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

bugfix in Measures.get_point_statistics for supporting multivariate models

Browse Source
This commit is contained in:
Petrônio Cândido 2018-06-29 12:19:56 -03:00
parent d15814b545
commit 393388f722
2 changed files with 40 additions and 11 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
pyFTS/benchmarks/Measures.py
View File

@ -324,7 +324,9 @@ def get_point_statistics(data, model, **kwargs):
if steps_ahead == 1: if steps_ahead == 1:
forecasts = model.predict(ndata, **kwargs) forecasts = model.predict(ndata, **kwargs)
if model.is_multivariate: if model.is_multivariate and model.has_seasonality:
ndata = model.indexer.get_data(ndata)
elif model.is_multivariate:
ndata = ndata[model.target_variable.data_label].values ndata = ndata[model.target_variable.data_label].values
if not isinstance(forecasts, (list, np.ndarray)): if not isinstance(forecasts, (list, np.ndarray)):

45
pyFTS/models/seasonal/SeasonalIndexer.py
View File

@ -31,7 +31,15 @@ class SeasonalIndexer(object):
class LinearSeasonalIndexer(SeasonalIndexer): class LinearSeasonalIndexer(SeasonalIndexer):
"""Use the data array/list position to index the seasonality """
def __init__(self, seasons, units, ignore=None, **kwargs): def __init__(self, seasons, units, ignore=None, **kwargs):
"""
Indexer for array/list position
:param seasons: A list with the season group (i.e: 7 for week, 30 for month, etc)
:param units: A list with the units used for each season group, the default is 1 for each
:param ignore:
:param kwargs:
"""
super(LinearSeasonalIndexer, self).__init__(len(seasons), **kwargs) super(LinearSeasonalIndexer, self).__init__(len(seasons), **kwargs)
self.seasons = seasons self.seasons = seasons
self.units = units self.units = units
@ -81,11 +89,19 @@ class LinearSeasonalIndexer(SeasonalIndexer):
class DataFrameSeasonalIndexer(SeasonalIndexer): class DataFrameSeasonalIndexer(SeasonalIndexer):
def __init__(self,index_fields,index_seasons, data_fields,**kwargs): """Use the Pandas.DataFrame index position to index the seasonality """
def __init__(self,index_fields,index_seasons, data_field,**kwargs):
"""
:param index_fields: DataFrame field to use as index
:param index_seasons: A list with the season group, i. e., multiples of positions that are considered a season (i.e: 7 for week, 30 for month, etc)
:param data_fields: DataFrame field to use as data
:param kwargs:
"""
super(DataFrameSeasonalIndexer, self).__init__(len(index_seasons), **kwargs) super(DataFrameSeasonalIndexer, self).__init__(len(index_seasons), **kwargs)
self.fields = index_fields self.fields = index_fields
self.seasons = index_seasons self.seasons = index_seasons
self.data_fields = data_fields self.data_field = data_field
def get_season_of_data(self,data): def get_season_of_data(self,data):
#data = data.copy() #data = data.copy()
@ -111,25 +127,34 @@ class DataFrameSeasonalIndexer(SeasonalIndexer):
data = data[data[f]== season[c]] data = data[data[f]== season[c]]
else: else:
data = data[(data[f] // self.seasons[c]) == season[c]] data = data[(data[f] // self.seasons[c]) == season[c]]
return data[self.data_fields] return data[self.data_field]
def get_index_by_season(self, indexes): def get_index_by_season(self, indexes):
raise Exception("Operation not available!") raise Exception("Operation not available!")
def get_data(self, data): def get_data(self, data):
return data[self.data_fields].tolist() return data[self.data_field].tolist()
def set_data(self, data, value): def set_data(self, data, value):
data.loc[:,self.data_fields] = value data.loc[:,self.data_field] = value
return data return data
class DateTimeSeasonalIndexer(SeasonalIndexer): class DateTimeSeasonalIndexer(SeasonalIndexer):
def __init__(self,date_field, index_fields, index_seasons, data_fields,**kwargs): """Use a Pandas.DataFrame date field to index the seasonality """
def __init__(self,date_field, index_fields, index_seasons, data_field,**kwargs):
"""
:param date_field: DataFrame field that contains the datetime field used on index
:param index_fields: List with commom.DataTime fields
:param index_seasons: Multiples of index_fields, the default is 1
:param data_field: DataFrame field with the time series data
:param kwargs:
"""
super(DateTimeSeasonalIndexer, self).__init__(len(index_seasons), **kwargs) super(DateTimeSeasonalIndexer, self).__init__(len(index_seasons), **kwargs)
self.fields = index_fields self.fields = index_fields
self.seasons = index_seasons self.seasons = index_seasons
self.data_fields = data_fields self.data_field = data_field
self.date_field = date_field self.date_field = date_field
def get_season_of_data(self, data): def get_season_of_data(self, data):
@ -151,7 +176,7 @@ class DateTimeSeasonalIndexer(SeasonalIndexer):
date = data[self.date_field] date = data[self.date_field]
season = [] season = []
for c, f in enumerate(self.fields, start=0): for c, f in enumerate(self.fields, start=0):
season.append(self.strip_datepart(date, f, self.seasons[c])) season.append(common.strip_datepart(date, f, self.seasons[c]))
ret.append(season) ret.append(season)
return ret return ret
@ -166,10 +191,12 @@ class DateTimeSeasonalIndexer(SeasonalIndexer):
raise Exception("Operation not available!") raise Exception("Operation not available!")
def get_data(self, data): def get_data(self, data):
return data[self.data_fields].tolist() return data[self.data_field].tolist()
def get_index(self, data): def get_index(self, data):
return data[self.date_field].tolist() if isinstance(data, pd.DataFrame) else data[self.date_field] return data[self.date_field].tolist() if isinstance(data, pd.DataFrame) else data[self.date_field]
def set_data(self, data, value): def set_data(self, data, value):
raise Exception("Operation not available!") raise Exception("Operation not available!")