Bugfix on FuzzySet.__binary_search for multivariate/seasonal data; Introducing the FuzzySet.transform method

pyFTS/common/Composite.py

@@ -35,7 +35,10 @@ class FuzzySet(FuzzySet.FuzzySet):
         if self.superset:
             return max([s.membership(x) for s in self.sets])
         else:
-            return min([self.mf[ct](x, self.parameters[ct]) for ct in np.arange(0, len(self.mf))])
+            return min([self.mf[ct](self.transform(x), self.parameters[ct]) for ct in np.arange(0, len(self.mf))])
+
+    def transform(self, x):
+        return self.sets[0].transform(x)
 
     def append(self, mf, parameters):
         """

pyFTS/common/FuzzySet.py

@@ -35,6 +35,16 @@ class FuzzySet(object):
             self.upper = parameters[0] + parameters[1]*3
         self.metadata = {}
 
+    def transform(self, x):
+        """
+        Preprocess the data point for non native types
+
+        :param x:
+        :return: return a native type value for the structured type
+        """
+
+        return x
+
     def membership(self, x):
         """
         Calculate the membership value of a given input
@@ -42,7 +52,7 @@ class FuzzySet(object):
         :param x: input value 
         :return: membership value of x at this fuzzy set
         """
-        return self.mf(x, self.parameters) * self.alpha
+        return self.mf(self.transform(x), self.parameters) * self.alpha
 
     def partition_function(self,uod=None, nbins=100):
         """
@@ -83,14 +93,14 @@ def __binary_search(x, fuzzy_sets, ordered_sets):
         fs1 = ordered_sets[midpoint - 1] if midpoint > 0 else ordered_sets[0]
         fs2 = ordered_sets[midpoint + 1] if midpoint < max_len else ordered_sets[max_len]
 
-        if fuzzy_sets[fs1].centroid <= x <= fuzzy_sets[fs2].centroid:
+        if fuzzy_sets[fs1].centroid <= fuzzy_sets[fs].transform(x) <= fuzzy_sets[fs2].centroid:
             return (midpoint-1, midpoint, midpoint+1)
         elif midpoint <= 1:
             return [0]
         elif midpoint >= max_len:
             return [max_len]
         else:
-            if x < fuzzy_sets[fs].centroid:
+            if fuzzy_sets[fs].transform(x) < fuzzy_sets[fs].centroid:
                 last = midpoint - 1
             else:
                 first = midpoint + 1

pyFTS/models/seasonal/common.py

@@ -89,14 +89,15 @@ class FuzzySet(FuzzySet.FuzzySet):
 
     def __init__(self, datepart, name, mf, parameters, centroid, alpha=1.0, **kwargs):
         super(FuzzySet, self).__init__(name, mf, parameters, centroid, alpha,
-                                       type=kwargs.get('type', 'datetime'),
+                                       type=kwargs.get('type', 'seasonal'),
                                        **kwargs)
         self.datepart = datepart
         self.type = 'seasonal'
 
-    def membership(self, x):
+    def transform(self, x):
-        if self.type == 'datetime':
+        if self.type == 'seasonal':
             dp = strip_datepart(x, self.datepart)
         else:
             dp = x
-        return self.mf(dp, self.parameters) * self.alpha
+
+        return dp

pyFTS/tests/multivariate.py

@@ -8,8 +8,6 @@ from pyFTS.data import Malaysia
 
 dataset = Malaysia.get_dataframe()
 
-print(dataset.head())
-
 dataset["date"] = pd.to_datetime(dataset["time"], format='%m/%d/%y %I:%M %p')
 
 from pyFTS.partitioners import Grid, Util as pUtil
@@ -48,3 +46,28 @@ vload = variable.Variable("load", data_label="load", partitioner=Grid.GridPartit
 vtemperature = variable.Variable("temperature", data_label="temperature", partitioner=Grid.GridPartitioner, npart=10, 
                        data=mv_train)
 
+
+variables = {
+    'month': vmonth,
+    'day': vday,
+    'hour': vhour,
+    'temperature': vtemperature,
+    'load': vload
+}
+
+var_list = [k for k in variables.keys()]
+
+models = []
+
+import itertools
+
+for k in [itertools.combinations(var_list, r) for r in range(2,len(var_list))]:
+    for x in k:
+      model = mvfts.MVFTS()
+      print(x)
+      for w in x:
+        model.append_variable(variables[w])
+        model.shortname += ' ' + w
+      model.target_variable = vload
+      model.fit(mv_train)
+      models.append(model)