Why and when is feature scaling necessary

Feature Scaling is useful for models that use distances like SVM and K-means

When Scaling Is Unnecessary

1. Tree-based Algorithms:
   • Algorithms like Decision Tree, Random Forests, and Gradient Boosted Trees are invariant to feature scaling because they split data based on thresholds, not distances.
   • Example: Splits are determined by feature values, not their magnitude.
2. Data with Uniform Scales:
   • If all features have the same range or are already normalized (e.g., percentages), scaling may not be required.