Hands-On Data Structures and Algorithms with Python - Second Edition

Feature scaling

The columns in a data frame are known as its features. The rows are known as records or observations. The performance of the machine learning algorithm decreases if one attribute has values in a higher range compared to other attributes' values. Thus, it is often required to scale or normalize the attribute values in a common range.

Consider an example, the following data matrix. This data will be referenced in subsections so please do take note:

data1= ([[  58.,    1.,   43.], [  10.,  200.,   65.], [  20.,   75.,    7.]]

Feature one, with data of 58, 10, and 20, has its values lying between 10 and 58. For feature two, the data lies between 1 and 200. Inconsistent results will be produced if we supply this data to any machine learning algorithm. ...