def test_prediction(self):
data_frame = datasets.get('gas_2016_hour')
data_frame_month = data_frame.resample('MS').sum().loc['2016', :]
data_frame_training = data_frame_month.iloc[:-1, :]
data_frame_pred = data_frame_month.iloc[[-1], :]
mvlr = regression.MultiVarLinReg(data_frame=data_frame_month,
dependent_var='313b',
options={'p_max': 0.04})
mvlr.do_analysis()
data_frame_pred_95 = mvlr._predict(
mvlr.fit, data_frame=data_frame_pred)
mvlr.confint = 0.98
data_frame_pred_98 = mvlr._predict(
mvlr.fit, data_frame=data_frame_pred)
self.assertAlmostEqual(
data_frame_pred_95.loc['2016-12-01', 'predicted'], data_frame_pred_98.loc['2016-12-01', 'predicted'])
self.assertTrue(data_frame_pred_98.loc['2016-12-01', 'interval_u']
> data_frame_pred_95.loc['2016-12-01', 'interval_u'])
self.assertTrue(data_frame_pred_98.loc['2016-12-01', 'interval_l']
< data_frame_pred_95.loc['2016-12-01', 'interval_l'])
# check limitation to zero
mvlr.allow_negative_predictions = False
mvlr.add_prediction()
self.assertTrue(mvlr.data_frame['predicted'].min() >= 0)
def test_standby(self):
df = datasets.get('elec_power_min_1sensor')
res = analysis.standby(df, 'D')
self.assertEqual(res.index.tz.zone, 'Europe/Brussels')
self.assertRaises(exceptions.EmptyDataFrame, analysis.standby,
pd.DataFrame)
def writeInputFileLists(sample_name, njobs, datadir, outdir):
dspath = datasets.get(sample_name)
if dspath is None:
print("Unknown sample:", sample_name)
print("Registered samples are:", list(datasets.keys()))
dir_reco = os.path.join(datadir, dspath) + 'tt.root'
dir_truth = os.path.join(datadir, dspath) + 'tt_truth.root'
dir_PL = os.path.join(datadir, dspath) + 'tt_PL.root'
dir_sumw = os.path.join(datadir, dspath) + 'sumWeights.root'
files_reco = [
os.path.join(dir_reco, fp) for fp in sorted(os.listdir(dir_reco))
]
files_truth = [
os.path.join(dir_truth, fp) for fp in sorted(os.listdir(dir_truth))
]
files_PL = [os.path.join(dir_PL, fp) for fp in sorted(os.listdir(dir_PL))]
files_sumw = [
os.path.join(dir_sumw, fp) for fp in sorted(os.listdir(dir_sumw))
]
lists_dir = os.path.join(outdir, 'input_lists')
if not os.path.isdir(lists_dir):
print("Create directory", lists_dir)
os.makedirs(lists_dir)
inlist_reco = os.path.join(lists_dir,
'input_' + sample_name + '_reco_{}.txt')
inlist_truth = os.path.join(lists_dir,
'input_' + sample_name + '_truth_{}.txt')
inlist_PL = os.path.join(lists_dir, 'input_' + sample_name + '_PL_{}.txt')
inlist_sumw = os.path.join(lists_dir,
'input_' + sample_name + '_sumw_{}.txt')
nfiles = len(files_reco)
nfilesPerJob = int(nfiles / njobs)
for j in range(njobs):
istart = j * nfilesPerJob
iend = istart + nfilesPerJob if j < njobs - 1 else None
f_list_reco = open(inlist_reco.format(j), 'w')
f_list_reco.write('\n'.join(files_reco[istart:iend]))
f_list_reco.close()
f_list_truth = open(inlist_truth.format(j), 'w')
f_list_truth.write('\n'.join(files_truth[istart:iend]))
f_list_truth.close()
f_list_PL = open(inlist_PL.format(j), 'w')
f_list_PL.write('\n'.join(files_PL[istart:iend]))
f_list_PL.close()
f_list_sumw = open(inlist_sumw.format(j), 'w')
f_list_sumw.write('\n'.join(files_sumw[istart:iend]))
f_list_sumw.close()
return inlist_reco, inlist_truth, inlist_PL, inlist_sumw
def test_init(self):
data_frame = datasets.get('gas_2016_hour')
data_frame_month = data_frame.resample('MS').sum()
mvlr = regression.MultiVarLinReg(data_frame=data_frame_month,
dependent_var='313b',
options={'p_max': 0.04})
mvlr.do_analysis()
self.assertTrue(hasattr(mvlr, 'list_of_fits'))
def test_strange_names(self):
data_frame = datasets.get('gas_2016_hour')
data_frame_month = data_frame.resample('MS').sum()
data_frame_month.rename(columns={'d5a7': '3*tempête !'}, inplace=True)
mvlr = regression.MultiVarLinReg(data_frame=data_frame_month,
dependent_var='313b',
options={'p_max': 0.04})
mvlr.do_analysis()
self.assertTrue(hasattr(mvlr, 'list_of_fits'))
def test_plot(self):
data_frame = datasets.get('gas_2016_hour')
data_frame_month = data_frame.resample('MS').sum()
mvlr = regression.MultiVarLinReg(data_frame=data_frame_month,
dependent_var='313b',
options={'p_max': 0.04})
mvlr.do_analysis()
with mock.patch.object(plt_mocked, 'subplots', return_value=(fig_mock, ax_mock)):
mvlr.plot()
def test_predict(self):
data_frame = datasets.get('gas_2016_hour')
data_frame_month = data_frame.resample('MS').sum()
data_frame_month.rename(columns={'d5a7': '3*tempête !'}, inplace=True)
mvlr = regression.MultiVarLinReg(data_frame=data_frame_month,
dependent_var='313b',
options={'p_max': 0.04})
mvlr.do_analysis()
mvlr.add_prediction()
self.assertListEqual(mvlr.data_frame.columns.tolist(),
data_frame_month.columns.tolist() + ['predicted', 'interval_l', 'interval_u'])
def test_raises(self):
data_frame = datasets.get('gas_2016_hour')
data_frame_month = data_frame.resample('MS').sum()
mvlr = regression.MultiVarLinReg(data_frame=data_frame_month,
dependent_var='313b',
options={'p_max': 0.04})
self.assertRaises(UnboundLocalError, mvlr.add_prediction)
try:
x = mvlr.list_of_fits
self.assertTrue(False)
except UnboundLocalError:
self.assertTrue(True)
def test_alternative_metrics(self):
data_frame = datasets.get('gas_2016_hour')
data_frame_month = data_frame.resample('MS').sum()
mvlr = regression.MultiVarLinReg(data_frame=data_frame_month,
dependent_var='313b',
options={'p_max': 0.04})
mvlr.do_analysis()
best_rsquared = mvlr.find_best_rsquared(mvlr.list_of_fits)
best_akaike = mvlr.find_best_akaike(mvlr.list_of_fits)
best_bic = mvlr.find_best_bic(mvlr.list_of_fits)
self.assertEqual(best_rsquared, best_akaike)
self.assertEqual(best_rsquared, best_bic)
def test_load_factor(self):
ts = datasets.get('electricity_2016_hour')
ts = ts['e1de'].truncate(after=pd.Timestamp('20160107'))
lf1 = analysis.calculate_load_factor(time_series=ts)
self.assertIsInstance(ts, pd.Series)
self.assertAlmostEqual(ts.iloc[0], (lf1 * ts.max()).iloc[0])
lf2 = analysis.calculate_load_factor(time_series=ts,
resolution='3h',
norm=800)
self.assertIsInstance(ts, pd.Series)
self.assertAlmostEqual(175.0345212009457, (lf2 * 800).iloc[0])
def test_standby_with_time_window(self):
df = datasets.get('elec_power_min_1sensor')
res = analysis.standby(df, 'D', time_window=('01:00', '06:00'))
self.assertEqual(res.index.tz.zone, 'Europe/Brussels')
self.assertEqual(
res.squeeze().to_json(),
'{"1507327200000":61.739999936,"1507413600000":214.9799999222,"1507500000000":53.0399997951,"1507586400000":55.7399999164,"1507672800000":59.94000006,"1507759200000":69.4800002407,"1507845600000":56.8200000236,"1507932000000":54.1799997864,"1508018400000":54.779999801,"1508104800000":54.7199997772,"1508191200000":98.5199999576,"1508277600000":55.6799999066,"1508364000000":53.9399997052,"1508450400000":109.5599999931,"1508536800000":144.3600001093,"1508623200000":52.7999997279}'
)
res = analysis.standby(df, 'D', time_window=('22:00', '06:00'))
self.assertEqual(res.index.tz.zone, 'Europe/Brussels')
self.assertEqual(
res.squeeze().to_json(),
'{"1507327200000":61.739999936,"1507413600000":119.2800000636,"1507500000000":53.0399997951,"1507586400000":55.7399999164,"1507672800000":59.94000006,"1507759200000":69.4800002407,"1507845600000":56.8200000236,"1507932000000":54.1799997864,"1508018400000":54.779999801,"1508104800000":54.7199997772,"1508191200000":98.5199999576,"1508277600000":55.6799999066,"1508364000000":53.9399997052,"1508450400000":96.3000000408,"1508536800000":133.9200000744,"1508623200000":52.7999997279}'
)
def test_pickle_round_trip(self):
"Pickle, unpickle and check results"
data_frame = datasets.get('gas_2016_hour')
data_frame_month = data_frame.resample('MS').sum().loc['2016', :]
data_frame_training = data_frame_month.iloc[:-1, :]
data_frame_pred = data_frame_month.iloc[[-1], :]
mvlr = regression.MultiVarLinReg(data_frame=data_frame_month,
dependent_var='313b',
options={'p_max': 0.04})
mvlr.do_analysis()
data_frame_pred_95_orig = mvlr._predict(
mvlr.fit, data_frame=data_frame_pred)
s = pickle.dumps(mvlr)
m = pickle.loads(s)
self.assertTrue(hasattr(m, 'list_of_fits'))
data_frame_pred_95_roundtrip = m._predict(
m.fit, data_frame=data_frame_pred)
self.assertAlmostEqual(
data_frame_pred_95_orig.loc['2016-12-01', 'predicted'], data_frame_pred_95_roundtrip.loc['2016-12-01', 'predicted'])
def test_prune(self):
"Create overfitted model and prune it"
data_frame = datasets.get('gas_2016_hour')
data_frame_month = data_frame.resample('MS').sum()
mvlr = regression.MultiVarLinReg(data_frame=data_frame_month,
dependent_var='313b')
mvlr.do_analysis()
self.assertTrue("ba14" in mvlr.fit.model.exog_names)
pruned = mvlr._prune(mvlr.fit, 0.05)
self.assertTrue("ba14" in pruned.model.exog_names)
# with this value, both x will be removed, which is a bit counter-intuitive because initially only ba14 has a pvalue > p_max.
pruned = mvlr._prune(mvlr.fit, 0.00009)
self.assertFalse("ba14" in pruned.model.exog_names)
self.assertFalse("d5a7" in pruned.model.exog_names)
mvlr = regression.MultiVarLinReg(data_frame=data_frame_month,
dependent_var='313b',
options={'p_max': 0.00009})
mvlr.do_analysis()
self.assertFalse("ba14" in mvlr.fit.model.exog_names)
self.assertFalse("d5a7" in mvlr.fit.model.exog_names)
def test_count_peaks(self):
df = datasets.get('gas_dec2016_min')
ts = df['313b'].head(100)
count = analysis.count_peaks(ts)
self.assertEqual(count, 13)
