StandardScaler vs MinMaxScaler vs RobustScaler — how do you choose?
easyAnswer
- StandardScaler: mean 0 / std 1.
- Best for approximately Gaussian data and models sensitive to variance (linear + regularization, SVM, PCA, neural nets).
- MinMaxScaler: to [0, 1].
- Useful when features must be strictly positive (e.g., neural nets with sigmoid outputs, image pixels) or when you want a bounded feature range.
- RobustScaler: subtract median, divide by IQR.
- Best when features contain outliers you don't want to remove — outliers don't distort the scale as they would with mean/std.
Check yourself — multiple choice
- They are all equivalent
- Standard: Gaussian data; MinMax: bounded/positive; Robust: outlier-heavy
- Only StandardScaler exists
- MinMaxScaler is only for images
Standard = Gaussian; MinMax = bounded; Robust = outlier-heavy. Match to data + model.
#feature-engineering#preprocessing#features
Practise Supervised Learning
215 interview questions in this topic.
Related questions
- Which models need feature scaling and which don't?
- When should you add interaction features to a linear model?
- Ordinal vs nominal encoding — how do you choose?
- What is target encoding and when is it useful?
- How do you encode cyclical features like hour-of-day or day-of-week?
- How do you turn text into features for a classical model?