Question

我正在解决这个看起来很简单的问题。数据如下：

例如，客户ID的当前事件ID为abc。现在，我需要以列表格式查找所有客户的所有eventId，以便记录第一个事件ID，然后记录下一个事件，直到最新的事件ID。

我为1位客户使用的方法如下：

import pandas as pd
import numpy as np
data = pd.read_csv('test.csv')
data.to_dict()
{'customerid': {0: 233, 1: 250, 2: 233, 3: 250, 4: 233},
 'eventid': {0: 'abc', 1: 'bcd', 2: 'edc', 3: 'abl', 4: 'cdl'},
 'date': {0: '2019-12-10',
  1: '2019-12-08',
  2: '2008-12-10',
  3: '2019-12-01',
  4: '2001-12-10'},
 'previouseventid': {0: 'edc', 1: 'abl', 2: 'cdl', 3: np.nan, 4: np.nan}}

    customerid  eventid date    previouseventid
0   233 abc 2019-12-10  edc
1   250 bcd 2019-12-08  abl
2   233 edc 2008-12-10  cdl
3   250 abl 2019-12-01  
4   233 cdl 2001-12-10

temp = [cust_233['eventid'][0]]

for i in range(len(cust_233['previouseventid'])-1):
    if pd.isna(cust_233['previouseventid'][i]) == False:
        # print(cust_233['previouseventid'][i])
        temp.append(cust_233['previouseventid'][i])
    else:
        # print('now exiting')
        break

我觉得我的方法有点笨拙，并且有很多代码。如何为所有客户有效地解决问题？

已更新：

我需要的输出是列表。客户233的预期输出是列表['cdl'，'edc'，'abc']，客户250的预期输出是['abl'，'bcd']

Answer 1

Groupby然后移动应该可以：

# First, make sure your data is sorted from oldest to newest
df['date'] = pd.to_datetime(df['date'])
df.sort_values('date', inplace=True)

# Get previous event through groupby operation
df['prev_id'] = df.groupby('customerid')['eventid'].shift(1)

如果要为每个客户提供清单：

# create a dictionary with stored values – keys are customer id
prev_events_dict = df.groupby('customerid')['eventid'].apply(list).to_dict()
# map dict to dataframe
df['list_of_prev_id'] = df['customerid'].map(prev_events_dict)

Answer 2

您可以创建如下列表：

df['previouseventid'] = df['customerid'].map(df.groupby('customerid')['eventid'].apply(list))

输出：

   customerid eventid        date  previouseventid
0         233     abc  2019-12-10  [abc, edc, cdl]
1         250     bcd  2019-12-08       [bcd, abl]
2         233     edc  2008-12-10  [abc, edc, cdl]
3         250     abl  2019-12-01       [bcd, abl]
4         233     cdl  2001-12-10  [abc, edc, cdl]

df.groupby（'customerid'）['eventid']。apply（list）仅列出列表

df.groupby('customerid')['eventid'].apply(list)                                                                                                                                     

customerid
233    [abc, edc, cdl]
250         [bcd, abl]
Name: eventid, dtype: object

Answer 3

尝试一下：

data.sort_values('date', ascending=True).groupby('customerid', sort=False)['eventid'].agg(list)

输出：

customerid
233    [cdl, edc, abc]
250         [abl, bcd]
Name: eventid, dtype: object

按客户查找所有先前事件ID的最佳方法

3 个答案: