识别出现在特定年份而不是另一组年份的记录答案

【问题标题】：Identify records that are present in particular set of years and not in the another set of years识别出现在特定年份而不是另一组年份的记录
【发布时间】：2022-07-20 01:48:37
【问题描述】：

我正在尝试根据 ID 和年份标记行，如果 ID 出现在 [2017,2018,2019] 年份并且 未出现 在 [2020,2021,2022] 中，则需要将其标记为 1 else 0。

df1 = pd.DataFrame({'ID': ['AX1', 'Ax1', 'AX1','AX1','AX1','AX1','AX2','AX2','AX2','AX3','AX3','AX4','AX4','AX4'],'year':[2017,2018,2019,2020,2021,2022,2019,2020,2022,2019,2020,2017,2018,2019]})

     ID  year
0   AX1  2017
1   Ax1  2018
2   AX1  2019
3   AX1  2020
4   AX1  2021
5   AX1  2022
6   AX2  2019
7   AX2  2020
8   AX2  2022
9   AX3  2019
10  AX3  2020
11  AX4  2017
12  AX4  2018
13  AX4  2019

预期输出：

     ID  year  label
0   AX1  2017      0
1   Ax1  2018      0
2   AX1  2019      0
3   AX1  2020      0
4   AX1  2021      0
5   AX1  2022      0
6   AX2  2019      0
7   AX2  2020      0
8   AX2  2022      0
9   AX3  2019      0
10  AX3  2020      0
11  AX4  2017      1
12  AX4  2018      1
13  AX4  2019      1

在上面的示例中 ID：AX4 被标记为 1，因为它是唯一出现在第一组年份 [2017,2018,2019] 并且没有出现在第二组年份 [2020] 中的 ID ,2021,2022]。

我如何做到这一点？

【问题讨论】：

标签： python python-3.x pandas dataframe

【解决方案1】：

按 ID 分组，使用集合操作检查是否需要的年份和不需要的年份，并将结果映射回df1。

df1 = pd.DataFrame({'ID': ['AX1', 'AX1', 'AX1','AX1','AX1','AX1','AX2','AX2','AX2','AX3','AX3','AX4','AX4','AX4'],'year':[2017,2018,2019,2020,2021,2022,2019,2020,2022,2019,2020,2017,2018,2019]})

# find group level labels by checking if all of 2017-19 and none of 2020-22 exist for each ID
gr_lbl = df1.groupby('ID')['year'].apply(lambda g: {2017,2018,2019}.issubset(g) and not bool({2020,2021,2022}.intersection(g)))*1
# map group level labels to ID
df1['labels'] = df1['ID'].map(gr_lbl)

另一个（更易读的代码）是交叉制表df1 并检查跨列的年份。 pd.crosstab() 对列进行排序（在本例中为年份），所以简单的 eq() 可以工作。

# cross tabulate and check for years across columns
labels = pd.crosstab(df1['ID'], df1['year']).eq([1,1,1,0,0,0], axis=1).all(1)*1
# map group level labels to ID
df1['labels'] = df1['ID'].map(labels)
df1

【讨论】：

【解决方案2】：

import pandas as pd

df1 = pd.DataFrame({'ID': ['AX1', 'Ax1', 'AX1','AX1','AX1','AX1','AX2','AX2','AX2','AX3','AX3','AX4','AX4','AX4'],'year':[2017,2018,2019,2020,2021,2022,2019,2020,2022,2019,2020,2017,2018,2019]})

include = set()
exclude = set()

for ID, year in zip(df1['ID'], df1['year']):
    if year in [2017,2018,2019]:
        include.add(ID.upper())
    if year in [2020,2021,2022]:
        exclude.add(ID.upper())
        
df1['label'] = [int(x.upper() in include - exclude) for x in df1['ID']]

print(df1)

【讨论】：

【解决方案3】：

通过聚合sets 创建Series，然后通过set.issubset 进行比较，最后将输出映射到新列：

y1 = set([2017,2018,2019])
y2 = set([2020,2021,2022])

s = df1.groupby('ID')['year'].agg(set)
df1['label'] = df1['ID'].map((s.map(y1.issubset) & ~s.map(y2.issubset)).astype(int))
print (df1)
     ID  year  label
0   AX1  2017      0
1   Ax1  2018      0
2   AX1  2019      0
3   AX1  2020      0
4   AX1  2021      0
5   AX1  2022      0
6   AX2  2019      0
7   AX2  2020      0
8   AX2  2022      0
9   AX3  2019      0
10  AX3  2020      0
11  AX4  2017      1
12  AX4  2018      1
13  AX4  2019      1

详情：

print (df1.groupby('ID')['year'].agg(set))
ID
AX1    {2017, 2019, 2020, 2021, 2022}
AX2                {2019, 2020, 2022}
AX3                      {2019, 2020}
AX4                {2017, 2018, 2019}
Ax1                            {2018}
Name: year, dtype: object()

print ((s.map(y1.issubset) & ~s.map(y2.issubset)).astype(int))
ID
AX1    0
AX2    0
AX3    0
AX4    1
Ax1    0
Name: year, dtype: int32

【讨论】：