sam/pyFTS
1
0
Fork 0
Code Issues 3 Pull Requests Actions Packages Projects Releases Wiki Activity

Improvements on IFTS and WIFTS (forecast_interval_ahead)

Browse Source
This commit is contained in:
Petrônio Cândido 2019-06-01 19:11:32 -03:00
parent e10aa1e872
commit 2149b4d041
3 changed files with 48 additions and 23 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
pyFTS/common/fts.py
View File

@ -91,6 +91,12 @@ class FTS(object):
return best
def clip_uod(self, ndata):
if self.uod_clip:
ndata = np.clip(ndata, self.original_min, self.original_max)
return ndata
def predict(self, data, **kwargs):
"""
Forecast using trained model
@ -116,8 +122,7 @@ class FTS(object):
else:
ndata = self.apply_transformations(data)
if self.uod_clip:
ndata = np.clip(ndata, self.original_min, self.original_max)
ndata = self.clip_uod(ndata)
if 'distributed' in kwargs:
distributed = kwargs.pop('distributed')

33
pyFTS/models/ifts.py
View File

@ -52,17 +52,16 @@ class IntervalFTS(hofts.HighOrderFTS):
mb = [fuzzySets[k].membership(data[k]) for k in np.arange(0, len(data))]
return mb
def forecast_interval(self, ndata, **kwargs):
ret = []
l = len(ndata)
if l <= self.order:
if l < self.order:
return ndata
for k in np.arange(self.max_lag, l):
for k in np.arange(self.max_lag, l+1):
sample = ndata[k - self.max_lag: k]
@ -88,6 +87,16 @@ class IntervalFTS(hofts.HighOrderFTS):
return ret
def forecast_ahead_interval(self, data, steps, **kwargs):
ret = [[x, x] for x in data[:self.max_lag]]
for k in np.arange(self.max_lag, self.max_lag + steps):
interval_lower = self.clip_uod(self.forecast_interval([x[0] for x in ret[k - self.max_lag: k]])[0])
interval_upper = self.clip_uod(self.forecast_interval([x[1] for x in ret[k - self.max_lag: k]])[0])
interval = [np.nanmin(interval_lower), np.nanmax(interval_upper)]
ret.append(interval)
return ret[-steps:]
class WeightedIntervalFTS(hofts.WeightedHighOrderFTS):
"""
@ -128,17 +137,15 @@ class WeightedIntervalFTS(hofts.WeightedHighOrderFTS):
mb = [fuzzySets[k].membership(data[k]) for k in np.arange(0, len(data))]
return mb
def forecast_interval(self, ndata, **kwargs):
ret = []
l = len(ndata)
if l <= self.order:
if l < self.order:
return ndata
for k in np.arange(self.max_lag, l):
for k in np.arange(self.max_lag, l+1):
sample = ndata[k - self.max_lag: k]
@ -163,3 +170,15 @@ class WeightedIntervalFTS(hofts.WeightedHighOrderFTS):
ret.append([lo_, up_])
return ret
def forecast_ahead_interval(self, data, steps, **kwargs):
ret = [[x, x] for x in data[:self.max_lag]]
for k in np.arange(self.max_lag, self.max_lag + steps):
interval_lower = self.clip_uod(self.forecast_interval([x[0] for x in ret[k - self.max_lag: k]])[0])
interval_upper = self.clip_uod(self.forecast_interval([x[1] for x in ret[k - self.max_lag: k]])[0])
interval = [np.nanmin(interval_lower), np.nanmax(interval_upper)]
ret.append(interval)
return ret[-steps:]

29
pyFTS/tests/general.py
View File

@ -61,18 +61,19 @@ methods_parameters = [
{'order':2 }
]
for dataset_name, dataset in datasets.items():
bchmk.sliding_window_benchmarks2(dataset, 1000, train=0.8, inc=0.2,
benchmark_models=True,
benchmark_methods=methods,
benchmark_methods_parameters=methods_parameters,
methods=[],
methods_parameters=[],
transformations=[None],
orders=[3],
steps_ahead=[10],
partitions=[None],
type='interval',
#distributed=True, nodes=['192.168.0.110', '192.168.0.107','192.168.0.106'],
file="tmp.db", dataset=dataset_name, tag="experiments")
#for dataset_name, dataset in datasets.items():
bchmk.sliding_window_benchmarks2(TAIEX.get_data()[:5000], 1000, train=0.8, inc=0.2,
benchmark_models=False,
benchmark_methods=methods,
benchmark_methods_parameters=methods_parameters,
methods=[ifts.IntervalFTS, ifts.WeightedIntervalFTS],
methods_parameters=[{},{}],
transformations=[None],
orders=[1,2,3],
steps_ahead=[10],
partitions=[33],
type='interval',
#distributed=True, nodes=['192.168.0.110', '192.168.0.107','192.168.0.106'],
#file="tmp.db", dataset=dataset_name, tag="experiments")
file="tmp.db", dataset='TAIEX', tag="experiments")
#'''