What is variance threshold feature selection?

Updated May 16, 2026

Short answer

Variance threshold removes features with very low variance as they provide little predictive information.

Deep explanation

Features with near-zero variance do not change much across samples and are often uninformative. Removing them simplifies models and reduces computation without losing predictive power.

Real-world example

Used in sensor datasets where some sensors produce constant readings.

Common mistakes

  • Removing low-variance features that might still be important in rare-event prediction.

Follow-up questions

  • Why remove low variance features?
  • Can variance threshold remove important features?

More Feature Engineering interview questions

View all →