midPCA
What are limitations of PCA in real-world applications?
Updated May 17, 2026
Short answer
PCA assumes linearity and is sensitive to scaling and outliers.
Deep explanation
PCA cannot capture nonlinear relationships, loses interpretability, and is affected by feature scaling and outliers. It may also discard small but important variance features.
Real-world example
Customer segmentation where nonlinear behavior exists.
Common mistakes
- Overusing PCA without understanding dataset structure.
Follow-up questions
- What is alternative to PCA?
- Is PCA deterministic?