第2章 索引

第2章 索引

在这里插入图片描述

import numpy as np
import pandas as pd
df = pd.read_csv('data/table.csv',index_col='ID')
df.head()
SchoolClassGenderAddressHeightWeightMathPhysics
ID
1101S_1C_1Mstreet_11736334.0A+
1102S_1C_1Fstreet_21927332.5B+
1103S_1C_1Mstreet_21868287.2B+
1104S_1C_1Fstreet_21678180.4B-
1105S_1C_1Fstreet_41596484.8B+

一、单级索引

1. loc方法、iloc方法、[]操作符

最常用的索引方法可能就是这三类,其中iloc表示位置索引,loc表示标签索引,[]也具有很大的便利性,各有特点
(a)loc方法(注意:所有在loc中使用的切片全部包含右端点!)
① 单行索引:
type(df.loc[1103])
pandas.core.series.Series
② 多行索引:
df.loc[[1102,2304]]
SchoolClassGenderAddressHeightWeightMathPhysics
ID
1102S_1C_1Fstreet_21927332.5B+
2304S_2C_3Fstreet_61648195.5A-
df.loc[1304:].head()
SchoolClassGenderAddressHeightWeightMathPhysics
ID
1304S_1C_3Mstreet_21957085.2A
1305S_1C_3Fstreet_51876961.7B-
2101S_2C_1Mstreet_71748483.3C
2102S_2C_1Fstreet_61616150.6B+
2103S_2C_1Mstreet_41576152.5B-
df.loc[2402::-1].head()
SchoolClassGenderAddressHeightWeightMathPhysics
ID
2402S_2C_4Mstreet_71668248.7B
2401S_2C_4Fstreet_21926245.3A
2305S_2C_3Mstreet_41877348.9B
2304S_2C_3Fstreet_61648195.5A-
2303S_2C_3Fstreet_71909965.9C
df.loc[2305:].head()
SchoolClassGenderAddressHeightWeightMathPhysics
ID
2305S_2C_3Mstreet_41877348.9B
2401S_2C_4Fstreet_21926245.3A
2402S_2C_4Mstreet_71668248.7B
2403S_2C_4Fstreet_61586059.7B+
2404S_2C_4Fstreet_21608467.7B
③ 单列索引:
df.loc[:,'Height'].head()
ID
1101    173
1102    192
1103    186
1104    167
1105    159
Name: Height, dtype: int64
④ 多列索引:
df.loc[:,['Height','Math']].head()
HeightMath
ID
110117334.0
110219232.5
110318687.2
110416780.4
110515984.8
df.loc[:,'Height':'Math'].head()
HeightWeightMath
ID
11011736334.0
11021927332.5
11031868287.2
11041678180.4
11051596484.8
⑤ 联合索引:
df.loc[1102:2401:3,'Height':'Math'].head()
HeightWeightMath
ID
11021927332.5
11051596484.8
12031605358.8
13011616831.5
13041957085.2
⑥ 函数式索引:
df.loc[lambda x:x['Gender']=='M'].head()
#loc中使用的函数,传入参数就是前面的df
SchoolClassGenderAddressHeightWeightMathPhysics
ID
1101S_1C_1Mstreet_11736334.0A+
1103S_1C_1Mstreet_21868287.2B+
1201S_1C_2Mstreet_51886897.0A-
1203S_1C_2Mstreet_61605358.8A+
1301S_1C_3Mstreet_41616831.5B+
def f(x):
    return [1101,1103]
df.loc[f]
SchoolClassGenderAddressHeightWeightMathPhysics
ID
1101S_1C_1Mstreet_11736334.0A+
1103S_1C_1Mstreet_21868287.2B+
⑦ 布尔索引(将重点在第2节介绍)
df.loc[df['Address'].isin(['street_7','street_4'])].head()
SchoolClassGenderAddressHeightWeightMathPhysics
ID
1105S_1C_1Fstreet_41596484.8B+
1202S_1C_2Fstreet_41769463.5B-
1301S_1C_3Mstreet_41616831.5B+
1303S_1C_3Mstreet_71888249.7B
2101S_2C_1Mstreet_71748483.3C
df.loc[[True if i[-1]=='4' or i[-1]=='7' else False for i in df['Address'].values]].head()
SchoolClassGenderAddressHeightWeightMathPhysics
ID
1105S_1C_1Fstreet_41596484.8B+
1202S_1C_2Fstreet_41769463.5B-
1301S_1C_3Mstreet_41616831.5B+
1303S_1C_3Mstreet_71888249.7B
2101S_2C_1Mstreet_71748483.3C
小节:本质上说,loc中能传入的只有布尔列表和索引子集构成的列表,只要把握这个原则就很容易理解上面那些操作
(b)iloc方法(注意与loc不同,切片右端点不包含)
① 单行索引:
df.iloc[3]
School          S_1
Class           C_1
Gender            F
Address    street_2
Height          167
Weight           81
Math           80.4
Physics          B-
Name: 1104, dtype: object
② 多行索引:
df.iloc[3:5]
SchoolClassGenderAddressHeightWeightMathPhysics
ID
1104S_1C_1Fstreet_21678180.4B-
1105S_1C_1Fstreet_41596484.8B+
③ 单列索引:
df.iloc[:,3].head()
ID
1101    street_1
1102    street_2
1103    street_2
1104    street_2
1105    street_4
Name: Address, dtype: object
④ 多列索引:
df.iloc[:,7::-2].head()
PhysicsWeightAddressClass
ID
1101A+63street_1C_1
1102B+73street_2C_1
1103B+82street_2C_1
1104B-81street_2C_1
1105B+64street_4C_1
⑤ 混合索引:
df.iloc[3::4,7::-2].head()
PhysicsWeightAddressClass
ID
1104B-81street_2C_1
1203A+53street_6C_2
1302A-57street_1C_3
2101C84street_7C_1
2105A81street_4C_1
⑥ 函数式索引:
df.iloc[lambda x:[3]].head()
SchoolClassGenderAddressHeightWeightMathPhysics
ID
1104S_1C_1Fstreet_21678180.4B-
小节:由上所述,iloc中接收的参数只能为整数或整数列表,不能使用布尔索引
(c) []操作符
如果不想陷入困境,请不要在行索引为浮点时使用[]操作符,因为在Series中的浮点[]并不是进行位置比较,而是值比较,非常特殊
(c.1)Series的[]操作
① 单元素索引:
s = pd.Series(df['Math'],index=df.index)
s[1101]
#使用的是索引标签
34.0
② 多行索引:
s[0:4]
#使用的是绝对位置的整数切片,与元素无关,这里容易混淆
ID
1101    34.0
1102    32.5
1103    87.2
1104    80.4
Name: Math, dtype: float64
③ 函数式索引:
s[lambda x: x.index[16::-6]]
#注意使用lambda函数时,直接切片(如:s[lambda x: 16::-6])就报错,此时使用的不是绝对位置切片,而是元素切片,非常易错
ID
2102    50.6
1301    31.5
1105    84.8
Name: Math, dtype: float64
④ 布尔索引:
s[s>80]
ID
1103    87.2
1104    80.4
1105    84.8
1201    97.0
1302    87.7
1304    85.2
2101    83.3
2205    85.4
2304    95.5
Name: Math, dtype: float64
(c.2)DataFrame的[]操作
① 单行索引:
df[1:2]
#这里非常容易写成df['label'],会报错
#同Series使用了绝对位置切片
#如果想要获得某一个元素,可用如下get_loc方法:
SchoolClassGenderAddressHeightWeightMathPhysics
ID
1102S_1C_1Fstreet_21927332.5B+
row = df.index.get_loc(1102)
df[row:row+1]
SchoolClassGenderAddressHeightWeightMathPhysics
ID
1102S_1C_1Fstreet_21927332.5B+
② 多行索引:
#用切片,如果是选取指定的某几行,推荐使用loc,否则很可能报错
df[3:5]
SchoolClassGenderAddressHeightWeightMathPhysics
ID
1104S_1C_1Fstreet_21678180.4B-
1105S_1C_1Fstreet_41596484.8B+
③ 单列索引:
df['School'].head()
ID
1101    S_1
1102    S_1
1103    S_1
1104    S_1
1105    S_1
Name: School, dtype: object
④ 多列索引:
df[['School','Math']].head()
SchoolMath
ID
1101S_134.0
1102S_132.5
1103S_187.2
1104S_180.4
1105S_184.8
⑤函数式索引:
df[lambda x:['Math','Physics']].head()
MathPhysics
ID
110134.0A+
110232.5B+
110387.2B+
110480.4B-
110584.8B+
⑥ 布尔索引:
df[df['Gender']=='F'].head()
SchoolClassGenderAddressHeightWeightMathPhysics
ID
1102S_1C_1Fstreet_21927332.5B+
1104S_1C_1Fstreet_21678180.4B-
1105S_1C_1Fstreet_41596484.8B+
1202S_1C_2Fstreet_41769463.5B-
1204S_1C_2Fstreet_51626333.8B
小节:一般来说,[]操作符常用于列选择或布尔选择,尽量避免行的选择

2. 布尔索引

(a)布尔符号:’&’,’|’,’~’:分别代表和and,或or,取反not
df[(df['Gender']=='F')&(df['Address']=='street_2')].head()
SchoolClassGenderAddressHeightWeightMathPhysics
ID
1102S_1C_1Fstreet_21927332.5B+
1104S_1C_1Fstreet_21678180.4B-
2401S_2C_4Fstreet_21926245.3A
2404S_2C_4Fstreet_21608467.7B
df[(df['Math']>85)|(df['Address']=='street_7')].head()
SchoolClassGenderAddressHeightWeightMathPhysics
ID
1103S_1C_1Mstreet_21868287.2B+
1201S_1C_2Mstreet_51886897.0A-
1302S_1C_3Fstreet_11755787.7A-
1303S_1C_3Mstreet_71888249.7B
1304S_1C_3Mstreet_21957085.2A
df[~((df['Math']>75)|(df['Address']=='street_1'))].head()
SchoolClassGenderAddressHeightWeightMathPhysics
ID
1102S_1C_1Fstreet_21927332.5B+
1202S_1C_2Fstreet_41769463.5B-
1203S_1C_2Mstreet_61605358.8A+
1204S_1C_2Fstreet_51626333.8B
1205S_1C_2Fstreet_61676368.4B-
loc和[]中相应位置都能使用布尔列表选择:
df.loc[df['Math']>60,(df[:8]['Address']=='street_6').values].head()
#如果不加values就会索引对齐发生错误,Pandas中的索引对齐是一个重要特征,很多时候非常使用
#但是若不加以留意,就会埋下隐患
Physics
ID
1103B+
1104B-
1105B+
1201A-
1202B-
(df[:8]['Address']=='street_6').values
array([False, False, False, False, False, False, False,  True])
(b) isin方法
df[df['Address'].isin(['street_1','street_4'])&df['Physics'].isin(['A','A+'])]
SchoolClassGenderAddressHeightWeightMathPhysics
ID
1101S_1C_1Mstreet_11736334.0A+
2105S_2C_1Mstreet_41708134.2A
2203S_2C_2Mstreet_41559173.8A+
#上面也可以用字典方式写:
df[df[['Address','Physics']].isin({'Address':['street_1','street_4'],'Physics':['A','A+']}).all(1)]
#all与&的思路是类似的,其中的1代表按照跨列方向判断是否全为True
SchoolClassGenderAddressHeightWeightMathPhysics
ID
1101S_1C_1Mstreet_11736334.0A+
2105S_2C_1Mstreet_41708134.2A
2203S_2C_2Mstreet_41559173.8A+

3. 快速标量索引

当只需要取一个元素时,at和iat方法能够提供更快的实现:
display(df.at[1101,'School'])
display(df.loc[1101,'School'])
display(df.iat[0,0])
display(df.iloc[0,0])
#可尝试去掉注释对比时间
#%timeit df.at[1101,'School']
#%timeit df.loc[1101,'School']
#%timeit df.iat[0,0]
#%timeit df.iloc[0,0]
'S_1'



'S_1'



'S_1'



'S_1'

4. 区间索引

此处介绍并不是说只能在单级索引中使用区间索引,只是作为一种特殊类型的索引方式,在此处先行介绍
(a)利用interval_range方法
pd.interval_range(start=0,end=5)
#closed参数可选'left''right''both''neither',默认左开右闭
IntervalIndex([(0, 1], (1, 2], (2, 3], (3, 4], (4, 5]],
              closed='right',
              dtype='interval[int64]')
pd.interval_range(start=0,periods=8,freq=5)
#periods参数控制区间个数,freq控制步长
IntervalIndex([(0, 5], (5, 10], (10, 15], (15, 20], (20, 25], (25, 30], (30, 35], (35, 40]],
              closed='right',
              dtype='interval[int64]')
(b)利用cut将数值列转为区间为元素的分类变量,例如统计数学成绩的区间情况:
math_interval = pd.cut(df['Math'],bins=[0,40,60,80,100])
#注意,如果没有类型转换,此时并不是区间类型,而是category类型
math_interval.head()
ID
1101      (0, 40]
1102      (0, 40]
1103    (80, 100]
1104    (80, 100]
1105    (80, 100]
Name: Math, dtype: category
Categories (4, interval[int64]): [(0, 40] < (40, 60] < (60, 80] < (80, 100]]
(c)区间索引的选取
df_i = df.join(math_interval,rsuffix='_interval')[['Math','Math_interval']]\
            .reset_index().set_index('Math_interval')
df_i.loc[65]
IDMath
Math_interval
(60, 80]120263.5
(60, 80]120568.4
(60, 80]130561.7
(60, 80]210472.2
(60, 80]220268.5
(60, 80]220373.8
(60, 80]230172.3
(60, 80]230365.9
(60, 80]240467.7
df_i.loc[65].head()
#包含该值就会被选中
IDMath
Math_interval
(60, 80]120263.5
(60, 80]120568.4
(60, 80]130561.7
(60, 80]210472.2
(60, 80]220268.5
df_i.loc[[62,100]].head()
IDMath
Math_interval
(60, 80]120263.5
(60, 80]120568.4
(60, 80]130561.7
(60, 80]210472.2
(60, 80]220268.5
如果想要选取某个区间,先要把分类变量转为区间变量,再使用overlap方法:
#df_i.loc[pd.Interval(70,75)].head() 报错
df_i[df_i.index.astype('interval').overlaps(pd.Interval(70, 85))].head()
IDMath
Math_interval
(80, 100]110387.2
(80, 100]110480.4
(80, 100]110584.8
(80, 100]120197.0
(60, 80]120263.5

二、多级索引

1. 创建多级索引

(a)通过from_tuple或from_arrays
① 直接创建元组
tuples = [('A','a'),('A','b'),('B','a'),('B','b')]
mul_index = pd.MultiIndex.from_tuples(tuples, names=('Upper', 'Lower'))
pd.DataFrame({'Score':['perfect','good','fair','bad']},index=mul_index)
mul_index
MultiIndex([('A', 'a'),
            ('A', 'b'),
            ('B', 'a'),
            ('B', 'b')],
           names=['Upper', 'Lower'])
pd.DataFrame({'Score':['perfect','good','fair','bad']},index=mul_index)
Score
UpperLower
Aaperfect
bgood
Bafair
bbad
② 利用zip创建元组
L1 = list('AABB')
L2 = list('abab')
tuples = list(zip(L1,L2))
mul_index = pd.MultiIndex.from_tuples(tuples, names=('Upper', 'Lower'))
pd.DataFrame({'Score':['perfect','good','fair','bad']},index=mul_index)
Score
UpperLower
Aaperfect
bgood
Bafair
bbad
③ 通过Array创建
arrays = [['A','a'],['A','b'],['B','a'],['B','b']]
mul_index = pd.MultiIndex.from_tuples(arrays, names=('Upper', 'Lower'))
pd.DataFrame({'Score':['perfect','good','fair','bad']},index=mul_index)
Score
UpperLower
Aaperfect
bgood
Bafair
bbad
mul_index
#由此看出内部自动转成元组
MultiIndex([('A', 'a'),
            ('A', 'b'),
            ('B', 'a'),
            ('B', 'b')],
           names=['Upper', 'Lower'])
(b)通过from_product
L1 = ['A','B']
L2 = ['a','b']
pd.MultiIndex.from_product([L1,L2],names=('Upper', 'Lower'))
#两两相乘
MultiIndex([('A', 'a'),
            ('A', 'b'),
            ('B', 'a'),
            ('B', 'b')],
           names=['Upper', 'Lower'])
(c)指定df中的列创建(set_index方法)
df_using_mul = df.set_index(['Class','Address'])
df_using_mul.head()
SchoolGenderHeightWeightMathPhysics
ClassAddress
C_1street_1S_1M1736334.0A+
street_2S_1F1927332.5B+
street_2S_1M1868287.2B+
street_2S_1F1678180.4B-
street_4S_1F1596484.8B+

2. 多层索引切片

df_using_mul.head()
SchoolGenderHeightWeightMathPhysics
ClassAddress
C_1street_1S_1M1736334.0A+
street_2S_1F1927332.5B+
street_2S_1M1868287.2B+
street_2S_1F1678180.4B-
street_4S_1F1596484.8B+
(a)一般切片
#df_using_mul.loc['C_2','street_5']
#当索引不排序时,单个索引会报出性能警告
#df_using_mul.index.is_lexsorted()
#该函数检查是否排序
df_using_mul.sort_index().loc['C_2','street_5']
#df_using_mul.sort_index().index.is_lexsorted()
SchoolGenderHeightWeightMathPhysics
ClassAddress
C_2street_5S_1M1886897.0A-
street_5S_1F1626333.8B
street_5S_2M19310039.1B
#df_using_mul.loc[('C_2','street_5'):] 报错
#当不排序时,不能使用多层切片
df_using_mul.sort_index().loc[('C_2','street_6'):('C_3','street_4')]
#注意此处由于使用了loc,因此仍然包含右端点
SchoolGenderHeightWeightMathPhysics
ClassAddress
C_2street_6S_1M1605358.8A+
street_6S_1F1676368.4B-
street_7S_2F1947768.5B+
street_7S_2F1837685.4B
C_3street_1S_1F1755787.7A-
street_2S_1M1957085.2A
street_4S_1M1616831.5B+
street_4S_2F1577872.3B+
street_4S_2M1877348.9B
df_using_mul.sort_index().loc[('C_2','street_7'):'C_3'].head()
#非元组也是合法的,表示选中该层所有元素
SchoolGenderHeightWeightMathPhysics
ClassAddress
C_2street_7S_2F1947768.5B+
street_7S_2F1837685.4B
C_3street_1S_1F1755787.7A-
street_2S_1M1957085.2A
street_4S_1M1616831.5B+
(b)第一类特殊情况:由元组构成列表
df_using_mul.sort_index().loc[[('C_2','street_7'),('C_3','street_2')]]
#表示选出某几个元素,精确到最内层索引
SchoolGenderHeightWeightMathPhysics
ClassAddress
C_2street_7S_2F1947768.5B+
street_7S_2F1837685.4B
C_3street_2S_1M1957085.2A
(c)第二类特殊情况:由列表构成元组
df_using_mul.sort_index().loc[(['C_2','C_3'],['street_4','street_7']),:]
#选出第一层在‘C_2’和'C_3'中且第二层在'street_4'和'street_7'中的行
SchoolGenderHeightWeightMathPhysics
ClassAddress
C_2street_4S_1F1769463.5B-
street_4S_2M1559173.8A+
street_7S_2F1947768.5B+
street_7S_2F1837685.4B
C_3street_4S_1M1616831.5B+
street_4S_2F1577872.3B+
street_4S_2M1877348.9B
street_7S_1M1888249.7B
street_7S_2F1909965.9C

3. 多层索引中的slice对象

L1,L2 = ['A','B','C'],['a','b','c']
mul_index1 = pd.MultiIndex.from_product([L1,L2],names=('Upper', 'Lower'))
L3,L4 = ['D','E','F'],['d','e','f']
mul_index2 = pd.MultiIndex.from_product([L3,L4],names=('Big', 'Small'))
df_s = pd.DataFrame(np.random.rand(9,9),index=mul_index1,columns=mul_index2)
df_s
BigDEF
Smalldefdefdef
UpperLower
Aa0.2761530.6978980.7516560.2210450.1176730.9844140.3870380.7197340.133291
b0.4518890.3333690.5256600.0521050.3341030.4629620.6953500.8755610.725070
c0.0703810.6630480.7034770.5937160.6408880.3207370.3801830.4332790.604086
Ba0.0131780.4930220.6227610.9257220.6771080.5314210.5020580.3701250.225989
b0.1963500.0824960.6956590.7950740.5815510.1300790.6826630.1939280.538473
c0.7289200.3448530.0293920.6469190.7605910.2329130.3603840.3360170.168119
Ca0.5692580.4720300.9878520.7478450.4665300.0513270.7640400.5012300.795936
b0.1316300.9920520.0314070.0148750.5395000.3569340.8391980.2884740.692343
c0.1933560.0831510.1924110.4814810.7727510.1464190.9267590.2399960.147642
idx=pd.IndexSlice
索引Slice的使用非常灵活:
df_s.loc[idx['B':,df_s['D']['d']>0.3],idx[df_s.sum()>4]]
#df_s.sum()默认为对列求和,因此返回一个长度为9的数值列表
BigDEF
Smallefdedf
UpperLower
Bc0.3448530.0293920.6469190.7605910.3603840.168119
Ca0.4720300.9878520.7478450.4665300.7640400.795936

4. 索引层的交换

(a)swaplevel方法(两层交换)
df_using_mul.head()
SchoolGenderHeightWeightMathPhysics
ClassAddress
C_1street_1S_1M1736334.0A+
street_2S_1F1927332.5B+
street_2S_1M1868287.2B+
street_2S_1F1678180.4B-
street_4S_1F1596484.8B+
df_using_mul.swaplevel(i=1,j=0,axis=0).sort_index().head()
SchoolGenderHeightWeightMathPhysics
AddressClass
street_1C_1S_1M1736334.0A+
C_2S_2M1757447.2B-
C_3S_1F1755787.7A-
street_2C_1S_1F1927332.5B+
C_1S_1M1868287.2B+
(b)reorder_levels方法(多层交换)
df_muls = df.set_index(['School','Class','Address'])
df_muls.head()
GenderHeightWeightMathPhysics
SchoolClassAddress
S_1C_1street_1M1736334.0A+
street_2F1927332.5B+
street_2M1868287.2B+
street_2F1678180.4B-
street_4F1596484.8B+
df_muls.reorder_levels([2,0,1],axis=0).sort_index().head()
GenderHeightWeightMathPhysics
AddressSchoolClass
street_1S_1C_1M1736334.0A+
C_3F1755787.7A-
S_2C_2M1757447.2B-
street_2S_1C_1F1927332.5B+
C_1M1868287.2B+
#如果索引有name,可以直接使用name
df_muls.reorder_levels(['Address','School','Class'],axis=0).sort_index().head()
GenderHeightWeightMathPhysics
AddressSchoolClass
street_1S_1C_1M1736334.0A+
C_3F1755787.7A-
S_2C_2M1757447.2B-
street_2S_1C_1F1927332.5B+
C_1M1868287.2B+

三、索引设定

1. index_col参数

index_col是read_csv中的一个参数,而不是某一个方法:
pd.read_csv('data/table.csv',index_col=['Address','School']).head()
ClassIDGenderHeightWeightMathPhysics
AddressSchool
street_1S_1C_11101M1736334.0A+
street_2S_1C_11102F1927332.5B+
S_1C_11103M1868287.2B+
S_1C_11104F1678180.4B-
street_4S_1C_11105F1596484.8B+

2. reindex和reindex_like

reindex是指重新索引,它的重要特性在于索引对齐,很多时候用于重新排序
df.head()
SchoolClassGenderAddressHeightWeightMathPhysics
ID
1101S_1C_1Mstreet_11736334.0A+
1102S_1C_1Fstreet_21927332.5B+
1103S_1C_1Mstreet_21868287.2B+
1104S_1C_1Fstreet_21678180.4B-
1105S_1C_1Fstreet_41596484.8B+
df.reindex(index=[1101,1203,1206,2402])
SchoolClassGenderAddressHeightWeightMathPhysics
ID
1101S_1C_1Mstreet_1173.063.034.0A+
1203S_1C_2Mstreet_6160.053.058.8A+
1206NaNNaNNaNNaNNaNNaNNaNNaN
2402S_2C_4Mstreet_7166.082.048.7B
df.reindex(columns=['Height','Gender','Average']).head()
HeightGenderAverage
ID
1101173MNaN
1102192FNaN
1103186MNaN
1104167FNaN
1105159FNaN
可以选择缺失值的填充方法:fill_value和method(bfill/ffill/nearest),其中method参数必须索引单调
df.reindex(index=[1101,1203,1206,2402],method='bfill')
#bfill表示用所在索引1206的后一个有效行填充,ffill为前一个有效行,nearest是指最近的
SchoolClassGenderAddressHeightWeightMathPhysics
ID
1101S_1C_1Mstreet_11736334.0A+
1203S_1C_2Mstreet_61605358.8A+
1206S_1C_3Mstreet_41616831.5B+
2402S_2C_4Mstreet_71668248.7B
df.reindex(index=[1101,1203,1206,2402],method='nearest')
#数值上1205比1301更接近1206,因此用前者填充
SchoolClassGenderAddressHeightWeightMathPhysics
ID
1101S_1C_1Mstreet_11736334.0A+
1203S_1C_2Mstreet_61605358.8A+
1206S_1C_2Fstreet_61676368.4B-
2402S_2C_4Mstreet_71668248.7B
reindex_like的作用为生成一个横纵索引完全与参数列表一致的DataFrame,数据使用被调用的表
df_temp = pd.DataFrame({'Weight':np.zeros(5),
                        'Height':np.zeros(5),
                        'ID':[1101,1104,1103,1106,1102]}).set_index('ID')
df_temp.reindex_like(df[0:5][['Weight','Height']])
WeightHeight
ID
11010.00.0
11020.00.0
11030.00.0
11040.00.0
1105NaNNaN
如果df_temp单调还可以使用method参数:
df_temp = pd.DataFrame({'Weight':range(5),
                        'Height':range(5),
                        'ID':[1101,1104,1103,1106,1102]}).set_index('ID').sort_index()
df_temp.reindex_like(df[0:5][['Weight','Height']],method='bfill')
#可以自行检验这里的1105的值是否是由bfill规则填充
WeightHeight
ID
110100
110244
110322
110411
110533

3. set_index和reset_index

先介绍set_index:从字面意思看,就是将某些列作为索引
使用表内列作为索引:
df.head()
SchoolClassGenderAddressHeightWeightMathPhysics
ID
1101S_1C_1Mstreet_11736334.0A+
1102S_1C_1Fstreet_21927332.5B+
1103S_1C_1Mstreet_21868287.2B+
1104S_1C_1Fstreet_21678180.4B-
1105S_1C_1Fstreet_41596484.8B+
df.set_index('Class').head()
SchoolGenderAddressHeightWeightMathPhysics
Class
C_1S_1Mstreet_11736334.0A+
C_1S_1Fstreet_21927332.5B+
C_1S_1Mstreet_21868287.2B+
C_1S_1Fstreet_21678180.4B-
C_1S_1Fstreet_41596484.8B+
利用append参数可以将当前索引维持不变
df.set_index('Class',append=True).head()
SchoolGenderAddressHeightWeightMathPhysics
IDClass
1101C_1S_1Mstreet_11736334.0A+
1102C_1S_1Fstreet_21927332.5B+
1103C_1S_1Mstreet_21868287.2B+
1104C_1S_1Fstreet_21678180.4B-
1105C_1S_1Fstreet_41596484.8B+
当使用与表长相同的列作为索引(需要先转化为Series,否则报错):
df.set_index(pd.Series(range(df.shape[0]))).head()
SchoolClassGenderAddressHeightWeightMathPhysics
0S_1C_1Mstreet_11736334.0A+
1S_1C_1Fstreet_21927332.5B+
2S_1C_1Mstreet_21868287.2B+
3S_1C_1Fstreet_21678180.4B-
4S_1C_1Fstreet_41596484.8B+
可以直接添加多级索引:
df.set_index([pd.Series(range(df.shape[0])),pd.Series(np.ones(df.shape[0]))]).head()
SchoolClassGenderAddressHeightWeightMathPhysics
01.0S_1C_1Mstreet_11736334.0A+
11.0S_1C_1Fstreet_21927332.5B+
21.0S_1C_1Mstreet_21868287.2B+
31.0S_1C_1Fstreet_21678180.4B-
41.0S_1C_1Fstreet_41596484.8B+
下面介绍reset_index方法,它的主要功能是将索引重置
默认状态直接恢复到自然数索引:
df.reset_index().head()
IDSchoolClassGenderAddressHeightWeightMathPhysics
01101S_1C_1Mstreet_11736334.0A+
11102S_1C_1Fstreet_21927332.5B+
21103S_1C_1Mstreet_21868287.2B+
31104S_1C_1Fstreet_21678180.4B-
41105S_1C_1Fstreet_41596484.8B+
用level参数指定哪一层被reset,用col_level参数指定set到哪一层:
L1,L2 = ['A','B','C'],['a','b','c']
mul_index1 = pd.MultiIndex.from_product([L1,L2],names=('Upper', 'Lower'))
L3,L4 = ['D','E','F'],['d','e','f']
mul_index2 = pd.MultiIndex.from_product([L3,L4],names=('Big', 'Small'))
df_temp = pd.DataFrame(np.random.rand(9,9),index=mul_index1,columns=mul_index2)
df_temp.head()
BigDEF
Smalldefdefdef
UpperLower
Aa0.0363000.1112970.5098190.3220650.1078460.3930020.9516950.3520450.055245
b0.5859760.8177810.5265120.5607630.6471260.8019220.5205110.7089810.774692
c0.8599070.8247120.6753480.2385580.8691720.9653630.8034850.5687710.734648
Ba0.1290400.2782340.9817280.9034230.6762400.3712910.8765710.3383530.021567
b0.2213860.7447650.0801380.0139360.6230340.3148590.5201990.7422330.834835
df_temp1 = df_temp.reset_index(level=1,col_level=1)
df_temp1.head()
BigDEF
SmallLowerdefdefdef
Upper
Aa0.0363000.1112970.5098190.3220650.1078460.3930020.9516950.3520450.055245
Ab0.5859760.8177810.5265120.5607630.6471260.8019220.5205110.7089810.774692
Ac0.8599070.8247120.6753480.2385580.8691720.9653630.8034850.5687710.734648
Ba0.1290400.2782340.9817280.9034230.6762400.3712910.8765710.3383530.021567
Bb0.2213860.7447650.0801380.0139360.6230340.3148590.5201990.7422330.834835
df_temp1.columns
#看到的确插入了level2
MultiIndex([( '', 'Lower'),
            ('D',     'd'),
            ('D',     'e'),
            ('D',     'f'),
            ('E',     'd'),
            ('E',     'e'),
            ('E',     'f'),
            ('F',     'd'),
            ('F',     'e'),
            ('F',     'f')],
           names=['Big', 'Small'])
df_temp1.index
#最内层索引被移出
Index(['A', 'A', 'A', 'B', 'B', 'B', 'C', 'C', 'C'], dtype='object', name='Upper')

4. rename_axis和rename

rename_axis是针对多级索引的方法,作用是修改某一层的索引名,而不是索引标签
df_temp.rename_axis(index={'Lower':'LowerLower'},columns={'Big':'BigBig'})
BigBigDEF
Smalldefdefdef
UpperLowerLower
Aa0.0363000.1112970.5098190.3220650.1078460.3930020.9516950.3520450.055245
b0.5859760.8177810.5265120.5607630.6471260.8019220.5205110.7089810.774692
c0.8599070.8247120.6753480.2385580.8691720.9653630.8034850.5687710.734648
Ba0.1290400.2782340.9817280.9034230.6762400.3712910.8765710.3383530.021567
b0.2213860.7447650.0801380.0139360.6230340.3148590.5201990.7422330.834835
c0.9052520.0375120.3778490.2258820.7811820.2325110.7785180.5727870.858842
Ca0.6782890.3246380.1656280.7370360.5919360.3121730.3196890.2890720.954463
b0.6478610.5273870.5059450.4883350.6310820.6395390.7850940.0260730.881210
c0.9907220.6917150.6976170.1158310.1298020.9941520.1768410.4490530.145351
rename方法用于修改列或者行索引标签,而不是索引名:
df_temp.rename(index={'A':'T'},columns={'e':'changed_e'}).head()
BigDEF
Smalldchanged_efdchanged_efdchanged_ef
UpperLower
Ta0.0363000.1112970.5098190.3220650.1078460.3930020.9516950.3520450.055245
b0.5859760.8177810.5265120.5607630.6471260.8019220.5205110.7089810.774692
c0.8599070.8247120.6753480.2385580.8691720.9653630.8034850.5687710.734648
Ba0.1290400.2782340.9817280.9034230.6762400.3712910.8765710.3383530.021567
b0.2213860.7447650.0801380.0139360.6230340.3148590.5201990.7422330.834835

四、常用索引型函数

1. where函数

当对条件为False的单元进行填充:
df.head()
SchoolClassGenderAddressHeightWeightMathPhysics
ID
1101S_1C_1Mstreet_11736334.0A+
1102S_1C_1Fstreet_21927332.5B+
1103S_1C_1Mstreet_21868287.2B+
1104S_1C_1Fstreet_21678180.4B-
1105S_1C_1Fstreet_41596484.8B+
df.where(df['Gender']=='M').head()
#不满足条件的行全部被设置为NaN
SchoolClassGenderAddressHeightWeightMathPhysics
ID
1101S_1C_1Mstreet_1173.063.034.0A+
1102NaNNaNNaNNaNNaNNaNNaNNaN
1103S_1C_1Mstreet_2186.082.087.2B+
1104NaNNaNNaNNaNNaNNaNNaNNaN
1105NaNNaNNaNNaNNaNNaNNaNNaN
通过这种方法筛选结果和[]操作符的结果完全一致:
df.where(df['Gender']=='M').dropna().head()
SchoolClassGenderAddressHeightWeightMathPhysics
ID
1101S_1C_1Mstreet_1173.063.034.0A+
1103S_1C_1Mstreet_2186.082.087.2B+
1201S_1C_2Mstreet_5188.068.097.0A-
1203S_1C_2Mstreet_6160.053.058.8A+
1301S_1C_3Mstreet_4161.068.031.5B+
第一个参数为布尔条件,第二个参数为填充值:
df.where(df['Gender']=='M',np.random.rand(df.shape[0],df.shape[1])).head()
SchoolClassGenderAddressHeightWeightMathPhysics
ID
1101S_1C_1Mstreet_1173.00000063.00000034.000000A+
11020.01524670.7084440.9171990.3021850.6896430.0101260.7246360.895387
1103S_1C_1Mstreet_2186.00000082.00000087.200000B+
11040.3691950.4592110.4641910.9644860.3657970.1276020.5014960.0287754
11050.8122320.9996340.8257820.2856920.3401970.0839820.7923100.133054

2. mask函数

mask函数与where功能上相反,其余完全一致,即对条件为True的单元进行填充
df.mask(df['Gender']=='M').dropna().head()
SchoolClassGenderAddressHeightWeightMathPhysics
ID
1102S_1C_1Fstreet_2192.073.032.5B+
1104S_1C_1Fstreet_2167.081.080.4B-
1105S_1C_1Fstreet_4159.064.084.8B+
1202S_1C_2Fstreet_4176.094.063.5B-
1204S_1C_2Fstreet_5162.063.033.8B
df.mask(df['Gender']=='M',np.random.rand(df.shape[0],df.shape[1])).head()
SchoolClassGenderAddressHeightWeightMathPhysics
ID
11010.2739620.250280.5874710.9772060.4424030.3194600.4609910.842498
1102S_1C_1Fstreet_2192.00000073.00000032.500000B+
11030.4366740.7415240.469960.6886030.9382410.5318110.7943520.17495
1104S_1C_1Fstreet_2167.00000081.00000080.400000B-
1105S_1C_1Fstreet_4159.00000064.00000084.800000B+

3. query函数

df.head()
SchoolClassGenderAddressHeightWeightMathPhysics
ID
1101S_1C_1Mstreet_11736334.0A+
1102S_1C_1Fstreet_21927332.5B+
1103S_1C_1Mstreet_21868287.2B+
1104S_1C_1Fstreet_21678180.4B-
1105S_1C_1Fstreet_41596484.8B+
query函数中的布尔表达式中,下面的符号都是合法的:行列索引名、字符串、and/not/or/&/|/~/not in/in/==/!=、四则运算符
df.query('(Address in ["street_6","street_7"])&(Weight>(70+10))&(ID in [1303,2304,2402])')
SchoolClassGenderAddressHeightWeightMathPhysics
ID
1303S_1C_3Mstreet_71888249.7B
2304S_2C_3Fstreet_61648195.5A-
2402S_2C_4Mstreet_71668248.7B

五、重复元素处理

1. duplicated方法

该方法返回了是否重复的布尔列表
df.duplicated('Class').head()
ID
1101    False
1102     True
1103     True
1104     True
1105     True
dtype: bool
df.loc[1105:1101:-1]
SchoolClassGenderAddressHeightWeightMathPhysics
ID
1105S_1C_1Fstreet_41596484.8B+
1104S_1C_1Fstreet_21678180.4B-
1103S_1C_1Mstreet_21868287.2B+
1102S_1C_1Fstreet_21927332.5B+
1101S_1C_1Mstreet_11736334.0A+
可选参数keep默认为first,即首次出现设为不重复,若为last,则最后一次设为不重复,若为False,则所有重复项为False
df.duplicated('Class',keep='last').tail()
ID
2401     True
2402     True
2403     True
2404     True
2405    False
dtype: bool
df.duplicated('Class',keep=False).head()
ID
1101    True
1102    True
1103    True
1104    True
1105    True
dtype: bool

2. drop_duplicates方法

从名字上看出为剔除重复项,这在后面章节中的分组操作中可能是有用的,例如需要保留每组的第一个值:
df.drop_duplicates('Class')
SchoolClassGenderAddressHeightWeightMathPhysics
ID
1101S_1C_1Mstreet_11736334.0A+
1201S_1C_2Mstreet_51886897.0A-
1301S_1C_3Mstreet_41616831.5B+
2401S_2C_4Fstreet_21926245.3A
参数与duplicate函数类似:
df.drop_duplicates('Class',keep='last')
SchoolClassGenderAddressHeightWeightMathPhysics
ID
2105S_2C_1Mstreet_41708134.2A
2205S_2C_2Fstreet_71837685.4B
2305S_2C_3Mstreet_41877348.9B
2405S_2C_4Fstreet_61935447.6B
在传入多列时等价于将多列共同视作一个多级索引,比较重复项:
df.drop_duplicates(['School','Class'])
SchoolClassGenderAddressHeightWeightMathPhysics
ID
1101S_1C_1Mstreet_11736334.0A+
1201S_1C_2Mstreet_51886897.0A-
1301S_1C_3Mstreet_41616831.5B+
2101S_2C_1Mstreet_71748483.3C
2201S_2C_2Mstreet_519310039.1B
2301S_2C_3Fstreet_41577872.3B+
2401S_2C_4Fstreet_21926245.3A

六、抽样函数

这里的抽样函数指的就是sample函数
(a)n为样本量
df.sample(n=5)
SchoolClassGenderAddressHeightWeightMathPhysics
ID
2103S_2C_1Mstreet_41576152.5B-
1102S_1C_1Fstreet_21927332.5B+
1301S_1C_3Mstreet_41616831.5B+
1304S_1C_3Mstreet_21957085.2A
1105S_1C_1Fstreet_41596484.8B+
(b)frac为抽样比
df.sample(frac=0.05)
SchoolClassGenderAddressHeightWeightMathPhysics
ID
1105S_1C_1Fstreet_41596484.8B+
2402S_2C_4Mstreet_71668248.7B
(c)replace为是否放回
df.sample(n=df.shape[0],replace=True).head()
SchoolClassGenderAddressHeightWeightMathPhysics
ID
2403S_2C_4Fstreet_61586059.7B+
2404S_2C_4Fstreet_21608467.7B
2405S_2C_4Fstreet_61935447.6B
2303S_2C_3Fstreet_71909965.9C
1203S_1C_2Mstreet_61605358.8A+
df.sample(n=35,replace=True).index.is_unique
False
(d)axis为抽样维度,默认为0,即抽行
df.sample(n=3,axis=1).head()
AddressWeightSchool
ID
1101street_163S_1
1102street_273S_1
1103street_282S_1
1104street_281S_1
1105street_464S_1
(e)weights为样本权重,自动归一化
df.sample(n=3,weights=np.random.rand(df.shape[0])).head()
SchoolClassGenderAddressHeightWeightMathPhysics
ID
1302S_1C_3Fstreet_11755787.7A-
1305S_1C_3Fstreet_51876961.7B-
2404S_2C_4Fstreet_21608467.7B
#以某一列为权重,这在抽样理论中很常见
df.sample(n=3,weights=df['Math']).head()
SchoolClassGenderAddressHeightWeightMathPhysics
ID
1305S_1C_3Fstreet_51876961.7B-
2103S_2C_1Mstreet_41576152.5B-
2105S_2C_1Mstreet_41708134.2A
df.head()
SchoolClassGenderAddressHeightWeightMathPhysics
ID
1101S_1C_1Mstreet_11736334.0A+
1102S_1C_1Fstreet_21927332.5B+
1103S_1C_1Mstreet_21868287.2B+
1104S_1C_1Fstreet_21678180.4B-
1105S_1C_1Fstreet_41596484.8B+
df.iloc[1::2]
SchoolClassGenderAddressHeightWeightMathPhysics
ID
1102S_1C_1Fstreet_21927332.5B+
1104S_1C_1Fstreet_21678180.4B-
1201S_1C_2Mstreet_51886897.0A-
1203S_1C_2Mstreet_61605358.8A+
1205S_1C_2Fstreet_61676368.4B-
1302S_1C_3Fstreet_11755787.7A-
1304S_1C_3Mstreet_21957085.2A
2101S_2C_1Mstreet_71748483.3C
2103S_2C_1Mstreet_41576152.5B-
2105S_2C_1Mstreet_41708134.2A
2202S_2C_2Fstreet_71947768.5B+
2204S_2C_2Mstreet_11757447.2B-
2301S_2C_3Fstreet_41577872.3B+
2303S_2C_3Fstreet_71909965.9C
2305S_2C_3Mstreet_41877348.9B
2402S_2C_4Mstreet_71668248.7B
2404S_2C_4Fstreet_21608467.7B

七、问题与练习

1. 问题

【问题一】 如何更改列或行的顺序?如何交换奇偶行(列)的顺序?
更改行或列顺序:df.loc[[索引值排序列表],[列名排序列表]]
交换奇偶行顺序:df.iloc[1::2,:].rename(index=['行1',...])
【问题二】 如果要选出DataFrame的某个子集,请给出尽可能多的方法实现。
loc、iloc、[]、query、布尔值、区间索引等
【问题三】 query函数比其他索引方法的速度更慢吗?在什么场合使用什么索引最高效?
1.使用query,速度最慢
2.把要查询的键用set_index设为index后再用loc切片查询会加快速度,根据StackOverflow上面的回答,pandas内部会对index做优化。
3.groupby把索引hash到一个字典,遍历循环这个字典。

尽管有很多小技巧可以提升速度,总体上pandas的速度比data.table还是要慢
【问题四】 单级索引能使用Slice对象吗?能的话怎么使用,请给出一个例子。
可以。
d=np.array(range(10)).reshape(2,5)
data=pd.DataFrame(d)
data.columns=['a','b']
data['test']=['m1901\t','m1902\t','m1903\t','m1904\t','m1905\t']
data['test']=data['test'].str.slice(1,5)#从第二个开始截取中间4位。    
【问题五】 如何快速找出某一列的缺失值所在索引?
df['列名'].isnull().index
【问题六】 索引设定中的所有方法分别适用于哪些场合?怎么直接把某个DataFrame的索引换成任意给定同长度的索引?
index_col参数:用于设定索引    
reindex主要特性是索引对齐,主要用于重新排序;re_index_like用于填充  
set_index可添加多级索引;reset_index将索引重置为自然数   
rename_axis针对多层索引,修改某一层索引名;rename修改索引标签  
可使用rename方法
【问题七】 多级索引有什么适用场合?
需要多特征组合作为索引时
【问题八】 什么时候需要重复元素处理?
例如需要保留每组第一个值
多级索引比较重复项

2. 练习

【练习一】 现有一份关于UFO的数据集,请解决下列问题:
df = pd.read_csv('data/UFO.csv')
df.head()
datetimeshapeduration (seconds)latitudelongitude
010/10/1949 20:30cylinder2700.029.883056-97.941111
110/10/1949 21:00light7200.029.384210-98.581082
210/10/1955 17:00circle20.053.200000-2.916667
310/10/1956 21:00circle20.028.978333-96.645833
410/10/1960 20:00light900.021.418056-157.803611
(a)在所有被观测时间超过60s的时间中,哪个形状最多?
(b)对经纬度进行划分:-180°至180°以30°为一个划分,-90°至90°以18°为一个划分,请问哪个区域中报告的UFO事件数量最多?
# (a)
df[df['duration (seconds)']>60].groupby('shape')['datetime'].count().idxmax()
'light'
#(b)
df['latidude_c'] = pd.cut(df['latitude'],bins=[x for x in range(-180,180,30)])
df['longitude_c'] = pd.cut(df['longitude'],bins=[x for x in range(-90,90,18)])
df.groupby(['latidude_c','longitude_c'])['datetime'].count().idxmax()

(Interval(30, 60, closed='right'), Interval(-90, -72, closed='right'))
【练习二】 现有一份关于口袋妖怪的数据集,请解决下列问题:
df = pd.read_csv('data/Pokemon.csv')
df.head()
#NameType 1Type 2TotalHPAttackDefenseSp. AtkSp. DefSpeedGenerationLegendary
01BulbasaurGrassPoison3184549496565451False
12IvysaurGrassPoison4056062638080601False
23VenusaurGrassPoison525808283100100801False
33VenusaurMega VenusaurGrassPoison62580100123122120801False
44CharmanderFireNaN3093952436050651False
(a)双属性的Pokemon占总体比例的多少?
(b)在所有种族值(Total)不小于580的Pokemon中,非神兽(Legendary=False)的比例为多少?
(c)在第一属性为格斗系(Fighting)的Pokemon中,物攻排名前三高的是哪些?
(d)请问六项种族指标(HP、物攻、特攻、物防、特防、速度)极差的均值最大的是哪个属性(只考虑第一属性,且均值是对属性而言)?
(e)哪个属性(只考虑第一属性)的神兽比例最高?该属性神兽的种族值也是最高的吗?
#(a)双属性的Pokemon占总体比例的多少?
df[~((df['Type 1'].isnull()) | (df['Type 2'].isnull()) )].shape[0]/df.shape[0]
0.5175
# (b)在所有种族值(Total)不小于580的Pokemon中,非神兽(Legendary=False)的比例为多少?
df1 = df[df['Total']>=580]
df1[df1['Legendary']==False].shape[0]/df1.shape[0]
0.4247787610619469
#(c)在第一属性为格斗系(Fighting)的Pokemon中,物攻排名前三高的是哪些?
df.iloc[df[df['Type 1']=='Fighting']['Attack'].nlargest(3).index,:]
#NameType 1Type 2TotalHPAttackDefenseSp. AtkSp. DefSpeedGenerationLegendary
498448LucarioMega LucarioFightingSteel6257014588140701124False
594534ConkeldurrFightingNaN505105140955565455False
7468MachampFightingNaN50590130806585551False
# (d)请问六项种族指标(HP、物攻、特攻、物防、特防、速度)极差的均值最大的是哪个属性(只考虑第一属性,且均值是对属性而言)?
df['range'] = df.iloc[:,5:11].max(axis=1)-df.iloc[:,5:11].min(axis=1)
attribute = df[['Type 1','range']].set_index('Type 1')
max_range = 0
result = ''
for i in attribute.index.unique():
    temp = attribute.loc[i,:].mean()
    if temp.values[0] > max_range:
        max_range = temp.values[0]
        result = i
result
'Steel'
# (e)哪个属性(只考虑第一属性)的神兽比例最高?该属性神兽的种族值也是最高的吗?
df[df['Legendary']==True].groupby('Type 1')['Name'].count().nlargest(1)
Type 1
Psychic    14
Name: Name, dtype: int64
attribute = df.query('Legendary == True')[['Type 1','Total']].set_index('Type 1')
max_value = 0
result = ''
for i in attribute.index.unique()[:-1]:
    temp = attribute.loc[i,:].mean()
    if temp[0] > max_value:
        max_value = temp[0]
        result = i
result
'Normal'
  • 0
    点赞
  • 0
    收藏
    觉得还不错? 一键收藏
  • 1
    评论

“相关推荐”对你有帮助么?

  • 非常没帮助
  • 没帮助
  • 一般
  • 有帮助
  • 非常有帮助
提交
评论 1
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值