Mahalanobis vs. Euclidean Distance for outlier detection.
Updated May 5, 2026
Short answer
Mahalanobis accounts for correlations between variables[cite: 1].
Deep explanation
Euclidean distance ignores feature distribution. Mahalanobis measures distance in units of standard deviation[cite: 1].
Real-world example
Detecting anomalies in correlated sensor data (e.g., pressure and temperature)[cite: 1].
Common mistakes
- Assuming variables are independent and using Euclidean[cite: 1].
Follow-up questions
- When to use Euclidean?