Binning (of a column or 1D Numerical Data)

Equal-frequency partitioning

Suppose a group of 12 sales price records has been sorted as follows:
5, 10, 11, 13, 15, 35, 50, 55, 72, 92, 204, 215

Partition them into three bins by each of the following methods. equal-frequency partitioning

bin 1: 5, 10, 11, 13
bin 2: 15, 35, 50, 55
bin 3: 72, 92, 204, 215

Equal-width partitioning

Suppose a group of 12 sales price records has been sorted as follows:
5, 10, 11, 13, 15, 35, 50, 55, 72, 92, 204, 215
Partition them into three bins by each of the following methods. 

The width of each interval is (215 – 5)/3 = 70.

bin 1: 5, 10, 11, 13, 15, 35, 50, 55, 72
bin 2: 92
bin 3: 204, 215

In [8]:

import pandas as pd
    

In [1]:

l = [5, 10, 11, 13, 15, 35, 50, 55, 72, 92, 204, 215]
    

In [3]:

s = pd.Series(l)
    

In [4]:

s.nunique()
    

Out[4]:

12

In [7]:

pd.cut(s, 4) # Equal width binning
    

Out[7]:

0       (4.79, 57.5]
1       (4.79, 57.5]
2       (4.79, 57.5]
3       (4.79, 57.5]
4       (4.79, 57.5]
5       (4.79, 57.5]
6       (4.79, 57.5]
7       (4.79, 57.5]
8      (57.5, 110.0]
9      (57.5, 110.0]
10    (162.5, 215.0]
11    (162.5, 215.0]
dtype: category
Categories (4, interval[float64, right]): [(4.79, 57.5] < (57.5, 110.0] < (110.0, 162.5] < (162.5, 215.0]]

For ex: if we had 3 bins instead of 4. The bins would look like this:

In [6]:

print(75-4.79)
print(145-75)
print(215-145)

70.21
70
70

In [9]:

pd.qcut(s, 4) # Equal frequency binning
    

Out[9]:

0     (4.999, 12.5]
1     (4.999, 12.5]
2     (4.999, 12.5]
3      (12.5, 42.5]
4      (12.5, 42.5]
5      (12.5, 42.5]
6      (42.5, 77.0]
7      (42.5, 77.0]
8      (42.5, 77.0]
9     (77.0, 215.0]
10    (77.0, 215.0]
11    (77.0, 215.0]
dtype: category
Categories (4, interval[float64, right]): [(4.999, 12.5] < (12.5, 42.5] < (42.5, 77.0] < (77.0, 215.0]]

In [ ]: