輸入：

    simpDat = [['r', 'z', 'h', 'j', 'p'],
               ['z', 'y', 'x', 'w', 'v', 'u', 't', 's'],
               ['z'],
               ['r', 'x', 'n', 'o', 's'],
               ['y', 'r', 'x', 'z', 'q', 't', 'p'],
               ['y', 'z', 'x', 'e', 'q', 's', 't', 'm']]

過程：

dataSet=simpDat
freqItems
 
=fpGrowth(dataSet,4)  #4代表最小頻繁度，即要找出出現4次或4次以上的頻繁項集
freqItems

dataSet是輸入，可按實際情況整理

輸出：

[{'x'}, {'z'}]  #simDat中出現4次貨以上的頻繁項集

以下是程式碼，當成一個FPgrowth的黑盒子：

#FP樹中節點的類定義
class treeNode:
    def __init__(self, nameValue, numOccur, parentNode):
        self.name = nameValue
        self.count = numOccur
        self.nodeLink = None
 
 #nodeLink 變數用於連結相似的元素項
        self.parent = parentNode #指向當前節點的父節點
        self.children = {} #空字典，存放節點的子節點

    def inc(self, numOccur):
        self.count += numOccur

#將樹以文字形式顯示
    def disp(self, ind=1):
        print ('  ' * ind, self.name, ' ', self.count)
        for child in self.children.values():
            child.disp(ind + 1
 
)


#構建FP-tree
def createTree(dataSet, minSup=1):
    headerTable = {}
    for trans in dataSet:  #第一次遍歷：統計各個資料的頻繁度
        for item in trans:
            headerTable[item] = headerTable.get(item, 0) + dataSet[trans]
            #用頭指標表統計各個類別的出現的次數，計算頻繁量：頭指標表[類別]=出現次數
    for  k in list(headerTable):  #刪除未達到最小頻繁度的資料
        if headerTable[k] < minSup:
            del (headerTable[k])
    freqItemSet = set(headerTable.keys())#儲存達到要求的資料
    # print ('freqItemSet: ',freqItemSet)
    if len(freqItemSet) == 0:
        return None, None  #若達到要求的數目為0
    for k in headerTable: #遍歷頭指標表
        headerTable[k] = [headerTable[k], None]  #儲存計數值及指向每種型別第一個元素項的指標
    # print ('headerTable: ',headerTable)
    retTree = treeNode('Null Set', 1, None)  #初始化tree
    for tranSet, count in dataSet.items():  # 第二次遍歷：
        localD = {}
        for item in tranSet:  # put transaction items in order
            if item in freqItemSet:#只對頻繁項集進行排序
                localD[item] = headerTable[item][0]

        #使用排序後的頻率項集對樹進行填充
        if len(localD) > 0:
            orderedItems = [v[0] for v in sorted(localD.items(), key=lambda p: p[1], reverse=True)]
            updateTree(orderedItems, retTree, headerTable, count)  # populate tree with ordered freq itemset
    return retTree, headerTable  #返回樹和頭指標表


def updateTree(items, inTree, headerTable, count):
    if items[0] in inTree.children:  # 首先檢查是否存在該節點
        inTree.children[items[0]].inc(count)  # 存在則計數增加
    else:  # 不存在則將新建該節點
        inTree.children[items[0]] = treeNode(items[0], count, inTree)#建立一個新節點
        if headerTable[items[0]][1] == None:  # 若原來不存在該類別，更新頭指標列表
            headerTable[items[0]][1] = inTree.children[items[0]]#更新指向
        else:#更新指向
            updateHeader(headerTable[items[0]][1], inTree.children[items[0]])
    if len(items) > 1:  #仍有未分配完的樹，迭代
        updateTree(items[1::], inTree.children[items[0]], headerTable, count)

#節點連結指向樹中該元素項的每一個例項。
# 從頭指標表的 nodeLink 開始,一直沿著nodeLink直到到達連結串列末尾
def updateHeader(nodeToTest, targetNode):
    while (nodeToTest.nodeLink != None):
        nodeToTest = nodeToTest.nodeLink
    nodeToTest.nodeLink = targetNode


def loadSimpDat():
        simpDat = [['r', 'z', 'h', 'j', 'p'],
                   ['z', 'y', 'x', 'w', 'v', 'u', 't', 's'],
                   ['z'],
                   ['r', 'x', 'n', 'o', 's'],
                   ['y', 'r', 'x', 'z', 'q', 't', 'p'],
                   ['y', 'z', 'x', 'e', 'q', 's', 't', 'm']]
    return simpDat
#createInitSet() 用於實現上述從列表到字典的型別轉換過程
def createInitSet(dataSet):
    retDict = {}
    for trans in dataSet:
        retDict[frozenset(trans)] = 1
    return retDict

#從FP樹中發現頻繁項集
def ascendTree(leafNode, prefixPath):  #遞迴上溯整棵樹
    if leafNode.parent != None:
        prefixPath.append(leafNode.name)
        ascendTree(leafNode.parent, prefixPath)


def findPrefixPath(basePat, treeNode):  #引數：指標，節點；
    condPats = {}
    while treeNode != None:
        prefixPath = []
        ascendTree(treeNode, prefixPath)#尋找當前非空節點的字首
        if len(prefixPath) > 1:
            condPats[frozenset(prefixPath[1:])] = treeNode.count #將條件模式基新增到字典中
        treeNode = treeNode.nodeLink
    return condPats

#遞迴查詢頻繁項集
def mineTree(inTree, headerTable, minSup, preFix, freqItemList):
    # 頭指標表中的元素項按照頻繁度排序,從小到大
    bigL = [v[0] for v in sorted(headerTable.items(), key=lambda p: str(p[1]))]#python3修改
    for basePat in bigL:  #從底層開始
        #加入頻繁項列表
        newFreqSet = preFix.copy()
        newFreqSet.add(basePat)
        print ('finalFrequent Item: ',newFreqSet)
        freqItemList.append(newFreqSet)
        #遞迴呼叫函式來建立基
        condPattBases = findPrefixPath(basePat, headerTable[basePat][1])
        print ('condPattBases :',basePat, condPattBases)

        #2. 構建條件模式Tree
        myCondTree, myHead = createTree(condPattBases, minSup)
        #將建立的條件基作為新的資料集新增到fp-tree
        print ('head from conditional tree: ', myHead)
        if myHead != None: #3. 遞迴
            print ('conditional tree for: ',newFreqSet)
            myCondTree.disp(1)
            mineTree(myCondTree, myHead, minSup, newFreqSet, freqItemList)

def fpGrowth(dataSet, minSup=3):
    initSet = createInitSet(dataSet)
    myFPtree, myHeaderTab = createTree(initSet, minSup)
    freqItems = []
    mineTree(myFPtree, myHeaderTab, minSup, set([]), freqItems)
    return freqItems