Choosing the appropriate scaling method depends on the characteristics of your data and the requirements of your machine-learning model. Here’s a brief overview of each scaling method and when you might use them:
Standard Scaler
How does it work?
Scales the features to have a mean of 0 and a standard deviation of 1 using the formula:
Range of output: range around a mean of 0 with a standard deviation of 1
Standard Sclaer works best when the data is normally distributed
Robust Scaling
How does it work?
Scales the feature using Interquartile Range (IQR) after removing the Median.
Range of output: range around a mean of 0 with a standard deviation of 1
MinMax Scaler
How does it work?
Scales the features in the range of [0, 1] using the formula:
Range of output: [0, 1]
Max Absolute Scaler
How does it work?
Scales the features to the range [-1, 1] by dividing each feature by the maximum absolute value
Normalizer
Normalization involves dividing each value by the square root of the sum of the squared values. Here’s how you would calculate it
How does it work?
Conclusion
Always consider the characteristics of your data and the requirements of the algorithms you plan to use when selecting a scaler. It’s often a good practice to experiment with different scaling methods and assess their impact on your specific machine learning task.