Python熊貓– GroupBy (Python Pandas – GroupBy)
GroupBy method can be used to work on group rows of data together and call aggregate functions. It allows to group together rows based off of a column and perform an aggregate function on them.
GroupBy方法可用于一起處理分組數據行并調用聚合函數。 它允許基于列將行分組在一起，并對它們執行聚合功能。
Consider the below example, there are three partitions of IDS (1, 2, and 3) and several values for them. We can now group by the ID column and aggregate them using some sort of aggregate function. Here we are sum-ing the values and putting the values.
考慮下面的示例，有三個IDS分區(1、2和3)，以及它們的幾個值。 現在，我們可以按ID列進行分組，并使用某種聚合函數對其進行聚合。 在這里，我們將這些值相加并放入這些值。
與熊貓團購 (Groupby with Pandas)
Create a dataframe from a dictionary
從字典創建數據框
import numpy as np
import pandas as pd
data = {'company':['Google','Microsoft','FB','Google','FB'], 'person':['Molly','Nathaniel', 'Sriansh', 'Carl','Sarah'], 'Sales':[200,123,130,144,122]}
df = pd.DataFrame(data)
print(df)
Output
輸出量
company person Sales
0 Google Molly 200
1 Microsoft Nathaniel 123
2 FB Sriansh 130
3 Google Carl 144
4 FB Sarah 122
Following examples illustrate the 'GroupBy' function,
以下示例說明了“ GroupBy”功能 ，
Example 1: GroupBy by 'company'
示例1：按“公司”分組
# returns the groubBy object
print(df.groupby('company'))
'''
<pandas.core.groupby.generic.DataFrameGroupBy object at 0x7f1721585350>
'''
by_company = df.groupby('company')
#invoke aggregate function
print(by_company.mean())
'''
Sales
company
FB 126
Google 172
Microsoft 123
'''
In the above example, we don't see the person column, because the data type is String and by no means, we can get mean of String variables, and hence Pandas automatically ignores any non-numeric values.
在上面的示例中，我們沒有看到person列，因為數據類型是String ，但絕不能獲得String變量的均值 ，因此Pandas自動忽略任何非數字值。
Below are some more examples of aggregate functions,
以下是聚合函數的更多示例，
print(by_company.sum())
'''
Output:
Sales
company
FB 252
Google 344
Microsoft 123
'''
print(by_company.std())
'''
Output:
Sales
company
FB 5.656854
Google 39.597980
Microsoft NaN
'''
Note the return type of the values are by default a DataFrame, as illustrated below,
請注意，默認情況下，值的返回類型為DataFrame，如下所示，
std = by_company.std()
print(type(std))
'''
Output:
<class 'pandas.core.frame.DataFrame'>
'''
And, hence we can perform all the dataFrame functions such as,
并且，因此我們可以執行所有dataFrame函數，例如，
print(by_company.std().loc['FB'])
'''
Output:
Sales 5.656854
Name: FB, dtype: float64
'''
The above mentioned steps, all can be performed in a single step as follows,
上述步驟全部可以在一個步驟中執行，如下所示：
print(df.groupby('company').sum().loc['FB'])
'''
Output:
Sales 252
Name: FB, dtype: int64
'''
Some more aggregate functions are,
還有一些聚合函數，
print(df.groupby('company').count())
'''
Output:
person Sales
company
FB 2 2
Google 2 2
Microsoft 1 1
'''
print(df.groupby('company').max())
'''
Output:
person Sales
company
FB Sriansh 130
Google Molly 200
Microsoft Nathaniel 123
'''
print(df.groupby('company').min())
'''
Output:
person Sales
company
FB Sarah 122
Google Carl 144
Microsoft Nathaniel 123
'''
使用具有描述方法的GroupBy (Using GroupBy with describe method)
The describe() method returns a bunch of useful information all at once.
describe()方法一次返回一堆有用的信息。
print(df.groupby('company').describe())
'''
Output:
Sales ...
count mean std ... 50% 75% max
company ...
FB 2.0 126.0 5.656854 ... 126.0 128.0 130.0
Google 2.0 172.0 39.597980 ... 172.0 186.0 200.0
Microsoft 1.0 123.0 NaN ... 123.0 123.0 123.0
[3 rows x 8 columns]
'''
The format of the description can be changed using transpose() method,
可以使用transpose()方法更改描述的格式，
print(df.groupby('company').describe().transpose())
'''
Output:
company FB Google Microsoft
Sales count 2.000000 2.00000 1.0
mean 126.000000 172.00000 123.0
std 5.656854 39.59798 NaN
min 122.000000 144.00000 123.0
25% 124.000000 158.00000 123.0
50% 126.000000 172.00000 123.0
75% 128.000000 186.00000 123.0
max 130.000000 200.00000 123.0
'''
翻譯自: https://www.includehelp.com/python/python-pandas-groupby.aspx
