我有一个类似于this question的数据透视表,似乎没有答案。我有一个名为grouped
的数据透视表,如下所示:
grouped = age_gender_bkts.pivot_table('population_in_thousands',index='gender',
columns='country_destination', aggfunc='sum').unstack()
这取自pandas dataframe age_gender_bkts:
age_gender_bkts = pd.read_csv('airbnb/age_gender_bkts.csv')
age_gender_bkts[:10]
age_bucket country_destination gender population_in_thousands year
0 100+ AU male 1 2015
1 95-99 AU male 9 2015
2 90-94 AU male 47 2015
3 85-89 AU male 118 2015
4 80-84 AU male 199 2015
5 75-79 AU male 298 2015
6 70-74 AU male 415 2015
7 65-69 AU male 574 2015
8 60-64 AU male 636 2015
9 55-59 AU male 714 2015
对于每个国家/地区,我希望得到男性和女性population_in_thousands
之间的比例,作为每个性别的百分比,例如12024/11899+12024
的{{1}}。
我对pandas非常陌生,numpy,寻找基于AU
计算列的通用解决方案。此外,如果回复有办法让我按性别和国家/地区创建这些群组而不使用pivot_table
,例如pivot_table
(我无法弄清楚),这对我的学习很有帮助。
答案 0 :(得分:0)
您可以使用groupby
,transform
和sum
。最后,您可以将merge
数据转换为原始DataFrame
:
print age_gender_bkts
age_bucket country_destination gender population_in_thousands year
0 100+ AU male 1 2015
1 95-99 AU male 9 2015
2 90-94 CA male 47 2015
3 85-89 CA male 118 2015
4 80-84 AU male 199 2015
5 75-79 NL male 298 2015
6 70-74 NL male 415 2015
7 65-69 AU male 574 2015
8 60-64 AU male 636 2015
9 55-59 AU male 714 2015
grouped = age_gender_bkts.pivot_table('population_in_thousands',index='gender', columns='country_destination', aggfunc='sum').unstack()
df = (grouped / grouped.groupby(level=0).transform(sum)).reset_index().rename(columns={0:'prop'})
print df
country_destination gender prop
0 AU male 1
1 CA male 1
2 NL male 1
print pd.merge(age_gender_bkts, df, on=['country_destination', 'gender'])
age_bucket country_destination gender population_in_thousands year prop
0 100+ AU male 1 2015 1
1 95-99 AU male 9 2015 1
2 80-84 AU male 199 2015 1
3 65-69 AU male 574 2015 1
4 60-64 AU male 636 2015 1
5 55-59 AU male 714 2015 1
6 90-94 CA male 47 2015 1
7 85-89 CA male 118 2015 1
8 75-79 NL male 298 2015 1
9 70-74 NL male 415 2015 1