详解python实现FP-TREE进行关联规则挖掘

__author__ = 'wei' from treebuilder import treebuilder from cpbtreebuilder import cpbtreebuilder from util import util from drawtree import drawtree class fptreemining: def __init__(self,saveimg=False,saveto=None): self.fsetdict={} self.util=util() self.saveimg=saveimg self.saveto=saveto if saveimg and saveto: self.drawtool=drawtree() def mine(self,facts,items,threshold): root=treebuilder(items,facts) print('所有特征项及其频度',root.itemcount) self.drawtool.saveimg(root.tree,self.saveto,'Root.jpg',title=None) self.multipathhandler(None,root,threshold) def gettreebuilder(self,facts,items): root=treebuilder(items=items,facts=facts) return root def getcpbtreebuilder(self,treebuilder,item): cpbtreebuilder=treebuilder.tree.getCPB(treebuilder.itemtable[item]) return cpbtreebuilder #检查是否为单路径的树,如果是返回1,不是返回0,如果空树则返回-1 def checksinglepathtree(self,node,threshold): if node.root and node.isleaf(): return 0 if node.isleaf(): return 1 else: if node.root or node.count>=threshold: ccc=0 #count of child has children for child in node.children: if not child.isleaf(): ccc+=1 notleafchild=child if child.isleaf() and not node.root: return 1 if ccc>1: return ccc elif ccc==1: if notleafchild.count<threshold: return 0 else: return self.checksinglepathtree(notleafchild,threshold) else: return 0 else: return 1 def getsinglepathtree(self,node,threshold,rs): if node.root: for child in node.children: if child.count>=threshold: return self.getsinglepathtree(child,threshold,[]) if node.count>=threshold: rs.append(node) if node.isleaf(): return rs for child in node.children: if child.count>=threshold: return self.getsinglepathtree(child,threshold,rs) return rs def unionhandler(self,item,cpb,threshold): #将树上所有节点与item做并集 for newitem in cpb.items: count=cpb.itemcount[newitem] if count>=threshold: pairs=((newitem,count),) self.getfrequentset(item,pairs) #单路径树求所有组合 def singlepathhandler(self,item,cpb,threshold): #寻找该单元路径上所有的组合,组合最小长度为2 spath=self.getsinglepathtree(cpb.tree,threshold,[]) nodepairs=[(node.item,node.count) for node in spath] for cpairs in self.getcombineoflist(nodepairs): self.getfrequentset(item,cpairs) def multipathhandler(self,orgitem,tb,threshold): itemrank=tb.getrandeditems() #print(itemrank) for newitem in itemrank: if tb.itemcount[newitem]<threshold: continue if orgitem is None or len(orgitem)==0: item=(newitem,) else: item=tuple(list(orgitem)+[newitem]) cpb_facts=tb.tree.getCPB(tb.itemtable[newitem]) cpb=cpbtreebuilder(cpb_facts) cpb_tree=cpb.tree if self.drawtool: title='-'.join(item)+"_cpb tree" filename=title+".jpg" self.drawtool.saveimg(cpb_tree,self.saveto,filename,title=title) if item == ('milk','eggs'): ddd=1 pathcount=self.checksinglepathtree(cpb_tree,threshold) if pathcount>1: self.multipathhandler(item,cpb,threshold) if pathcount==1: self.singlepathhandler(item,cpb,threshold) self.unionhandler(item,cpb,threshold) def getcombineoflist(self,nodepairs): for i in range(2,len(nodepairs)+1): combs=self.util.getcombination(nodepairs,i) for comb in combs: yield comb def getfrequentset(self,item,cpairs): orgitem=list(item) items=tuple(orgitem+[pair[0] for pair in cpairs]) count=min([pair[1] for pair in cpairs]) if items in self.fsetdict: print('repeat frequent items',items,item ) self.fsetdict[items]=count #对挖掘结果排序显示 def showfset(self): rankkeys=sorted(self.fsetdict.keys(),key= lambda k:self.fsetdict[k],reverse=True) for key in rankkeys: print(key,self.fsetdict[key])