Comparison between methods
This chart provides guidance on which scaling technique is suitable for various data scenarios.
|
Scaling method |
When to use |
Pros |
Cons |
|
Min-max scaling |
Features have a clear, known range Normal distribution is assumed Data does not contain outliers |
Simple and easy to understand Preserves relative relationships Memory efficient |
Sensitive to outliers |
|
Z-score scaling |
Data follows a normal distribution No strong assumptions about the range Handling outliers is not a priority |
Standardizes data to zero mean and deviation of 1 |
Sensitive to outliers May not be suitable for skewed data |
