利用python将连续指标离散化，分类变量哑元化

 1 #导入模块
 2 import pandas as pd
 3 
 4 # 将范围划分为(0, 18]、(18, 25]、(25, 35]、(35, 60]、(60, 100]
 5 ages = [10, 21, 24, 25, 32, 23, 22, 42, 35, 66, 21]
 6 bins = [0, 18, 25, 35, 60, 100]
 7 cats = pd.cut(ages, bins, right=True)  #right=False表示区间是左闭右开的
 8   cats
 9   [(0, 18], (18, 25], (18, 25], (18, 25], (25, 35], ..., (18, 25],  
10   (35, 60], (25, 35], (60, 100], (18, 25]]
11   Length: 11
12   Categories (5, interval[int64]): [(0, 18] < (18, 25] < (25, 35] < (35, 60] < (60, 100]]
13 #查看ages中不同值对应的标签
14 cats.codes
15 array([0, 1, 1, 1, 2, 1, 1, 3, 2, 4, 1], dtype=int8)                                       
16 #对不同区间的数进行统计
17 cats.value_counts()
18   (0, 18]      1
19   (18, 25]     6
20   (25, 35]     2
21   (35, 60]     1
22   (60, 100]    1 
23   dtype: int64　

 1 import numpy as np
 2 
 3 data = np.random.rand(20)
 4 #cut传入的是区间数量时候，他会根据数据的最小值和最大值计算等长面元
 5 pd.cut(data, 4, precision=2)
 6   [(0.53, 0.76], (0.76, 0.99], (0.068, 0.3], (0.3, 0.53], (0.53, 0.76], ..., (0.76,0.99], (0.76, 0.99], (0.068, 0.3], (0.53, 0.76], (0.068, 0.3]]
 7   Length: 20 
 8   Categories (4, interval[float64]): [(0.068, 0.3] < (0.3, 0.53] < (0.53, 0.76] <(0.76, 0.99]]
 9 
10 #qcut()是pandas中的一个类似于cut的函数，根据样本分位数对数据进行面元划分
11 data = np.random.rand(1000)
12 cats = pd.qcut(data, 4)   #按四分位数切割
13 cats.value_count()
14   (-0.000589, 0.257]    250
15   (0.257, 0.516]        250
16   (0.516, 0.761]        250
17   (0.761, 0.994]        250
18   dtype: int64
19 
20 #自定义分位数
21 cats = pd.qcut(data, [0, 0.1, 0.5, 0.9, 1.])
22 cats.value_counts()
23   (-0.000589, 0.115]    100
24   (0.115, 0.516]        400
25   (0.516, 0.908]        400
26   (0.908, 0.994]        100
27   dtype: int64

1 df = pd.DataFrame({‘key‘:[‘b‘, ‘b‘, ‘a‘, ‘c‘, ‘a‘, ‘b‘],
2                   ‘data1‘:range(6)})
3 dummies = pd.get_dummies(df[‘key‘], prefix=‘key‘)

5 #合并
6 df[[‘data1‘]].join(dummies)
      

 1 df = pd.DataFrame({‘id‘:range(6), 
 2                   ‘key‘:[‘b|a|c‘, ‘a|b|d‘, ‘e|f‘, ‘c‘, ‘a‘, ‘b‘],
 3                   })
 4 df
      

 5 keys_iter = (set(x.split(‘|‘)) for x in df.key)   
 6 keys = sorted(set.union(*keys_iter))
 7 keys   #[‘a‘, ‘b‘, ‘c‘, ‘d‘, ‘e‘, ‘f‘]

 8 dummies = pd.DataFrame(np.zeros((len(df), len(keys))), columns=keys)
 9 dummies

10 for i, key in enumerate(df.key):    #enumerate(df.key)   返回迭代对象的（索引，值）
11 　　dummies.loc[i, key.split(‘|‘)] = 1 
12 dummies
      

13 df.join(dummies.add_prefix("key_"))

1 np.random.seed(12345)
2 values = np.random.randn(10)
3 bins = [0, 0.2, 0.4, 0.6, 0.8, 1]
4 
5 pd.get_dummies(pd.cut(values, bins))