Improvements on cmvfts and pwfts for multivariate forecasting

pyFTS/common/flrg.py

@@ -106,5 +106,9 @@ class FLRG(object):
     def __len__(self):
         return len(self.RHS)
 
+    def reset_calculated_values(self):
+        self.midpoint = None
+        self.upper = None
+        self.lower = None
 
 

pyFTS/common/fts.py

@@ -507,8 +507,22 @@ class FTS(object):
         return len(self.flrgs)
 
     def len_total(self):
+        """
+        Total length of the model, adding the number of terms in all rules
+
+        :return:
+        """
         return sum([len(k) for k in self.flrgs])
 
+    def reset_calculated_values(self):
+        """
+        Reset all pre-calculated values on the FLRG's
+
+        :return:
+        """
+
+        for flrg in self.flrgs.keys():
+            self.flrgs[flrg].reset_calculated_values()
 
 
 

pyFTS/models/multivariate/cmvfts.py

@@ -82,11 +82,15 @@ class ClusteredMVFTS(mvfts.MVFTS):
 
         ret = {}
         for var in self.explanatory_variables:
+            if self.target_variable.name != var.name:
+                self.target_variable = var
                 self.cluster.change_target_variable(var)
+                self.model.partitioner = self.cluster
+                self.reset_calculated_values()
+
             ret[var.name] = self.model.forecast(ndata, fuzzyfied=self.pre_fuzzyfy, **kwargs)
 
-        columns = ret.keys()
+        return pd.DataFrame(ret, columns=ret.keys())
-        return pd.DataFrame(ret, columns=columns)
 
     def __str__(self):
         """String representation of the model"""

pyFTS/models/multivariate/common.py

@@ -28,6 +28,12 @@ class MultivariateFuzzySet(Composite.FuzzySet):
         if variable == self.target_variable.name:
             self.centroid = set.centroid
 
+    def set_target_variable(self, variable):
+        #print(self.target_variable, variable)
+        self.target_variable = variable
+        #print(self.centroid,self.sets[variable.name].centroid)
+        self.centroid = self.sets[variable.name].centroid
+
     def membership(self, x):
         mv = []
         for var in self.sets.keys():

pyFTS/models/multivariate/grid.py

@@ -108,4 +108,4 @@ class GridCluster(partitioner.Partitioner):
 
     def change_target_variable(self, variable):
         for fset in self.sets:
-            self.sets[fset].target_variable = variable
+            self.sets[fset].set_target_variable(variable)

pyFTS/models/pwfts.py

@@ -202,18 +202,18 @@ class ProbabilisticWeightedFTS(ifts.IntervalFTS):
             #return self.flrg_lhs_unconditional_probability(flrg)
 
     def flrg_lhs_conditional_probability(self, x, flrg):
-        mv = flrg.get_membership(x, self.sets)
+        mv = flrg.get_membership(x, self.partitioner.sets)
         pb = self.flrg_lhs_unconditional_probability(flrg)
         return mv * pb
 
     def get_midpoint(self, flrg):
         if flrg.get_key() in self.flrgs:
             tmp = self.flrgs[flrg.get_key()]
-            ret = tmp.get_midpoint(self.sets) #sum(np.array([tmp.rhs_unconditional_probability(s) * self.setsDict[s].centroid for s in tmp.RHS]))
+            ret = tmp.get_midpoint(self.partitioner.sets) #sum(np.array([tmp.rhs_unconditional_probability(s) * self.setsDict[s].centroid for s in tmp.RHS]))
         else:
             if len(flrg.LHS) > 0:
                 pi = 1 / len(flrg.LHS)
-                ret = sum(np.array([pi * self.sets[s].centroid for s in flrg.LHS]))
+                ret = sum(np.array([pi * self.partitioner.sets[s].centroid for s in flrg.LHS]))
             else:
                 ret = np.nan
         return ret
@@ -224,19 +224,19 @@ class ProbabilisticWeightedFTS(ifts.IntervalFTS):
             _flrg = self.flrgs[flrg.get_key()]
             cond = []
             for s in _flrg.RHS.keys():
-                _set = self.sets[s]
+                _set = self.partitioner.sets[s]
                 tmp = _flrg.rhs_unconditional_probability(s) * (_set.membership(x) / _set.partition_function(uod=self.get_UoD()))
                 cond.append(tmp)
             ret = sum(np.array(cond))
         else:
             pi = 1 / len(flrg.LHS)
-            ret = sum(np.array([pi * self.sets[s].membership(x) for s in flrg.LHS]))
+            ret = sum(np.array([pi * self.partitioner.sets[s].membership(x) for s in flrg.LHS]))
         return ret
 
     def get_upper(self, flrg):
         if flrg.get_key() in self.flrgs:
             tmp = self.flrgs[flrg.get_key()]
-            ret = tmp.get_upper(self.sets)
+            ret = tmp.get_upper(self.partitioner.sets)
         else:
             ret = 0
         return ret
@@ -244,7 +244,7 @@ class ProbabilisticWeightedFTS(ifts.IntervalFTS):
     def get_lower(self, flrg):
         if flrg.get_key() in self.flrgs:
             tmp = self.flrgs[flrg.get_key()]
-            ret = tmp.get_lower(self.sets)
+            ret = tmp.get_lower(self.partitioner.sets)
         else:
             ret = 0
         return ret
@@ -398,8 +398,8 @@ class ProbabilisticWeightedFTS(ifts.IntervalFTS):
                 for s in flrgs:
                     if s.get_key() in self.flrgs:
                         flrg = self.flrgs[s.get_key()]
-                        pk = flrg.lhs_conditional_probability(sample, self.sets, self.global_frequency_count, uod, nbins)
+                        pk = flrg.lhs_conditional_probability(sample, self.partitioner.sets, self.global_frequency_count, uod, nbins)
-                        wi = flrg.rhs_conditional_probability(bin, self.sets, uod, nbins)
+                        wi = flrg.rhs_conditional_probability(bin, self.partitioner.sets, uod, nbins)
                         num.append(wi * pk)
                         den.append(pk)
                     else:

pyFTS/tests/multivariate.py

@@ -26,13 +26,23 @@ from pyFTS.models.multivariate import variable, cmvfts
 vx = variable.Variable("x", data_label="x", partitioner=Grid.GridPartitioner, npart=15, data=df)
 vy = variable.Variable("y", data_label="y", partitioner=Grid.GridPartitioner, npart=15, data=df)
 
-model = cmvfts.ClusteredMVFTS(pre_fuzzyfy=False, knn=3, fts_method=pwfts.ProbabilisticWeightedFTS)
+model = cmvfts.ClusteredMVFTS(pre_fuzzyfy=False, knn=3, order=2, fts_method=pwfts.ProbabilisticWeightedFTS)
 model.append_variable(vx)
 model.append_variable(vy)
 model.target_variable = vx
 
 model.fit(df.iloc[:800])
 
-df = model.predict(df.iloc[800:], type='multivariate')
+test = df.iloc[800:]
 
-print(df)
+forecasts = model.predict(test, type='multivariate')
+
+fig, ax = plt.subplots(nrows=2, ncols=2, figsize=[15,7])
+ax[0][0].plot(test['x'].values)
+ax[0][0].plot(forecasts['x'].values)
+ax[0][1].scatter(test['x'].values,test['y'].values)
+ax[0][1].scatter(forecasts['x'].values,forecasts['y'].values)
+ax[1][0].scatter(test['y'].values,test['x'].values)
+ax[1][0].scatter(forecasts['y'].values,forecasts['x'].values)
+ax[1][1].plot(test['y'].values)
+ax[1][1].plot(forecasts['y'].values)