Numerical Data at Scale

Picture a dataset with two features: age (0–100) and income (0–1,000,000). To the model, every unit counts the same — so a $1 change in income looks identical to a 1-year change in age, and income's huge range lets it dominate every gradient step.

Normalization rescales features onto a common footing so no column bullies the rest. Two workhorses:

• Min-max scaling squashes everything into the range 0 to 1
• Z-score standardization recenters each feature around a mean of 0 with a spread of 1

After scaling, the model compares features on their merits — not their units.