def finallyResult(currentList,simpDat):
data= fpTree.createInitSet(simpDat)
myFPtree,myHeaderTab = fpTree.createTree(data,2)
freqItems = []
fpTree.mineTree(myFPtree,myHeaderTab,2,set([]),freqItems)
result = []
temp = igemRecomdData.getResult(currentList,freqItems)
for item in temp:
t = list(item)
if (len(t) > 0):
for item2 in t:
if (item2 not in result):
result.append(item2)
return result
def ibm_fptree(min_support, min_confidence):
path_ibm = './dataset/IBM-Quest-Data-Generator.exe/ttt.data.txt'
member_group = DataReader.readDataIBM(path_ibm)
result_df = pd.DataFrame(member_group)
result_df_T = result_df.T
L1 = apriori.createL1(result_df_T, min_support)
for i in range(len(L1)):
if str(L1[0][i]) == 'None':
L1 = L1.drop([i], axis=0)
L1 = L1.reset_index(drop=True)
l1_sort = L1.sort_values(['count'], ascending=False)
l1_sort = l1_sort.reset_index(drop=True)
# member_group = (np.array(result_df)).tolist()
new_member_group = []
for i in range(len(member_group)):
member = []
for j in range(len(l1_sort)):
if l1_sort[0][j] in member_group[i]:
member.append(l1_sort[0][j])
new_member_group.append(member)
# 建立FP tree
fp_tree_root = fpTree.createTree(new_member_group)
# fp growth
freq_pats = fpTree.fpGrowth(l1_sort, fp_tree_root, min_support)
freq_itemset = []
freq_itemset_count = []
for i in range(len(freq_pats)):
if (freq_pats[i].val in freq_itemset) == False:
freq_itemset.append(freq_pats[i].val)
freq_itemset_count.append(freq_pats[i].count)
df_itemset = pd.DataFrame(freq_itemset)
df_itemset_count = pd.DataFrame(freq_itemset_count, columns=['count'])
df_ans = pd.concat([df_itemset_count, df_itemset], axis=1)
print('freq itemsets:')
print(df_ans)
generateRule.generateRule(freq_itemset, freq_itemset_count, L1,
min_confidence)
def kaggle_fptree(min_support, min_confidence):
path_kaggle = './dataset/BreadBasket_DMS.csv'
member_group = DataReader.readDataKaggle(path_kaggle)
result_df = pd.DataFrame(member_group)
result_df_T = result_df.T
L1 = apriori.createL1(result_df_T, min_support)
l1_sort = L1.sort_values(['count'], ascending=False)
l1_sort = l1_sort.reset_index(drop=True)
# member_group = (np.array(result_df)).tolist()
new_member_group = []
for i in range(len(member_group)):
member = []
for j in range(len(l1_sort)):
if l1_sort[0][j] in member_group[i]:
member.append(l1_sort[0][j])
new_member_group.append(member)
# 建立FP tree
fp_tree_root = fpTree.createTree(new_member_group)
# fp growth
freq_pats = fpTree.fpGrowth(l1_sort, fp_tree_root, min_support)
freq_itemset = []
freq_itemset_count = []
for i in range(len(freq_pats)):
if (freq_pats[i].val in freq_itemset) == False:
freq_itemset.append(freq_pats[i].val)
freq_itemset_count.append(freq_pats[i].count)
df_itemset = pd.DataFrame(freq_itemset)
df_itemset_count = pd.DataFrame(freq_itemset_count, columns=['count'])
df_ans = pd.concat([df_itemset_count, df_itemset], axis=1)
print('freq itemsets:')
print(df_ans)
generateRule.generateRule(freq_itemset, freq_itemset_count, L1,
min_confidence)
# print str(rule.cause) + '->' + str(rule.effect) + '(support:' + str(rule.support) + ' confidence:' + str(
# rule.confidence) + ')'
counter+=1
return counter
if __name__ == "__main__":
total_data = [[0, 1, 2, 3], [0, 1, 2], [2, 3], [0, 1, 4]]
# 最小支持度
minSup = 200
test_data = ge_association_data(10000)
# 初始集合格式化
initSet = fpTree.createInitSet(test_data)
# 构建FP树
myFPtree, myHeaderTab = fpTree.createTree(initSet, minSup)
# 创建空列表，保存频繁项集
myFreqList = []
fpTree.mineTree(myFPtree, myHeaderTab, minSup, set([]), myFreqList)
print "频繁项集个数：", len(myFreqList)
print "频繁项集：", myFreqList
rules = generate_rules(myFreqList,initSet)
rule_nums = []
ratios = np.linspace(0.2, 0.6, 30)
for ratio in ratios:
print ratio
rule_num = print_rules(rules,0.02,ratio)
rule_nums.append(rule_num)
print rule_num
plt.plot(rule_nums)
def test_fptree(min_support, min_confidence):
path = './dataset/Test.csv'
df = pd.read_csv(path)
# print(df)
# 將每個組合分開
member = []
member_group = []
# group_id=[]
for i in range(len(df)):
if i == 0:
# member.append(df['tid'][i])
member.append(df['Item'][i])
elif df['TID'][i] == df['TID'][i - 1]:
member.append((df['Item'][i]))
else:
member_group.append(list(member))
# group_id.append(df['t_id'][i-1])
member = []
# member.append(df['tid'][i])
member.append(df['Item'][i])
member_group.append(list(member))
# member_group = DataReader.readDataKaggle(path_kaggle)
result_df = pd.DataFrame(member_group)
result_df_T = result_df.T
L1 = apriori.createL1(result_df_T, min_support)
l1_sort = L1.sort_values(['count'], ascending=False)
l1_sort = l1_sort.reset_index(drop=True)
# member_group = (np.array(result_df)).tolist()
new_member_group = []
for i in range(len(member_group)):
member = []
for j in range(len(l1_sort)):
if l1_sort[0][j] in member_group[i]:
member.append(l1_sort[0][j])
new_member_group.append(member)
# 建立FP tree
fp_tree_root = fpTree.createTree(new_member_group)
# fp growth
freq_pats = fpTree.fpGrowth(l1_sort, fp_tree_root, min_support)
freq_itemset = []
freq_itemset_count = []
for i in range(len(freq_pats)):
if (freq_pats[i].val in freq_itemset) == False:
freq_itemset.append(freq_pats[i].val)
freq_itemset_count.append(freq_pats[i].count)
df_itemset = pd.DataFrame(freq_itemset)
df_itemset_count = pd.DataFrame(freq_itemset_count, columns=['count'])
df_ans = pd.concat([df_itemset_count, df_itemset], axis=1)
print('freq itemsets:')
print(df_ans)
generateRule.generateRule(freq_itemset, freq_itemset_count, L1,
min_confidence)