apyori apriori 损坏的项目集输出
Posted
技术标签:
【中文标题】apyori apriori 损坏的项目集输出【英文标题】:apyori apriori broken itemset output 【发布时间】:2020-06-08 19:00:21 【问题描述】:我正在尝试使用 apyori 模块运行关联规则。 我的“项目”是各种手术(行 = 患者病例),正如您在下面的数据框示例中看到的那样。 Apyori 未能捕捉到正确的标签,它似乎正在用字母切割标签。我过去从未见过这样的行为。除非我遗漏了什么,否则我的数据集已正确格式化以供 apyori 使用。任何时候进行的手术不超过 2 次。
这是我得到的一个例子:
RelationRecord(items=frozenset('v', '_'), support=0.10309278350515463, ordered_statistics=[OrderedStatistic(items_base=frozenset(), items_add=frozenset('v', '_'), confidence=0.10309278350515463, lift=1.0), OrderedStatistic(items_base=frozenset('_'), items_add=frozenset('v'), confidence=0.10638297872340426, lift=1.0319148936170213), OrderedStatistic(items_base=frozenset('v'), items_add=frozenset('_'), confidence=1.0, lift=1.0319148936170213)]) Support: 0.10309278350515463 Confidence: frozenset('v', '_') Lift:
0.10309278350515463
frozenset 坏了... 这是我的输入 dataframe.head():
sm-to-sm_bowel_anastom small_bowel_incision_nec sm_bowel_exteriorization \
0 0 0 0
1 0 0 0
2 0 0 0
3 0 0 0
4 0 0 0
5 0 0 0
6 0 0 0
7 0 0 0
8 0 0 0
9 0 0 0
incisional_hernia_repair colonoscopy anal_anastomosis \
0 0 0 0
1 0 0 0
2 0 0 0
3 0 0 0
4 0 0 0
5 0 0 0
6 0 0 0
7 0 0 0
8 0 0 0
9 0 0 0
c.a.t._scan_of_abdomen open_sigmoidectomy_nec small_bowel_suture_nec \
0 0 0 0
1 0 0 0
2 0 0 0
3 0 0 0
4 0 0 0
5 0 0 0
6 0 0 0
7 0 0 0
8 0 0 0
9 0 0 0
lap_pt_ex_lrg_intest_nec ... abdperneal_res_rectm_nos \
0 0 ... 0
1 0 ... 0
2 0 ... 0
3 0 ... 0
4 0 ... 0
5 0 ... 0
6 0 ... 0
7 0 ... 0
8 0 ... 0
9 0 ... 0
ureteral_catheterization cv_cath_plcmt_w_guidance \
0 0 0
1 0 0
2 0 0
3 0 0
4 0 0
5 0 0
6 0 0
7 0 0
8 0 0
9 0 0
clos_large_bowel_biopsy lap_right_hemicolectomy continent_ileostomy \
0 0 0 0
1 0 0 0
2 0 0 0
3 0 0 0
4 0 0 0
5 0 0 0
6 0 0 0
7 0 0 0
8 0 0 0
9 0 0 1
insert_endotracheal_tube mult_seg_sm_bowel_excis \
0 0 0
1 0 0
2 0 0
3 0 0
4 0 0
5 0 0
6 0 0
7 0 0
8 0 0
9 0 0
small-to-large_bowel_nec opn_lft_hemicolectmy_nec
0 1 1
1 0 0
2 0 0
3 0 0
4 0 0
5 0 0
6 1 0
7 0 0
8 0 0
9 0 0
[10 rows x 97 columns]
我这样运行规则:
from apyori import apriori as ap
rulez = ap(ohe_df, min_support = 0.1, min_length = 2,use_colnames=True)
我只有 2 次手术同时进行,所以我不希望有 >2 项的组合。
frozenset 发生了什么?
谢谢
【问题讨论】:
【参考方案1】:您需要将输入数据放在列表列表中,其中每个列表是一对组合在一起的事物。我编了一些数据:
# Replace 1's with the column name
df = df.replace(1, pd.Series(df.columns, df.columns))
# get a list of non-zero values per row into an array of lists
ops = df.apply(lambda x: [v for v in x.values if v!=0], axis=1).values
ops 变量现在看起来不错:
array([list(['small_bowel_incision_nec', 'colonoscopy']),
list(['sm_bowel_exteriorization', 'colonoscopy']),
list(['sm-to-sm_bowel_anastom', 'small_bowel_suture_nec']),
list(['small_bowel_incision_nec', 'colonoscopy']),
list(['anal_anastomosis', 'open_sigmoidectomy_nec']),
list(['colonoscopy', 'c.a.t._scan_of_abdomen']),
list(['sm-to-sm_bowel_anastom', 'open_sigmoidectomy_nec']),
list(['c.a.t._scan_of_abdomen', 'small_bowel_suture_nec']),
list(['incisional_hernia_repair', 'small_bowel_suture_nec']),
list(['small_bowel_incision_nec', 'colonoscopy'])], dtype=object)
# Run apriori, getting them as a list
rulez = list(ap(ops, min_support = 0.1, min_length = 2,use_colnames=True))
样本输出
[RelationRecord(items=frozenset('anal_anastomosis'), support=0.1, ordered_statistics=[OrderedStatistic(items_base=frozenset(), items_add=frozenset('anal_anastomosis'), confidence=0.1, lift=1.0)]),
RelationRecord(items=frozenset('c.a.t._scan_of_abdomen'), support=0.2, ordered_statistics=[OrderedStatistic(items_base=frozenset(), items_add=frozenset('c.a.t._scan_of_abdomen'), confidence=0.2, lift=1.0)]),
RelationRecord(items=frozenset('colonoscopy'), support=0.5, ordered_statistics=[OrderedStatistic(items_base=frozenset(), items_add=frozenset('colonoscopy'), confidence=0.5, lift=1.0)]),...]
【讨论】:
效果很好。这有点令人困惑,因为 MLXtend 库想要一个 1/0 的数据帧(单热编码),每一列都是项目......以上是关于apyori apriori 损坏的项目集输出的主要内容,如果未能解决你的问题,请参考以下文章
Python --深入浅出Apriori关联分析算法 Apriori关联规则实战