Python 使用基于其他列的值填充np.nan

我尝试将

offer\u id

与相应的事务匹配。这是数据集：

       time            event                          offer_id  amount
2077      0   offer received  f19421c1d4aa40978ebb69ca19b0e20d     NaN
15973     6     offer viewed  f19421c1d4aa40978ebb69ca19b0e20d     NaN
15974     6      transaction                               NaN    3.43
18470    12      transaction                               NaN    6.01
18471    12  offer completed  f19421c1d4aa40978ebb69ca19b0e20d     NaN
43417   108      transaction                               NaN   11.00
44532   114      transaction                               NaN    1.69
50587   150      transaction                               NaN    3.23
55277   168   offer received  9b98b8c7a33c4b65b9aebfe6a799e6d9     NaN
96598   258      transaction                               NaN    2.18

规则是，当查看报价时，交易属于此报价id。如果报价已收到，但未查看，则交易不属于报价id。我希望

time

变量可以清楚地说明这一点。这是期望的结果：

       time            event                          offer_id  amount
2077      0   offer received  f19421c1d4aa40978ebb69ca19b0e20d     NaN
15973     6     offer viewed  f19421c1d4aa40978ebb69ca19b0e20d     NaN
15974     6      transaction  f19421c1d4aa40978ebb69ca19b0e20d    3.43
18470    12      transaction  f19421c1d4aa40978ebb69ca19b0e20d    6.01
18471    12  offer completed  f19421c1d4aa40978ebb69ca19b0e20d     NaN
43417   108      transaction                               NaN   11.00
44532   114      transaction                               NaN    1.69
50587   150      transaction                               NaN    3.23
55277   168   offer received  9b98b8c7a33c4b65b9aebfe6a799e6d9     NaN
96598   258      transaction                               NaN    2.18

示例代码：

import pandas as pd
import numpy as np

d = {'time': [0, 6, 6, 12, 12, 108, 144, 150, 168, 258], 
     'event': ["offer received", "offer viewed", "transaction", "transaction", "offer completed", "transaction", "transaction", "transaction", "offer received", "transaction"], 
     'offer_id': ["f19421c1d4aa40978ebb69ca19b0e20d", "f19421c1d4aa40978ebb69ca19b0e20d", np.nan, np.nan, "f19421c1d4aa40978ebb69ca19b0e20d", np.nan, np.nan, np.nan, "9b98b8c7a33c4b65b9aebfe6a799e6d9", np.nan]}

df = pd.DataFrame(d)

print("Original data:\n{}\n".format(df))

is_offer_viewed = False
now_offer_id = np.nan
for index, row in df.iterrows():
    if row['event'] == "offer viewed":
        is_offer_viewed = True
        now_offer_id = row['offer_id']
        
    elif row['event'] == "transaction" and is_offer_viewed:
        df.at[index, 'offer_id'] = now_offer_id

    elif row['event'] == "offer completed":
        is_offer_viewed = False
        now_offer_id = np.nan

print("Processed data:\n{}\n".format(df))

产出：

Original data:
   time            event                          offer_id
0     0   offer received  f19421c1d4aa40978ebb69ca19b0e20d
1     6     offer viewed  f19421c1d4aa40978ebb69ca19b0e20d
2     6      transaction                               NaN
3    12      transaction                               NaN
4    12  offer completed  f19421c1d4aa40978ebb69ca19b0e20d
5   108      transaction                               NaN
6   144      transaction                               NaN
7   150      transaction                               NaN
8   168   offer received  9b98b8c7a33c4b65b9aebfe6a799e6d9
9   258      transaction                               NaN

Processed data:
   time            event                          offer_id
0     0   offer received  f19421c1d4aa40978ebb69ca19b0e20d
1     6     offer viewed  f19421c1d4aa40978ebb69ca19b0e20d
2     6      transaction  f19421c1d4aa40978ebb69ca19b0e20d
3    12      transaction  f19421c1d4aa40978ebb69ca19b0e20d
4    12  offer completed  f19421c1d4aa40978ebb69ca19b0e20d
5   108      transaction                               NaN
6   144      transaction                               NaN
7   150      transaction                               NaN
8   168   offer received  9b98b8c7a33c4b65b9aebfe6a799e6d9
9   258      transaction                               NaN

---

直到报价完成，对吗？是的，完全正确，在那之后必须收到新的报价，并且viewed@DataMastery很乐意帮忙：）