熊猫转换时间戳和每月摘要

Question

我有几个通过 Pandas 导入的 .csv 文件，然后计算出数据摘要（最小值、最大值、平均值），最好是每周和每月报告。我有以下代码，但似乎无法使月份摘要起作用，我确定问题出在时间戳转换上。

我做错了什么？

import pandas as pd
import numpy as np

#Format of the data that is been imported
#2017-05-11 18:29:14+00:00,264.0,987.99,26.5,23.70,512.0,11.763,52.31

df = pd.read_csv('data.csv')
df['timestamp'] = pd.to_datetime(df['time'], format='%Y-%m-%d %H:%M:%S')

print 'month info'
print [g for n, g in df.groupby(pd.Grouper(key='timestamp',freq='M'))]
print(data.groupby('timestamp')['light'].mean())

Answer 1

IIUC，您几乎拥有它，并且您的日期时间转换很好。这是一个例子：

从这样的数据框开始（这是您的示例行，经过轻微修改重复）：

>>> df
                        time      x       y     z     a      b       c      d
0  2017-05-11 18:29:14+00:00  264.0  947.99  24.5  53.7  511.0  11.463  12.31
1  2017-05-15 18:29:14+00:00  265.0  957.99  25.5  43.7  512.0  11.563  22.31
2  2017-05-21 18:29:14+00:00  266.0  967.99  26.5  33.7  513.0  11.663  32.31
3  2017-06-11 18:29:14+00:00  267.0  977.99  26.5  23.7  514.0  11.763  42.31
4  2017-06-22 18:29:14+00:00  268.0  997.99  27.5  13.7  515.0  11.800  52.31

你可以用你的日期时间做你以前做过的事情：

df['timestamp'] = pd.to_datetime(df['time'], format='%Y-%m-%d %H:%M:%S')

然后分别获取您的摘要：

monthly_mean = df.groupby(pd.Grouper(key='timestamp',freq='M')).mean()
monthly_max = df.groupby(pd.Grouper(key='timestamp',freq='M')).max()
monthly_min = df.groupby(pd.Grouper(key='timestamp',freq='M')).min()

weekly_mean = df.groupby(pd.Grouper(key='timestamp',freq='W')).mean()
weekly_min = df.groupby(pd.Grouper(key='timestamp',freq='W')).min()
weekly_max = df.groupby(pd.Grouper(key='timestamp',freq='W')).max()

# Examples:
>>> monthly_mean
                x       y     z     a      b        c      d
timestamp                                                   
2017-05-31  265.0  957.99  25.5  43.7  512.0  11.5630  22.31
2017-06-30  267.5  987.99  27.0  18.7  514.5  11.7815  47.31

>>> weekly_mean
                x       y     z     a      b       c      d
timestamp                                                  
2017-05-14  264.0  947.99  24.5  53.7  511.0  11.463  12.31
2017-05-21  265.5  962.99  26.0  38.7  512.5  11.613  27.31
2017-05-28    NaN     NaN   NaN   NaN    NaN     NaN    NaN
2017-06-04    NaN     NaN   NaN   NaN    NaN     NaN    NaN
2017-06-11  267.0  977.99  26.5  23.7  514.0  11.763  42.31
2017-06-18    NaN     NaN   NaN   NaN    NaN     NaN    NaN
2017-06-25  268.0  997.99  27.5  13.7  515.0  11.800  52.31

或将它们全部聚合在一起以获得带有摘要的多索引数据框：

monthly_summary = df.groupby(pd.Grouper(key='timestamp',freq='M')).agg(['mean', 'min', 'max'])
weekly_summary = df.groupby(pd.Grouper(key='timestamp',freq='W')).agg(['mean', 'min', 'max'])

# Example of summary of row 'x':
>>> monthly_summary['x']
             mean    min    max
timestamp                      
2017-05-31  265.0  264.0  266.0
2017-06-30  267.5  267.0  268.0

>>> weekly_summary['x']
             mean    min    max
timestamp                      
2017-05-14  264.0  264.0  264.0
2017-05-21  265.5  265.0  266.0
2017-05-28    NaN    NaN    NaN
2017-06-04    NaN    NaN    NaN
2017-06-11  267.0  267.0  267.0
2017-06-18    NaN    NaN    NaN
2017-06-25  268.0  268.0  268.0