Normalization transforms numeric values onto a common bounded scale, usually from 0 to 1. It helps models process data consistently when values differ greatly in magnitude or originate from systems using different scales. section 3.3 section 10.5.3
Scaling a set of numbers
Apply Normalized Value = (Value − MIN) / (MAX − MIN) to 2, 4, 5, 6, 20, 56, and 62. With a minimum of 2 and maximum of 62, the values become 0, 0.03, 0.05, 0.07, 0.30, 0.90, and 1 after rounding. The transformation preserves ordering while placing every value within the same range. section 3.3
Table 3.1 Normalization procedure
|
Original number
|
Normalized value
|
|---|---|
|
2
|
(2 – 2)/60 = 0
|
|
4
|
(4 – 2)/60 = 0.03
|
|
5
|
(5 – 2)/60 = 0.05
|
|
6
|
(6 – 2)/60 = 0.07
|
|
20
|
(20 – 2)/60 = 0.3
|
|
56
|
(56 – 2)/60 = 0.9
|
|
62
|
(62 – 2)/60 = 1
|
Use the normalization table to inspect each original number alongside its calculated value. section 3.3
Normalization is especially useful when combining measurements represented differently, such as percentages and scores from 1 to 10. It forms part of data transformation alongside techniques such as clustering. section 10.5.3
