Added predictProbability method.

Author: Olcay Taner YILDIZ

Classification/Model/DecisionTree/DecisionNode.py

@@ -253,3 +253,12 @@ def predict(self, instance: Instance) -> str:
                 if node.__condition.satisfy(instance):
                     return node.predict(instance)
             return self.__classLabel
+
+    def predictProbabilityDistribution(self, instance: Instance) -> dict:
+        if self.leaf:
+            return self.__data.classDistribution().getProbabilityDistribution()
+        else:
+            for node in self.children:
+                if node.__condition.satisfy(instance):
+                    return node.predictProbabilityDistribution(instance)
+            return self.__data.classDistribution().getProbabilityDistribution()

Classification/Model/DecisionTree/DecisionTree.py

@@ -40,6 +40,9 @@ def predict(self, instance: Instance) -> str:
             predictedClass = instance.getPossibleClassLabels()
         return predictedClass
 
+    def predictProbability(self, instance: Instance) -> dict:
+        return self.__root.predictProbabilityDistribution(instance)
+
     def pruneNode(self, node: DecisionNode, pruneSet: InstanceList):
         """
         The prune method takes a DecisionNode and an InstanceList as inputs. It checks the classification performance

Classification/Model/DummyModel.py

@@ -40,3 +40,6 @@ def predict(self, instance: Instance) -> str:
             return self.distribution.getMaxItemIncludeTheseOnly(possibleClassLabels)
         else:
             return self.distribution.getMaxItem()
+
+    def predictProbability(self, instance: Instance) -> dict:
+        return self.distribution.getProbabilityDistribution()

Classification/Model/KnnModel.py

@@ -54,6 +54,10 @@ def predict(self, instance: Instance) -> str:
             predictedClass = Model.getMaximum(nearestNeighbors.getClassLabels())
         return predictedClass
 
+    def predictProbability(self, instance: Instance) -> dict:
+        nearestNeighbors = self.nearestNeighbors(instance)
+        return nearestNeighbors.classDistribution().getProbabilityDistribution()
+
     def makeComparator(self):
         def compare(instanceA: KnnInstance, instanceB: KnnInstance):
             if instanceA.distance < instanceB.distance:

Classification/Model/Model.py

@@ -24,6 +24,10 @@ def predict(self, instance: Instance) -> str:
         """
         pass
 
+    @abstractmethod
+    def predictProbability(self, instance: Instance) -> dict:
+        pass
+
     @staticmethod
     def getMaximum(classLabels: list) -> str:
         """

Classification/Model/NeuralNetworkModel.py

@@ -225,3 +225,9 @@ def predict(self, instance: Instance) -> str:
             return self.predictWithCompositeInstance(instance.getPossibleClassLabels())
         else:
             return self.classLabels[self.y.maxIndex()]
+
+    def predictProbability(self, instance: Instance) -> dict:
+        result = {}
+        for i in range(len(self.classLabels)):
+            result[self.classLabels[i]] = self.y.getValue(i)
+        return result

Classification/Model/RandomModel.py

@@ -46,3 +46,9 @@ def predict(self, instance: Instance) -> str:
             size = len(self.__classLabels)
             index = random.randrange(size)
             return self.__classLabels[index]
+
+    def predictProbability(self, instance: Instance) -> dict:
+        result = {}
+        for classLabel in self.__classLabels:
+            result[classLabel] = 1.0 / len(self.__classLabels)
+        return result

Classification/Model/TreeEnsembleModel.py

@@ -38,3 +38,9 @@ def predict(self, instance: Instance) -> str:
         for tree in self.__forest:
             distribution.addItem(tree.predict(instance))
         return distribution.getMaxItem()
+
+    def predictProbability(self, instance: Instance) -> dict:
+        distribution = DiscreteDistribution()
+        for tree in self.__forest:
+            distribution.addItem(tree.predict(instance))
+        return distribution.getProbabilityDistribution()

setup.py

@@ -2,7 +2,7 @@
 
 setup(
     name='NlpToolkit-Classification',
-    version='1.0.7',
+    version='1.0.8',
     packages=['Classification', 'Classification.Model', 'Classification.Model.DecisionTree', 'Classification.Filter',
               'Classification.DataSet', 'Classification.Instance', 'Classification.Attribute',
               'Classification.Parameter', 'Classification.Classifier', 'Classification.Experiment',