挖掘DBLP作者合作关系，FP-Growth算法实践（5）：挖掘研究者合作关系

就是频繁项集挖掘，FP-Growth算法。

先产生headerTable：

数据结构（其实也是调了好几次代码才确定的，因为一开始总有想不到的东西）：entry: entry: {authorName: frequence, firstChildPointer, startYear, endYear}

def CreateHeaderTable(tranDB, minSupport=1):
    headerTable={} #entry: entry: {authorName: frequence, firstChildPointer, startYear, endYear}
    authorDB={} #entry: {frozenset(authorListSet): frequence}
    for i, (conf, year, authorList) in enumerate(tranDB):
        authorListSet=set([])
        print "trans", i, "=="*20
        if conf is np.nan or year is np.nan or authorList is np.nan:
            continue #for tranDB[2426,:]
        for author in authorList.split("|"):
            authorListSet.add(author)
            if headerTable.has_key(author):
                headerTable[author][0]+=1
                if year<headerTable[author][2]:
                    headerTable[author][2]=year
                elif year>headerTable[author][3]:
                    headerTable[author][3]=year
            else:
                headerTable[author]=[1, None, year, year]
        if authorDB.has_key(frozenset(authorListSet)):
            authorDB[frozenset(authorListSet)]+=1
        else:
            authorDB[frozenset(authorListSet)]=1
    for author in headerTable.keys():
        if headerTable[author][0]<minSupport:
            del headerTable[author]
    return headerTable, authorDB

再构建FP-Tree：

每个treeNode又五元组来描述：

class TREE_NODE:
    def __init__(self, authorName, frequence, parentPointer):
        self.authorName=authorName
        self.frequence=frequence
        s