Random Machine Learning #2

What is Feature Scaling?

During the data pre-processing stage, feature scaling is used to scale the independent variables within the same range (say between 0 and 1 or between -1 and +1).

Why we need to do the feature scaling?

This will ensure no independent variable dominates the machine learning algorithms.

Do you know?

Feature Scaling is not required for all the Machine Learning algorithms. For some algorithms scaling will not have any impact hence we can ignore the feature scaling for them.

For example, Feature Scaling is required for K-Nearest Neighbors and Gradient Descent. Scaling may not be required for Decision Trees and Random Forest.

Question?

How to decide whether feature scaling is required or not? If the algorithm calculates distance, we need feature scaling. Is this correct?

Your comments please.