文章目录
- 3. Summary Functions and Maps
- 3.1 Summary Functions 数据总结函数
- 3.1.1 describe()
- 3.1.2 mean(),median(),idxmax(),unique(),value_counts()
- 3.2 Maps 映射
- 3.2.1 map()
- 3.2.2 apply()
- 3.2.3 内置转换方法
- 4. Grouping and Sorting
- 4.1 Grouping 分组
- 4.1.1 groupby()
- 4.1.2 agg()
- 4.1.3 multi_indexes
- 4.2 sort_values() 排序
learn from https://www.kaggle.com/learn/pandas
上一篇:Pandas入门1(DataFrame+Series读写/Index+Select+Assign)
下一篇:Pandas入门3(dtype+fillna+replace+rename+concat+join)
3. Summary Functions and Maps
3.1 Summary Functions 数据总结函数
3.1.1 describe()
wine_rev.points.describe()
,各种统计信息,数字信息总结
# 数字列的总结
count 129971.000000
mean 88.447138
std 3.039730
min 80.000000
25% 86.000000
50% 88.000000
75% 91.000000
max 100.000000
Name: points, dtype: float64
wine_rev.country.describe()
,文字信息总结
# 文字类列的总结
count 129908
unique 43
top US
freq 54504
Name: country, dtype: object
3.1.2 mean(),median(),idxmax(),unique(),value_counts()
wine_rev.points.mean()
,均值,median()
,中位数,idxmax()
,最大数据的下标wine_rev.country.unique()
,不同的值多少个
array(['Italy', 'Portugal', 'US', 'Spain', 'France', 'Germany','Argentina', 'Chile', 'Australia', 'Austria', 'South Africa','New Zealand', 'Israel', 'Hungary', 'Greece', 'Romania', 'Mexico','Canada', nan, 'Turkey', 'Czech Republic', 'Slovenia','Luxembourg', 'Croatia', 'Georgia', 'Uruguay', 'England','Lebanon', 'Serbia', 'Brazil', 'Moldova', 'Morocco', 'Peru','India', 'Bulgaria', 'Cyprus', 'Armenia', 'Switzerland','Bosnia and Herzegovina', 'Ukraine', 'Slovakia', 'Macedonia','China', 'Egypt'], dtype=object)
wine_rev.country.value_counts()
,各个值的计数
US 54504
France 22093
Italy 19540
Spain 6645
.....
Egypt 1
China 1
Name: country, dtype: int64
3.2 Maps 映射
3.2.1 map()
wine_points_mean = wine_rev.points.mean()
wine_rev.points.map(lambda p : p-wine_points_mean)
,将数据变到均值上下(产生一个Series,原DF数据没变)
0 -1.447138
1 -1.447138
2 -1.447138
3 -1.447138
4 -1.447138...
129966 1.552862
129967 1.552862
129968 1.552862
129969 1.552862
129970 1.552862
Name: points, Length: 129971, dtype: float64
3.2.2 apply()
通过定义函数,使用apply对整个表进行转换,对每一行进行操作
def remean_points(row):row.points = row.points - wine_points_meanreturn row
wine_rev.apply(remean_points,axis='columns')
Note that map() and apply() return new, transformed Series and DataFrames, respectively.
They don’t modify the original data they’re called on.
上面两种方法都不会修改原始数据
3.2.3 内置转换方法
wine_rev.points - wine_points_mean
,直接相减就可以,每个数据都会减去右边的单个value
0 -1.447138
1 -1.447138
2 -1.447138
3 -1.447138
4 -1.447138...
129966 1.552862
129967 1.552862
129968 1.552862
129969 1.552862
129970 1.552862
Name: points, Length: 129971, dtype: float64
wine_rev.country + '-' + wine_rev.region_1
,相等长度的两个Series操作,直接1v1对应起来
0 Italy-Etna
1 NaN
2 US-Willamette Valley
3 US-Lake Michigan Shore
4 US-Willamette Valley...
129966 NaN
129967 US-Oregon
129968 France-Alsace
129969 France-Alsace
129970 France-Alsace
Length: 129971, dtype: object
4. Grouping and Sorting
4.1 Grouping 分组
4.1.1 groupby()
wine_rev.groupby('points').points.count()
points
80 397
81 692
82 1836
83 3025
84 6480
85 9530
86 12600
87 16933
88 17207
89 12226
90 15410
91 11359
92 9613
93 6489
94 3758
95 1535
96 523
97 229
98 77
99 33
100 19
Name: points, dtype: int64
wine_rev.groupby('points').price.min()
,按得分分组,然后每组里面价格最低的
points
80 5.0
81 5.0
82 4.0
83 4.0
84 4.0
85 4.0
86 4.0
87 5.0
88 6.0
89 7.0
90 8.0
91 7.0
92 11.0
93 12.0
94 13.0
95 20.0
96 20.0
97 35.0
98 50.0
99 44.0
100 80.0
Name: price, dtype: float64
wine_rev.groupby('points').apply(lambda df : df.title.iloc[0])
,按得分分组后,每个DataFrame的 title的第一行,代码产生的是一个Series
points
80 Viña Tarapacá 2015 Gran Reserva Chardonnay (Le...
81 Pura 8 2010 Grand Reserve Pinot Noir (Rapel Va...
82 Mémoires 2015 Rosé (Coteaux Varois en Provence)
83 Koyle 2015 Costa Pinot Noir (Colchagua Costa)
84 Three Brothers 2014 Zero Degree Dry Riesling (...
85 Casa Silva 2008 Gran Reserva Petit Verdot (Col...
86 Clarksburg Wine Company 2010 Chenin Blanc (Cla...
87 Nicosia 2013 Vulkà Bianco (Etna)
88 Fattoria Sardi 2015 Rosato (Toscana)
89 David Fulton 2008 Petite Sirah (St. Helena)
90 Beaumont 2005 Hope Marguerite Chenin Blanc (Wa...
91 Le Riche 2003 Cabernet Sauvignon Reserve Caber...
92 Dopff & Irion 2004 Schoenenbourg Grand Cru Ven...
93 Claiborne & Churchill 2014 Twin Creeks Estate ...
94 Sandeman 2015 Quinta do Seixo Vintage (Port)
95 Jasper Hill 2013 Georgia's Paddock Shiraz (Hea...
96 Oremus 2005 Eszencia (Tokaji)
97 Robert Weil 2014 Kiedrich Gräfenberg Trockenbe...
98 Chambers Rosewood Vineyards NV Rare Muscadelle...
99 Quilceda Creek 2008 Cabernet Sauvignon (Columb...
100 Chambers Rosewood Vineyards NV Rare Muscat (Ru...
dtype: object
wine_rev.groupby(['country','province']).apply(lambda df : df.loc[df.points.idxmax()])
按照,先按国家分组、再按省份分组,每个组里得分最大的,产生的是一个DataFrame
4.1.2 agg()
wine_rev.groupby(['country']).price.agg([len,min,max])
,后面可以跟一些统计量
4.1.3 multi_indexes
country_rev = wine_rev.groupby(['country','province']).description.agg([len])
,多个特征的分组是多索引的
country_rev.index
,MultiIndex 多级索引
MultiIndex([('Argentina', 'Mendoza Province'),('Argentina', 'Other'),( 'Armenia', 'Armenia'),('Australia', 'Australia Other'),('Australia', 'New South Wales'),('Australia', 'South Australia'),('Australia', 'Tasmania'),('Australia', 'Victoria'),('Australia', 'Western Australia'),( 'Austria', 'Austria'),...( 'US', 'Washington'),( 'US', 'Washington-Oregon'),( 'Ukraine', 'Ukraine'),( 'Uruguay', 'Atlantida'),( 'Uruguay', 'Canelones'),( 'Uruguay', 'Juanico'),( 'Uruguay', 'Montevideo'),( 'Uruguay', 'Progreso'),( 'Uruguay', 'San Jose'),( 'Uruguay', 'Uruguay')],names=['country', 'province'], length=425)
- 转换多级索引为普通索引,
cr = country_rev.reset_index()
,需赋值给一个新的DF
4.2 sort_values() 排序
上面例子可以看出,输出都是按照 index 排序的,我们有时希望按值排序。
cr.sort_values(by='len')
,默认升序
cr.sort_values(by='len',ascending=False)
,降序(升序=False)cr.sort_index()
,恢复按 index 升序
- 按多个值进行排序,
cr.sort_values(by=['country', 'len'])
,先按国家字符串升序,然后按长度升序
cr.sort_values(by=['country', 'len'],ascending=[False,True])
,还可分别指定,每个特征是升序还是降序
上一篇:Pandas入门1(DataFrame+Series读写/Index+Select+Assign)
下一篇:Pandas入门3(dtype+fillna+replace+rename+concat+join)