Python 使用df、groupby时组内记录数不正确

我从中获得了以下修改代码，用于根据时间戳将行拆分为5秒组

df = pd.read_csv(file_name, delimiter=',')
df['dt'] = pd.to_datetime(df['datetime'], unit='s')
for g in df.groupby(pd.Grouper(freq='5s', key='dt')):
    print(f'Start time {g[0]} has {len(g)} records within 5 secs')

但我在组中得到的记录数量不正确

输出

Start time 2017-05-02 16:00:45 has 2 records within 5 secs
...

示例CSV如下所示

datetime,x,y,z,label
1493740845,0.0004,-0.0001,0.0045,bad
1493740846,0.0004,0.0006,0.0049,bad
1493740847,0.0002,0.0013,0.0044,bad
1493740848,0.0002,0.0005,0.0046,bad
1493740849,0.0006,0.0006,0.0038,bad
1493740850,0.0009,0.0002,0.0038,bad
...

---

有两个值的

g

元组，所以总是得到

2

我认为您可以将元组解压为

name

和

g

变量，然后按照需要工作：

for name, g in df.groupby(pd.Grouper(freq='5s', key='dt')):
    print(f'Start time {name} has {len(g)} records within 5 secs')

Start time 2017-05-02 16:00:45 has 5 records within 5 secs
Start time 2017-05-02 16:00:50 has 1 records within 5 secs

在您的解决方案中，将

g[1]

用于

length

s：

for g in df.groupby(pd.Grouper(freq='5s', key='dt')):
    print(f'Start time {g[0]} has {len(g[1])} records within 5 secs')

---

@耶斯雷尔一世在汉克斯身上加了很多东西，效果很好。我喜欢你的个人资料哲学，关于否决投票的推理。我和你一样same@user158-嗯，事实上应该更好，但现在我认为反对票越少越好，就像我意识到的那样；）它应该小心使用，主要是用于答案。我想你是说使用新的API的新答案？如果是这样的话，可以通过按活动对答案进行排序来解决。耶，重复的答案很烦人，原因是人们试图利用答案排序设置。如果读卡器答案排序设置设置为“活动”，则最新答案将位于顶部，这将导致新读卡器更多的向上投票。@user158-当然，但我这里没有魔力。对于我来说，工作编码，许多行代码都是从其他人那里学习的。所以对我来说这真的是一个漫长的过程。但如果你有

R

或matlab或类似软件的经验，你就可以理解数据处理的原理，这样在so中搜索就更容易了。有很多问题都解决了，只是发现不是那么容易。