Improvements on generators for multi step ahead forecasting at MVFTS

pyFTS/common/fts.py

@@ -101,7 +101,8 @@ class FTS(object):
         :keyword explain: try to explain, step by step, the one-step-ahead point forecasting result given the input data.
         :keyword generators: for multivariate methods on multi step ahead forecasting, generators is a dict where the keys
                             are the variables names (except the target_variable) and the values are lambda functions that
-                            accept one value (the actual value of the variable) and return the next value.
+                            accept one value (the actual value of the variable) and return the next value or trained FTS
+                            models that accept the actual values and forecast new ones.
 
         :return: a numpy array with the forecasted data
         """

pyFTS/models/multivariate/mvfts.py

@@ -2,6 +2,7 @@ from pyFTS.common import fts, FuzzySet, FLR, Membership
 from pyFTS.partitioners import Grid
 from pyFTS.models.multivariate import FLR as MVFLR, common, flrg as mvflrg
 from itertools import product
+from types import LambdaType
 
 import numpy as np
 import pandas as pd
@@ -52,11 +53,15 @@ class MVFTS(fts.FTS):
     def apply_transformations(self, data, params=None, updateUoD=False, **kwargs):
         ndata = data.copy(deep=True)
         for var in self.explanatory_variables:
+            try:
+                values = ndata[var.data_label].values #if isinstance(ndata, pd.DataFrame) else ndata[var.data_label]
                 if self.uod_clip and var.partitioner.type == 'common':
-                ndata[var.data_label] = np.clip(ndata[var.data_label].values,
+                    ndata[var.data_label] = np.clip(values,
                                                     var.partitioner.min, var.partitioner.max)
 
-            ndata[var.data_label] = var.apply_transformations(ndata[var.data_label].values)
+                ndata[var.data_label] = var.apply_transformations(values)
+            except:
+                pass
 
         return ndata
 
@@ -120,7 +125,7 @@ class MVFTS(fts.FTS):
         ret = []
         ndata = self.apply_transformations(data)
         c = 0
-        for index, row in ndata.iterrows():
+        for index, row in ndata.iterrows() if isinstance(ndata, pd.DataFrame) else enumerate(ndata):
             data_point = self.format_data(row)
             flrs = self.generate_lhs_flrs(data_point)
             mvs = []
@@ -159,7 +164,8 @@ class MVFTS(fts.FTS):
             raise Exception('You must provide parameter \'generators\'! generators is a dict where the keys' +
                             ' are the variables names (except the target_variable) and the values are ' +
                             'lambda functions that accept one value (the actual value of the variable) '
-                            ' and return the next value.')
+                            ' and return the next value or trained FTS models that accept the actual values and '
+                            'forecast new ones.')
 
         ndata = self.apply_transformations(data)
 
@@ -174,13 +180,20 @@ class MVFTS(fts.FTS):
 
             ret.append(tmp)
 
-            last_data_point = sample.loc[sample.index[-1]]
-
             new_data_point = {}
 
             for var in self.explanatory_variables:
                 if var.name != self.target_variable.name:
+                    if isinstance(generators[var.name], LambdaType):
+                        last_data_point = ndata.loc[sample.index[-1]]
                         new_data_point[var.data_label] = generators[var.name](last_data_point[var.data_label])
+                    elif isinstance(generators[var.name], fts.FTS):
+                        model = generators[var.name]
+                        last_data_point = ndata.loc[[sample.index[-model.order]]]
+                        if not model.is_multivariate:
+                            last_data_point = last_data_point[var.data_label].values
+
+                        new_data_point[var.data_label] = model.forecast(last_data_point)[0]
 
             new_data_point[self.target_variable.data_label] = tmp
 

pyFTS/models/multivariate/variable.py

@@ -30,9 +30,13 @@ class Variable:
         self.mask = kwargs.get('mask', None)
         """The mask for format the data column on Pandas Dataframe"""
         self.transformation = kwargs.get('transformation', None)
+        """Pre processing transformation for the variable"""
         self.transformation_params = kwargs.get('transformation_params', None)
         self.partitioner = None
+        """UoD partitioner for the variable data"""
         self.alpha_cut = kwargs.get('alpha_cut', 0.0)
+        """Minimal membership value to be considered on fuzzyfication process"""
+
 
         if kwargs.get('data', None) is not None:
             self.build(**kwargs)

pyFTS/tests/multivariate.py

@@ -128,30 +128,21 @@ vtemp = variable.Variable("Temperature", data_label="temperature", alias='temper
 
 from pyFTS.models.multivariate import mvfts, wmvfts, cmvfts, grid
 
-vars = [vhour, vday, vload]
+from pyFTS.models.multivariate import mvfts, wmvfts, cmvfts, grid
 
-#fs = grid.GridCluster(explanatory_variables=vars, target_variable=vload)
+mtemp = wmvfts.WeightedMVFTS(explanatory_variables=[vhour, vmonth, vtemp], target_variable=vtemp)
+mtemp.fit(train_mv)
 
-#model = mvfts.MVFTS(explanatory_variables=vars, target_variable=vload)
+from pyFTS.models.multivariate import mvfts, wmvfts, cmvfts, grid
-model = wmvfts.WeightedMVFTS(explanatory_variables=vars, target_variable=vload)
-#model = cmvfts.ClusteredMVFTS(explanatory_variables=vars, target_variable=vload,order=2, knn=3, partitioner=fs)
-model.fit(train_mv)
-print(model.shortname)
-Util.persist_obj(model, model.shortname)
-#'''
 
-#model = Util.load_obj('MVFTS')
+mload = wmvfts.WeightedMVFTS(explanatory_variables=[vhour, vday, vtemp, vload], target_variable=vload)
+mload.fit(train_mv)
 
-with open("rules.txt","w") as file:
-    file.write(str(model))
 
-forecasts = model.predict(test_mv.iloc[:100])
+time_generator = lambda x : pd.to_datetime(x) + pd.to_timedelta(1, unit='h')
-forecasts.insert(0,None)
+temp_generator = mtemp
 
-fig, ax = plt.subplots(nrows=1, ncols=1, figsize=[15,3])
-ax.plot(test_mv['load'].values[:100],label='Original')
-ax.plot(forecasts, label='predicted')
-handles, labels = ax.get_legend_handles_labels()
-lgd = ax.legend(handles, labels, loc=2, bbox_to_anchor=(1, 1))
 
-Util.show_and_save_image(fig, model.shortname, True)
+forecasts = mload.predict(test_mv.iloc[:1], steps_ahead=48, generators={'Hour': time_generator,
+                                                                        'DayOfWeek': time_generator,
+                                                                        "Temperature": mtemp})