数据编码的十种方式¶

在线刷题

检查 or 强化 Pandas 数据分析操作？👉在线体验「Pandas进阶修炼300题」

Note

在使用Python进行机器学习时，很多算法都需要我们对分类特征进行转换（编码），即根据某一列的值，新增（修改）一列。

这个操作在pandas中也有多种解决方案，本文就将介绍十种方法，代码拿走就用，希望你在遇到不同类型的数据时，可以灵活使用。

下面先创建用于示例的数据👇

import pandas as pd
df = pd.DataFrame({
    "Sex": pd.Series(['Male','Female','Male','Male','Male','Female','Male','Male','Female','Female']),
    "Course Name": pd.Series(['Python','Java','C','Sql','Linux','Python','Python','Java','C','Php']),
    "Score":[95,85,75,65,55,95,75,65,55,85]})

df

	Sex	Course Name	Score
0	Male	Python	95
1	Female	Java	85
2	Male	C	75
3	Male	Sql	65
4	Male	Linux	55
5	Female	Python	95
6	Male	Python	75
7	Male	Java	65
8	Female	C	55
9	Female	Php	85

数值型数据编码¶

1 - 使用自定义函数 + 循环遍历¶

首先然是最简单，最笨的方法，自己写一个函数来转换数据，并用循环遍历，肯定就是一个def加一个for

df1 = df.copy()

def myfun(x):
    if x>90:
        return 'A'
    elif x>=80 and x<90:
        return 'B'
    elif x>=70 and x<80:
        return 'C'
    elif x>=60 and x<70:
        return 'D'
    else:
        return 'E'
    
df1['Score_Label'] = None
for i in range(len(df1)):
    df1.iloc[i,3] = myfun(df1.iloc[i,2])
    
df1

	Sex	Course Name	Score	Score_Label
0	Male	Python	95	A
1	Female	Java	85	B
2	Male	C	75	C
3	Male	Sql	65	D
4	Male	Linux	55	E
5	Female	Python	95	A
6	Male	Python	75	C
7	Male	Java	65	D
8	Female	C	55	E
9	Female	Php	85	B

这段代码，相信所有人都能看懂，简单好想但比较麻烦

有没有更简单的办法呢？pandas当然提供了很多高效的操作的函数，继续往下看。

2 - 使用 map + 自定义函数¶

现在，可以使用map来干掉循环（虽然本质上也是循环）

df2 = df.copy()

def mapfun(x):
    if x>90:
        return 'A'
    elif x>=80 and x<90:
        return 'B'
    elif x>=70 and x<80:
        return 'C'
    elif x>=60 and x<70:
        return 'D'
    else:
        return 'E'

df2['Score_Label'] = df2['Score'].map(mapfun)
df2

	Sex	Course Name	Score	Score_Label
0	Male	Python	95	A
1	Female	Java	85	B
2	Male	C	75	C
3	Male	Sql	65	D
4	Male	Linux	55	E
5	Female	Python	95	A
6	Male	Python	75	C
7	Male	Java	65	D
8	Female	C	55	E
9	Female	Php	85	B

3 - 使用 apply + 匿名函数¶

如果还想简洁代码，可以使用自定义函数 + apply来干掉自定义函数(结果和上面是一致的，只不过这么写容易被打。)

df3 = df.copy()
df3['Score_Label'] = df3['Score'].apply(lambda x: 'A' if x > 90 else (
    'B' if 90 > x >= 80 else ('C' if 80 > x >= 70 else ('D' if 70 > x >= 60 else 'E'))))
df3

	Sex	Course Name	Score	Score_Label
0	Male	Python	95	A
1	Female	Java	85	B
2	Male	C	75	C
3	Male	Sql	65	D
4	Male	Linux	55	E
5	Female	Python	95	A
6	Male	Python	75	C
7	Male	Java	65	D
8	Female	C	55	E
9	Female	Php	85	B

4 - 使用cut¶

现在，让我们继续了解更高级的pandas函数，依旧是对 Score 进行编码，使用pd.cut，并指定划分的区间后，可以直接帮你分好组

df4 = df.copy()
bins = [0, 59, 70, 80, 100]
df4['Score_Label'] = pd.cut(df4['Score'], bins)
df4

	Sex	Course Name	Score	Score_Label
0	Male	Python	95	(80, 100]
1	Female	Java	85	(80, 100]
2	Male	C	75	(70, 80]
3	Male	Sql	65	(59, 70]
4	Male	Linux	55	(0, 59]
5	Female	Python	95	(80, 100]
6	Male	Python	75	(70, 80]
7	Male	Java	65	(59, 70]
8	Female	C	55	(0, 59]
9	Female	Php	85	(80, 100]

也可以直接使用labels参数来修改对应组的名称，是不是方便多了

df4['Score_Label_new'] = pd.cut(df4['Score'], bins, labels=[
                                'low', 'middle', 'good', 'perfect'])
df4

	Sex	Course Name	Score	Score_Label	Score_Label_new
0	Male	Python	95	(80, 100]	perfect
1	Female	Java	85	(80, 100]	perfect
2	Male	C	75	(70, 80]	good
3	Male	Sql	65	(59, 70]	middle
4	Male	Linux	55	(0, 59]	low
5	Female	Python	95	(80, 100]	perfect
6	Male	Python	75	(70, 80]	good
7	Male	Java	65	(59, 70]	middle
8	Female	C	55	(0, 59]	low
9	Female	Php	85	(80, 100]	perfect

5 - 使用 sklearn 二值化¶

既然是和机器学习相关，sklearn肯定跑不掉，如果需要新增一列并判定成绩是否及格，就可以使用Binarizer函数，代码也是简洁好懂

from sklearn.preprocessing import Binarizer

df5 = df.copy()
binerize = Binarizer(threshold = 60)
trans = binerize.fit_transform(np.array(df1['Score']).reshape(-1,1))
df5['Score_Label'] = trans

df5

	Sex	Course Name	Score	Score_Label
0	Male	Python	95	1
1	Female	Java	85	1
2	Male	C	75	1
3	Male	Sql	65	1
4	Male	Linux	55	0
5	Female	Python	95	1
6	Male	Python	75	1
7	Male	Java	65	1
8	Female	C	55	0
9	Female	Php	85	1

文本型数据编码¶

下面介绍更常见的，对文本数据进行转换打标签。例如新增一列，将性别男、女分别标记为0、1

6 - 使用 replace¶

首先介绍replace，但要注意的是，上面说过的自定义函数相关方法依旧是可行的

df6 = df.copy()
df6['Sex_Label'] = df6['Sex'].replace(['Male','Female'],[0,1])
df6

	Sex	Course Name	Score	Sex_Label
0	Male	Python	95	0
1	Female	Java	85	1
2	Male	C	75	0
3	Male	Sql	65	0
4	Male	Linux	55	0
5	Female	Python	95	1
6	Male	Python	75	0
7	Male	Java	65	0
8	Female	C	55	1
9	Female	Php	85	1

上面是对性别操作，因为只有男女，所以可以手动指定0、1，但要是类别很多，也可以使用pd.value_counts()来自动指定标签，例如对Course Name列分组

df6 = df.copy()
value = df6['Course Name'].value_counts()
value_map = dict((v, i) for i,v in enumerate(value.index))
df6['Course Name_Label'] = df6.replace({'Course Name':value_map})['Course Name']
df6

	Sex	Course Name	Score	Course Name_Label
0	Male	Python	95	0
1	Female	Java	85	1
2	Male	C	75	2
3	Male	Sql	65	5
4	Male	Linux	55	3
5	Female	Python	95	0
6	Male	Python	75	0
7	Male	Java	65	1
8	Female	C	55	2
9	Female	Php	85	4

7 - 使用map¶

额外强调的是，新增一列，一定要能够想到map

df7 = df.copy()
Map = {elem:index for index,elem in enumerate(set(df["Course Name"]))}
df7['Course Name_Label'] = df7['Course Name'].map(Map)
df7

	Sex	Course Name	Score	Course Name_Label
0	Male	Python	95	5
1	Female	Java	85	2
2	Male	C	75	0
3	Male	Sql	65	3
4	Male	Linux	55	1
5	Female	Python	95	5
6	Male	Python	75	5
7	Male	Java	65	2
8	Female	C	55	0
9	Female	Php	85	4

8 - 使用astype¶

这个方法应该很多人不知道，这就属于上面提到的知乎问题，能实现的方法太多了

df8 = df.copy()
value = df8['Course Name'].astype('category')
df8['Course Name_Label'] = value.cat.codes
df8

	Sex	Course Name	Score	Course Name_Label
0	Male	Python	95	4
1	Female	Java	85	1
2	Male	C	75	0
3	Male	Sql	65	5
4	Male	Linux	55	2
5	Female	Python	95	4
6	Male	Python	75	4
7	Male	Java	65	1
8	Female	C	55	0
9	Female	Php	85	3

9 - 使用 sklearn¶

同数值型一样，这种机器学习中的经典操作，sklearn一定有办法，使用LabelEncoder可以对分类数据进行编码

from sklearn.preprocessing import LabelEncoder
df9 = df.copy()
le = LabelEncoder()
le.fit(df9['Sex'])
df9['Sex_Label'] = le.transform(df9['Sex'])
le.fit(df9['Course Name'])
df9['Course Name_Label'] = le.transform(df9['Course Name'])
df9

	Sex	Course Name	Score	Sex_Label	Course Name_Label
0	Male	Python	95	1	4
1	Female	Java	85	0	1
2	Male	C	75	1	0
3	Male	Sql	65	1	5
4	Male	Linux	55	1	2
5	Female	Python	95	0	4
6	Male	Python	75	1	4
7	Male	Java	65	1	1
8	Female	C	55	0	0
9	Female	Php	85	0	3

一次性转换两列也是可以的

from sklearn.preprocessing import LabelEncoder
df9 = df.copy()
le = OrdinalEncoder()
le.fit(df9[['Sex','Course Name']])
df9[['Sex_Label','Course Name_Label']] = le.transform(df9[['Sex','Course Name']])

df9

	Sex	Course Name	Score	Sex_Label	Course Name_Label
0	Male	Python	95	1.0	4.0
1	Female	Java	85	0.0	1.0
2	Male	C	75	1.0	0.0
3	Male	Sql	65	1.0	5.0
4	Male	Linux	55	1.0	2.0
5	Female	Python	95	0.0	4.0
6	Male	Python	75	1.0	4.0
7	Male	Java	65	1.0	1.0
8	Female	C	55	0.0	0.0
9	Female	Php	85	0.0	3.0

10 - 使用factorize¶

最后，再介绍一个小众但好用的pandas方法，我们需要注意到，在上面的方法中，自动生成的Course Name_Label列，虽然一个数据对应一个语言，因为避免写自定义函数或者字典，这样可以自动生成，所以大多是无序的。

如果我们希望它是有序的，也就是 Python 对应 0，Java对应1，除了自己指定，还有什么优雅的办法？这时可以使用factorize，它会根据出现顺序进行编码

df10 = df.copy()
df10['Course Name_Label'] = pd.factorize(df10['Course Name'])[0]
df10

	Sex	Course Name	Score	Course Name_Label
0	Male	Python	95	0
1	Female	Java	85	1
2	Male	C	75	2
3	Male	Sql	65	3
4	Male	Linux	55	4
5	Female	Python	95	0
6	Male	Python	75	0
7	Male	Java	65	1
8	Female	C	55	2
9	Female	Php	85	5

结合匿名函数，我们可以做到对多列进行有序编码转换

df10 = df.copy()
cat_columns = df10.select_dtypes(['object']).columns

df10[['Sex_Label', 'Course Name_Label']] = df10[cat_columns].apply(
    lambda x: pd.factorize(x)[0])
df10

	Sex	Course Name	Score	Sex_Label	Course Name_Label
0	Male	Python	95	0	0
1	Female	Java	85	1	1
2	Male	C	75	0	2
3	Male	Sql	65	0	3
4	Male	Linux	55	0	4
5	Female	Python	95	1	0
6	Male	Python	75	0	0
7	Male	Java	65	0	1
8	Female	C	55	1	2
9	Female	Php	85	1	5

注意

你无需完全记住所有方法与细节，只需知道有这么个函数能完成这样操作，需要用时能想到，想到再来查就行。