3.7 高级处理–数据离散化

• 目标
  • 应用cut、qcut实现数据的区间分组
  • 应用get_dummies实现数据的one-hot编码
• 内容预览
  • 3.7.1 什么是数据的离散化
  • 3.7.2 为什么要离散化
  • 3.7.3 如何实现数据的离散化

3.7.1 什么是数据的离散化

连续属性的离散化就是在连续属性的值域上，将值域划分为若干个离散的区间，最后用不同的符号或整数值代表落在每个子区间中的属性值
[外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传(img-cFUxvyId-1586742259769)(attachment:image.png)]

非离散化数据：

	性别	年龄
A	1	23
B	2	30
C	1	18

非离散化数据：

	物种	毛发
A	1
B	2
C	3

数据的离散化

---

one-hot编码/哑变量：

---

	男	女	年龄
A	1	0	23
B	0	1	30
C	1	0	18

3.7.2 为什么要离散化

连续属性数据的离散化是为了简化数据结构，数据离散化技术可以用来减少给定连续属性值的个数。离散化数据经常作为数据挖掘的工具

3.7.3 如何实现数据的离散化

流程：

1. 对数据进行分组
   • 自动分组：pd.qcut(data, bins) # bins为分组的组数，返回一个Series。
   • 自定义分组：pd.cut(data, []) # []中为设置好的分组区间，返回一个Series。
   • 对数据进行分组一般会与value_counts搭配使用
     • Series.value_counts()：统计分组次数
2. 对分好组的数据求哑变量
   • pd.get_dummies(data, prefix=None)
     • data：array-like, Series or DataFrame
     • prefix：分组名字

# 1）准备数据
import pandas as pd
data = pd.Series([165, 174, 160, 180, 159, 163, 192, 184], index=['NO1:165', 'NO2:174', 'NO3:160', 'NO4:180', 'NO5:159', 'NO6:163', 'NO7:192', 'NO8:184'])
data

NO1:165    165
NO2:174    174
NO3:160    160
NO4:180    180
NO5:159    159
NO6:163    163
NO7:192    192
NO8:184    184
dtype: int64

# 2）分组
# 自动分组
sr = pd.qcut(data, 3)
sr
type(sr)

NO1:165      (163.667, 178.0]
NO2:174      (163.667, 178.0]
NO3:160    (158.999, 163.667]
NO4:180        (178.0, 192.0]
NO5:159    (158.999, 163.667]
NO6:163    (158.999, 163.667]
NO7:192        (178.0, 192.0]
NO8:184        (178.0, 192.0]
dtype: category
Categories (3, interval[float64]): [(158.999, 163.667] < (163.667, 178.0] < (178.0, 192.0]]

pandas.core.series.Series

# 统计分组次数
sr.value_counts()

(178.0, 192.0]        3
(158.999, 163.667]    3
(163.667, 178.0]      2
dtype: int64

# 3）转换成哑变量
pd.get_dummies(sr, prefix='height')

	height_(158.999, 163.667]	height_(163.667, 178.0]	height_(178.0, 192.0]
NO1:165	0	1	0
NO2:174	0	1	0
NO3:160	1	0	0
NO4:180	0	0	1
NO5:159	1	0	0
NO6:163	1	0	0
NO7:192	0	0	1
NO8:184	0	0	1

# 2）分组
# 自定义分组
bins = [150, 165, 180, 195]
sr2 = pd.cut(data, bins)
sr2

NO1:165    (150, 165]
NO2:174    (165, 180]
NO3:160    (150, 165]
NO4:180    (165, 180]
NO5:159    (150, 165]
NO6:163    (150, 165]
NO7:192    (180, 195]
NO8:184    (180, 195]
dtype: category
Categories (3, interval[int64]): [(150, 165] < (165, 180] < (180, 195]]

sr2.value_counts()

(150, 165]    4
(180, 195]    2
(165, 180]    2
dtype: int64

pd.get_dummies(sr2, prefix='身高')

	身高_(150, 165]	身高_(165, 180]	身高_(180, 195]
NO1:165	1	0	0
NO2:174	0	1	0
NO3:160	1	0	0
NO4:180	0	1	0
NO5:159	1	0	0
NO6:163	1	0	0
NO7:192	0	0	1
NO8:184	0	0	1

案例：股票的涨跌幅离散化

# 1）读取数据
import pandas as pd
stock = pd.read_excel('stock.xls')
stock

	trade_date	close	open	high	low	pre_close	change	pct_chg	vol	amount
0	20200313	2887.4265	2804.2322	2910.8812	2799.9841	2923.4856	-36.0591	-1.2334	366450436.0	3.930197e+08
1	20200312	2923.4856	2936.0163	2944.4651	2906.2838	2968.5174	-45.0318	-1.5170	307778457.0	3.282092e+08
2	20200311	2968.5174	3001.7616	3010.0286	2968.5174	2996.7618	-28.2444	-0.9425	352470970.0	3.787666e+08
3	20200310	2996.7618	2918.9347	3000.2963	2904.7989	2943.2907	53.4711	1.8167	393296648.0	4.250172e+08
4	20200309	2943.2907	2987.1805	2989.2051	2940.7138	3034.5113	-91.2206	-3.0061	414560736.0	4.381439e+08
...	...	...	...	...	...	...	...	...	...	...
6997	19910719	136.7000	137.6600	138.5400	136.6600	137.1700	-0.4700	-0.3426	10823.0	5.242826e+03
6998	19910718	137.1700	137.1700	137.1700	135.8100	135.8100	1.3600	1.0014	847.0	4.644160e+02
6999	19910717	135.8100	135.8100	135.8100	135.3900	134.4700	1.3400	0.9965	660.0	3.975240e+02
7000	19910716	134.4700	134.3900	134.4700	133.1400	133.1400	1.3300	0.9989	2796.0	1.328502e+03
7001	19910715	133.1400	133.9000	134.1000	131.8700	132.8000	0.3400	0.2560	11938.0	5.534900e+03

7002 rows × 10 columns

change = stock['change']
change

0      -36.0591
1      -45.0318
2      -28.2444
3       53.4711
4      -91.2206
         ...   
6997    -0.4700
6998     1.3600
6999     1.3400
7000     1.3300
7001     0.3400
Name: change, Length: 7002, dtype: float64

# 2）分组
# 自动分组
sr3 = pd.qcut(change, 10)
sr3

0       (-354.685, -33.319]
1       (-354.685, -33.319]
2         (-33.319, -17.08]
3           (37.551, 649.5]
4       (-354.685, -33.319]
               ...         
6997        (-3.416, 0.934]
6998          (0.934, 4.84]
6999          (0.934, 4.84]
7000          (0.934, 4.84]
7001        (-3.416, 0.934]
Name: change, Length: 7002, dtype: category
Categories (10, interval[float64]): [(-354.685, -33.319] < (-33.319, -17.08] < (-17.08, -9.298] < (-9.298, -3.416] ... (4.84, 10.614] < (10.614, 19.612] < (19.612, 37.551] < (37.551, 649.5]]

sr3.value_counts()

(37.551, 649.5]        701
(0.934, 4.84]          701
(-354.685, -33.319]    701
(19.612, 37.551]       700
(10.614, 19.612]       700
(-3.416, 0.934]        700
(-9.298, -3.416]       700
(-17.08, -9.298]       700
(-33.319, -17.08]      700
(4.84, 10.614]         699
Name: change, dtype: int64

# 3）离散化(获得哑变量/one-hot编码)
stock_change = pd.get_dummies(sr3, prefix='涨跌幅')
stock_change

	涨跌幅_(-354.685, -33.319]	涨跌幅_(-33.319, -17.08]	涨跌幅_(-17.08, -9.298]	涨跌幅_(-9.298, -3.416]	涨跌幅_(-3.416, 0.934]	涨跌幅_(0.934, 4.84]	涨跌幅_(4.84, 10.614]	涨跌幅_(10.614, 19.612]	涨跌幅_(19.612, 37.551]	涨跌幅_(37.551, 649.5]
0	1	0	0	0	0	0	0	0	0	0
1	1	0	0	0	0	0	0	0	0	0
2	0	1	0	0	0	0	0	0	0	0
3	0	0	0	0	0	0	0	0	0	1
4	1	0	0	0	0	0	0	0	0	0
...	...	...	...	...	...	...	...	...	...	...
6997	0	0	0	0	1	0	0	0	0	0
6998	0	0	0	0	0	1	0	0	0	0
6999	0	0	0	0	0	1	0	0	0	0
7000	0	0	0	0	0	1	0	0	0	0
7001	0	0	0	0	1	0	0	0	0	0

7002 rows × 10 columns

# 自定义分组
bins = [-600, -300, 0, 300, 600, 900]
sr = pd.cut(change, bins)
sr

0       (-300, 0]
1       (-300, 0]
2       (-300, 0]
3        (0, 300]
4       (-300, 0]
          ...    
6997    (-300, 0]
6998     (0, 300]
6999     (0, 300]
7000     (0, 300]
7001     (0, 300]
Name: change, Length: 7002, dtype: category
Categories (5, interval[int64]): [(-600, -300] < (-300, 0] < (0, 300] < (300, 600] < (600, 900]]

sr.value_counts()

(0, 300]        3702
(-300, 0]       3290
(-600, -300]       7
(300, 600]         2
(600, 900]         1
Name: change, dtype: int64

stock_change = pd.get_dummies(sr, prefix='涨跌幅') # onr-hot/哑变量
stock_change

	涨跌幅_(-600, -300]	涨跌幅_(-300, 0]	涨跌幅_(0, 300]	涨跌幅_(300, 600]	涨跌幅_(600, 900]
0	0	1	0	0	0
1	0	1	0	0	0
2	0	1	0	0	0
3	0	0	1	0	0
4	0	1	0	0	0
...	...	...	...	...	...
6997	0	1	0	0	0
6998	0	0	1	0	0
6999	0	0	1	0	0
7000	0	0	1	0	0
7001	0	0	1	0	0

7002 rows × 5 columns