在python中按时间分组和绘制数据

Question

我有一个 csv 文件，我正在尝试绘制每月某些值的平均值。我的 csv 文件的结构如下所示，所以我认为我应该每天分组我的数据，然后每月分组以计算平均值。

timestamp,heure,lat,lon,impact,type
2007-01-01 00:00:00,13:58:43,33.837,-9.205,10.3,1
2007-01-02 00:00:00,00:07:28,34.5293,-10.2384,17.7,1
2007-01-02 00:00:00,23:01:03,35.0617,-1.435,-17.1,2
2007-01-03 00:00:00,01:14:29,36.5685,0.9043,36.8,1
2007-01-03 00:00:00,05:03:51,34.1919,-12.5061,-48.9,1

我正在使用此代码：

import pandas as pd
import seaborn as sns
import numpy as np
import matplotlib.pyplot as plt

df= pd.read_csv("ave.txt", sep=',', names =["timestamp","heure","lat","lon","impact","type"])
daily = df.set_index('timestamp').groupby(pd.TimeGrouper(key='timestamp', freq='D', axis=1), axis=1)['impact'].count()
monthly = daily.groupby(pd.TimeGrouper(freq='M')).mean()
ax = monthly.plot(kind='bar')
plt.show()

但是，我不断收到这样的错误：

KeyError: '找不到石斑鱼名称时间戳'

有什么想法吗？？

Answer 1

您收到此错误是因为您已将 timestamp 列设置为 index。尝试从TimeGrouper() 或set_index 方法中删除key='timestamp'，它应该按您的预期分组：

daily = df.set_index('timestamp').groupby(pd.TimeGrouper(freq='D', axis=1), axis=1)['impact'].count()

或

daily = df.groupby(pd.TimeGrouper(key='timestamp', freq='D', axis=1), axis=1)['impact'].count()

Answer 2

我相信你需要DataFrame.resample。

还必须通过read_csv 中的参数parse_dates 和index_col 将timestamp 转换为DataTimeindex。

names =["timestamp","heure","lat","lon","impact","type"]
data = pd.read_csv('fou.txt',names=names, parse_dates=['timestamp'],index_col=['timestamp'])
print (data.head())

#your code
daily = data.groupby(pd.TimeGrouper(freq='D'))['impact'].count()
monthly = daily.groupby(pd.TimeGrouper(freq='M')).mean()
ax = monthly.plot(kind='bar')
plt.show()

#more simpliest
daily = data.resample('D')['impact'].count()
monthly = daily.resample('M').mean()
ax = monthly.plot(kind='bar')
plt.show()

---

还要检查是否真的需要count，而不是size。 What is the difference between size and count in pandas?

daily = data.resample('D')['impact'].size()
monthly = daily.resample('M').mean()
ax = monthly.plot(kind='bar')
plt.show()