists...')
obs_id = list(priors)[0]
obs = list(priors[obs_id])
hidden = [i for i in priors[ids].to_records(index=False)]
sshmm.build(obs, hidden)
sshmms.append(sshmm)
train_times.append((time() - tm_start) / 60)
print()
print('Train Time was', round(sum(train_times), 2), ' min (avg ',
round(sum(train_times) / len(train_times), 2), ' min/fold).')
datasets_dir = './datasets/%s.csv'
logs_dir = './logs/%s.log'
models_dir = './models/%s.json'
print()
print('Loading saved model %s from JSON storage (%s)...' % (modeldb, models_dir % modeldb))
fp = open(models_dir % modeldb, 'r')
jdata = json.load(fp)
fp.close()
folds = len(jdata)
print('\tModel set for %d-fold cross-validation.' % folds)
print('\tLoading JSON data into SSHMM objects...')
sshmms = []
for data in jdata:
sshmm = SuperStateHMM()
sshmm._fromdict(data)
sshmms.append(sshmm)
del jdata
labels = sshmms[0].labels
print('\tModel lables are: ', labels)
print()
print('Testing %s algorithm load disagg...' % algo_name)
acc = Accuracy(len(labels), folds)
test_times = []
print()
folds = Folding(dataset_loader(datasets_dir % dataset, labels, precision, denoised), folds)
for (fold, priors, testing) in folds:
del priors