Python：如何基于datetime获取值计数

我编写了以下代码，创建了两个数据帧

nq

和

cmnt

nq

包含

UserId

和获得徽章的相应时间

日期

cmnt

包含

OwnerUserId

和用户发表评论的时间

CreationDate

我想统计一下在获得徽章1周之前和之后的所有日子里的评论，这样我就可以从中创建一个时间序列线图

以下代码执行相同的操作，但会产生一个KeyError。请提供为所有用户执行此操作的代码

nq

 UserId |   date 
     1      2009-10-17 17:38:32.590
     2      2009-10-19 00:37:23.067
     3      2009-10-20 08:37:14.143
     4      2009-10-21 18:07:51.247
     5      2009-10-22 21:25:24.483

cmnt

OwnerUserId | CreationDate
1             2009-10-16 17:38:32.590
1             2009-10-18 17:38:32.590
2             2009-10-18 00:37:23.067
2             2009-10-17 00:37:23.067
2             2009-10-20 00:37:23.067
3             2009-10-19 08:37:14.143
4             2009-10-20 18:07:51.247
5             2009-10-21 21:25:24.483

UserId     |   date                 |-7|-6|-5|-4|-3|-2|-1|0 |1 |2 |3 |4 |5 |6 |7
     1      2009-10-17 17:38:32.590 |0 |0 |0 |0 |0 |0 |1 |0 |1 |0 |0 |0 |0 |0 |0  
     2      2009-10-19 00:37:23.067 |0 |0 |0 |0 |0 |1 |1 |0 |1 |0 |0 |0 |0 |0 |0    
     3      2009-10-20 08:37:14.143 |0 |0 |0 |0 |0 |0 |1 |0 |0 |0 |0 |0 |0 |0 |0 
     4      2009-10-21 18:07:51.247 |0 |0 |0 |0 |0 |0 |1 |0 |0 |0 |0 |0 |0 |0 |0 
     5      2009-10-22 21:25:24.483 |0 |0 |0 |0 |0 |0 |1 |0 |0 |0 |0 |0 |0 |0 |0 

代码

 nq.date = pd.to_datetime(nq.date)
 cmnt.CreationDate = pd.to_datetime(cmnt.CreationDate)

 count= []
   
 for j in range(len(nq)): 
      for i in range(-7,8):
        
          check_date = nq.date.iloc[j] + timedelta(days=i)
          
          count = cmnt.loc[(cmnt.OwnerUserId == nq.UserId.iloc[j]) & (cmnt.CreationDate == check_date)].shape[0]
          nq.iloc[j].append({nq[i]:count})

预期产出

OwnerUserId | CreationDate
1             2009-10-16 17:38:32.590
1             2009-10-18 17:38:32.590
2             2009-10-18 00:37:23.067
2             2009-10-17 00:37:23.067
2             2009-10-20 00:37:23.067
3             2009-10-19 08:37:14.143
4             2009-10-20 18:07:51.247
5             2009-10-21 21:25:24.483

UserId     |   date                 |-7|-6|-5|-4|-3|-2|-1|0 |1 |2 |3 |4 |5 |6 |7
     1      2009-10-17 17:38:32.590 |0 |0 |0 |0 |0 |0 |1 |0 |1 |0 |0 |0 |0 |0 |0  
     2      2009-10-19 00:37:23.067 |0 |0 |0 |0 |0 |1 |1 |0 |1 |0 |0 |0 |0 |0 |0    
     3      2009-10-20 08:37:14.143 |0 |0 |0 |0 |0 |0 |1 |0 |0 |0 |0 |0 |0 |0 |0 
     4      2009-10-21 18:07:51.247 |0 |0 |0 |0 |0 |0 |1 |0 |0 |0 |0 |0 |0 |0 |0 
     5      2009-10-22 21:25:24.483 |0 |0 |0 |0 |0 |0 |1 |0 |0 |0 |0 |0 |0 |0 |0 

此处，

-1

列表示在获得徽章前一天发表的评论，

1

列表示在获得徽章后一天发表的评论，依此类推

注意

---

有一种完全交替的方法可以做到这一点。我的主要目标是绘制一个时间序列线图，显示用户在获得徽章前后的评论数量。

可能您需要交叉合并、筛选，然后是交叉表。：

# merge the two dataframes
merged = (nq.merge(cmnt, left_on='UserId', 
         right_on='OwnerUserId',
         how='left')
)

# extract the date difference between `date` and `CreationDate`
merged['date_diff'] = merged['date'].dt.normalize() - merged['CreationDate'].dt.normalize()
merged['date_diff'] = (merged['date_diff'] / pd.to_timedelta('1D')).astype(int)

# filter the comments within the range
merged = merged[merged['date_diff'].between(-7,7)]

# crosstab
pd.crosstab([merged['UserId'],merged['date']], merged['date_diff'])

输出：

date_diff                       -1   1   2
UserId date                               
1      2009-10-17 17:38:32.590   1   1   0
2      2009-10-19 00:37:23.067   1   1   1
3      2009-10-20 08:37:14.143   0   1   0
4      2009-10-21 18:07:51.247   0   1   0
5      2009-10-22 21:25:24.483   0   1   0

---

这将提供正确的输出。您能否向解决方案中添加如何将此交叉表转换为数据帧？该交叉表命令将返回一个数据帧。只需将其分配给某个对象，例如，

out=pd.crosstab（…）

。它是，但我希望它有以下列['UserId'，“date”，-7，-6，…0…，6，7]，我可以像普通数据帧列一样访问这些列。但是现在的列是

Int64Index（[-7，-6，-5，-4，-3，-2，-1,0,1,2,3,4,5,6,7]，dtype='int64'，name='date_diff'）

。因此，现在命令df['UserId']给出错误，因为'UserId'不是df的一列。（df=pd.crosstab（…）chain

reset_index（）

使用该

pd.crosstab（）

。对于某些输入，我也会得到此错误

ValueError:无法将非有限值（NA或inf）转换为整数。如何解决这个问题？