在 Pandas 中，str 值随时间和客户维度发生变化

Question

我有一些日期的客户数据，我想看看他们是否会随着时间的推移选择另一种产品。理想情况下，我想将发生更改的两列复制到一个新列中。

所以，如果我有一张这样的桌子

period, Customer , product
2020-01, Cust1, 12 TS
2020-02, Cust1, 12 TS
2020-03, Cust1, 14 SLM
2020-01, Cust2, 12 SLM
2020-02, Cust2, 12 TS
2020-03, Cust2, 14 SLM

所以 cust1 随着时间的推移从 TS 转到 SLM，而 Cust2 从 SLM 转到 TS，然后相反。 最后一列应如下所示：

period, Customer , product , change
2020-01, Cust1, 12 TS , NAN
2020-02, Cust1, 12 TS , NAN
2020-03, Cust1, 14 SLM, from TS to SLM
2020-01, Cust2, 12 SLM, NAN
2020-02, Cust2, 12 TS, from SLM to TS
2020-03, Cust2, 14 SLM, from TS to SLM

我查看了许多可用的解决方案，例如 here，但我无法按照我想要的方式进行操作。

Answer 1

我们可以通过多种方式做到这一点，我建议使用shift 和groupby 查找最大记录，然后使用.loc 适当地过滤您的查询集。

设置。

from io import StringIO
import pandas as pd


d = """period, Customer, quantity , product
2020-01, Cust1, 12, TS
2020-02, Cust1, 12, TS
2020-03, Cust1, 14, SLM
2020-01, Cust2, 12, SLM
2020-02, Cust2, 12, TS
2020-03, Cust2, 14, SLM"""

---

df = pd.read_csv(StringIO(d),sep=',',parse_dates=['period'])
# as you have spaces in your csv above.
#df.columns = df.columns.str.strip()

#create a record end date. 
df['period_end_date'] = df.groupby('Customer')['period'].shift(-1)
#find the previous product. 
df.loc[df['period_end_date'].isna(), 
                             'previous_product'] = df.groupby('Customer')['product'].shift(1)

---

此处的当前记录将是preiod_end_date 为空的位置。

print(df)

      period Customer  quantity product period_end_date previous_product
0 2020-01-01    Cust1        12      TS      2020-02-01              NaN
1 2020-02-01    Cust1        12      TS      2020-03-01              NaN
2 2020-03-01    Cust1        14     SLM             NaT               TS
3 2020-01-01    Cust2        12     SLM      2020-02-01              NaN
4 2020-02-01    Cust2        12      TS      2020-03-01              NaN
5 2020-03-01    Cust2        14     SLM             NaT               TS

如果您需要按照上面概述的预定义格式。

df.loc[df['period_end_date'].isna(), 
                            'previous_product'] = ("FROM " 
                            + df.groupby('Customer')['product'].shift(1) 
                            + " TO " 
                            + df['product'] )

      period Customer  quantity product period_end_date  previous_product
0 2020-01-01    Cust1        12      TS      2020-02-01               NaN
1 2020-02-01    Cust1        12      TS      2020-03-01               NaN
2 2020-03-01    Cust1        14     SLM             NaT  FROM  TS TO  SLM
3 2020-01-01    Cust2        12     SLM      2020-02-01               NaN
4 2020-02-01    Cust2        12      TS      2020-03-01               NaN
5 2020-03-01    Cust2        14     SLM             NaT  FROM  TS TO  SLM

Answer 2

我们可以先通过Customergroupdataframe，然后shift检查是否有变化。之后我们可以比较并确定变化。

df['prev_product'] = df.groupby(['Customer'])['product'].shift().bfill()

df['change'] = df[['product', 'prev_product']].apply(lambda x: None if(x[0] == x[1]) else f'from {x[1]}  to {x[0]}', axis=1)

    period  Customer    n   product prev_product    change
0   2020-01 Cust1       12  TS      TS              None
1   2020-02 Cust1       12  TS      TS              None
2   2020-03 Cust1       14  SLM     TS              from TS to SLM
3   2020-01 Cust2       12  SLM     SLM             None
4   2020-02 Cust2       12  TS      SLM             from SLM to TS
5   2020-03 Cust2       14  SLM     TS              from TS to SLM
​

注意：df.drop('prev_product',axis=1) 不需要。