Python实现FP-growth【链表】【递归】

糖逗

参考书籍：《机器学习实战》

1. import numpy as np
   
2. 
   
3. class treeNode:
   
4.     def __init__(self, nameValue, numOccur, parentNode):
   
5.         self.name = nameValue
   
6.         self.count = numOccur
   
7.         self.nodeLink = None
   
8.         self.parent = parentNode
   
9.         self.children = {}
   
10.         
    
11.     def inc(self, numOccur):#增加当前节点的值
    
12.         self.count += numOccur
    
13.    
    
14.     def disp(self, ind = 1):#打印树，ind设置成0也可以吗？？？？？
    
15.         print('  '*ind, self.name, ' ', self.count)
    
16.         for child in self.children.values():
    
17.             child.disp(ind + 1)#递归
    
18. 
    
19. 
    
20. def createTree(dataSet, minSup = 1):#扫描头指针表，删掉哪些出现次数少于minSup的项
    
21.     headerTable = {}
    
22.     for trans in dataSet:
    
23.         for item in trans:
    
24.             #https://www.runoob.com/python/att-dictionary-get.html
    
25.             headerTable[item] = headerTable.get(item, 0) + dataSet[trans]#统计每个字符出现的次数
    
26.     for k in list(headerTable.keys()):  
    
27.         if headerTable[k] < minSup:
    
28.             del(headerTable[k])
    
29.     freqItemSet = set(headerTable.keys())#通过minSup筛选的频繁集
    
30.     if len(freqItemSet) == 0:#如果每个元素只出现一次，那就直接返回None（没有规律可循）
    
31.         return None, None  
    
32.     for k in headerTable:
    
33.         headerTable[k] = [headerTable[k], None]#None可能存储指针
    
34.     retTree = treeNode('Null Set', 1, None)#根节点设置成空集
    
35.     for tranSet, count in dataSet.items():  
    
36.         localD = {}
    
37.         for item in tranSet:  
    
38.             if item in freqItemSet:#单独出现
    
39.                 localD[item] = headerTable[item][0]
    
40.         if len(localD) > 0:
    
41.             orderedItems = [v[0] for v in sorted(localD.items(),
    
42.                             key = lambda p: p[1], reverse = True)]#降序排列
    
43.             updateTree(orderedItems, retTree, headerTable, count)
    
44.     return retTree, headerTable
    
45. 
    
46. def updateTree(items, inTree, headerTable, count):
    
47.     if items[0] in inTree.children:#如果在节点中，直接更新节点count
    
48.         inTree.children[items[0]].inc(count)
    
49.     else:  
    
50.         inTree.children[items[0]] = treeNode(items[0], count, inTree)#如果没有则新增节点
    
51.         if headerTable[items[0]][1] == None:#当前item不属于频繁集中，更新headerTable
    
52.             headerTable[items[0]][1] = inTree.children[items[0]]
    
53.         else:
    
54.             updateHeader(headerTable[items[0]][1], inTree.children[items[0]])
    
55.     if len(items) > 1:
    
56.         updateTree(items[1:], inTree.children[items[0]], headerTable, count)
    
57.         
    
58. def updateHeader(nodeToTest, targetNode):   
    
59.     while (nodeToTest.nodeLink != None):#移动到链表的最后端
    
60.         nodeToTest = nodeToTest.nodeLink
    
61.     nodeToTest.nodeLink = targetNode#将链表最后段关联当前端        
    
62. 
    
63. def ascendTree(leafNode, prefixPath):#从叶子节点递归至根节点
    
64.     if leafNode.parent != None:
    
65.         prefixPath.append(leafNode.name)
    
66.         ascendTree(leafNode.parent, prefixPath)
    
67.    
    
68. def findPrefixPath(basePat, treeNode): #headerTable中存储的是元素链表的起始指针
    
69.     condPats = {}
    
70.     while treeNode != None:
    
71.         prefixPath = []
    
72.         ascendTree(treeNode, prefixPath)
    
73.         if len(prefixPath) > 1:
    
74.             condPats[frozenset(prefixPath[1:])] = treeNode.count
    
75.         treeNode = treeNode.nodeLink
    
76.     return condPats
    
77. 
    
78. def loadSimpDat():
    
79.     simpDat = [['r', 'z', 'h', 'j', 'p'],
    
80.                ['z', 'y', 'x', 'w', 'v', 'u', 't', 's'],
    
81.                ['z'],
    
82.                ['r', 'x', 'n', 'o', 's'],
    
83.                ['y', 'r', 'x', 'z', 'q', 't', 'p'],
    
84.                ['y', 'z', 'x', 'e', 'q', 's', 't', 'm']]
    
85.     return simpDat
    
86. 
    
87. def createInitSet(dataSet):
    
88.     retDict = {}
    
89.     for trans in dataSet:
    
90.         retDict[frozenset(trans)] = 1
    
91.     return retDict
    
92. 
    
93. def mineTree(inTree, headerTable, minSup, preFix, freqItemList):
    
94.     bigL = [v[0] for v in sorted(headerTable.items(), key = lambda p: p[1][0])]
    
95.     for basePat in bigL:
    
96.         newFreqSet = preFix.copy()
    
97.         newFreqSet.add(basePat)
    
98.         freqItemList.append(newFreqSet)
    
99.         condPattBases = findPrefixPath(basePat, headerTable[basePat][1])#提取前缀路径
    
100.         myCondTree, myHead = createTree(condPattBases, minSup)#根据前缀路径构建条件FP树
     
101.         if myHead != None:        
     
102.             mineTree(myCondTree, myHead, minSup, newFreqSet, freqItemList)#从二元扩展到三元，从三元扩展到四元，以此类推递归
     
103.     return
     
104.             
     
105. if __name__ == "__main__":
     
106.     simpDat = loadSimpDat()
     
107.     initSet = createInitSet(simpDat)
     
108.     myFPtree, myHeaderTab = createTree(initSet, 3)
     
109.     #myFPtree.disp()
     
110.     #res = findPrefixPath('r', myHeaderTab['r'][1])
     
111.     freqItem = []
     
112.     mineTree(myFPtree, myHeaderTab, 3, set(), freqItem)
     
113.     print(freqItem)

复制代码

糖逗

未完待续，明天接着学习。