Source Code for Module rdkit.ML.KNN.KNNClassificationModel

 1  # $Id: KNNClassificationModel.py 997 2009-02-25 06:12:43Z glandrum $ 
 2  # 
 3  #  Copyright (C) 2003 Rational Discovery LLC 
 4  #      All Rights Reserved 
 5  # 
 6   
 7  """ Define the class _KNNClassificationModel_, used to represent a k-nearest neighbhors classification model 
 8   
 9      Inherits from _KNNModel_ 
10  """ 
11   
12  from rdkit.ML.KNN import KNNModel 
13  from rdkit.ML.KNN import DistFunctions 
14  import math 
15   
16   
17 -class KNNClassificationModel(KNNModel.KNNModel) : 
18    """ This is used to represent a k-nearest neighbor classifier 
19   
20    """ 
21   
22 -  def __init__(self, k, attrs, dfunc) : 
23      self._setup(k, attrs, dfunc) 
24           
25      self._badExamples = [] # list of examples incorrectly classified 
26   
27 -  def type(self): 
28      return "Classification Model" 
29   
30 -  def SetBadExamples(self, examples) : 
31      self._badExamples = examples 
32 -  def GetBadExamples(self) : 
33      return self._badExamples 
34   
35 -  def NameModel(self, varNames) : 
36      self.SetName(self.type()) 
37   
38 -  def ClassifyExample(self, example, appendExamples=0) : 
39      """ Classify a an example by looking at its closest neighbors 
40   
41      The class assigned to this example is same as the class for the mojority of its 
42      _k neighbors 
43   
44      **Arguments** 
45   
46      - examples: the example to be classified 
47   
48      - appendExamples: if this is nonzero then the example will be stored on this model 
49   
50      **Returns** 
51   
52        - the classification of _example_ 
53      """ 
54      if appendExamples: 
55        self._examples.append(example) 
56         
57      # first find the k-closest examples in the traning set 
58      knnLst = self.GetNeighbors(example) 
59   
60      # find out how many of the neighbors belong to each of the classes 
61      clsCnt = {} 
62      for knn in knnLst : 
63        cls = knn[1][-1] 
64        if (clsCnt.has_key(cls)) : 
65          clsCnt[cls] += 1 
66        else : 
67          clsCnt[cls] = 1 
68   
69      # now return the class with the maximum count 
70      mkey = -1 
71      mcnt = -1 
72      for key in clsCnt.keys() : 
73        if (mcnt < clsCnt[key]) : 
74          mkey = key 
75          mcnt = clsCnt[key] 
76   
77      return mkey 
78