The process of scaling numeric data to a standard range so that different variables can be compared fairly.
Without normalization, a variable measured in thousands (like income) can dominate a model's learning simply due to scale, even if a variable measured in single digits is actually more predictive.
