Support vector regression with orthogonal constraints for outlier suppression
Abstract
This paper introduces a novel support vector regression framework with an orthogonal constraint (OC-SVR) for outlier suppression. The proposed method detects outliers via angular deviations in the feature space and modifies the support vector regression (SVR) optimization problem by enforcing orthogonality between the regression function and the detected outliers' feature directions, thereby neutralizing their influence. Experimental evaluations on benchmark datasets demonstrate that OC-SVR achieves significantly improved robustness and predictive accuracy compared to standard SVR and robust SVR variants.
Fields of Study & Subjects
Anomaly Detection Techniques and Applications
Face and Expression Recognition
Advanced Statistical Methods and Models
Publication details
- Journal / source
- AIMS Mathematics
- Volume
- 11
- Issue
- 1
- Pages
- 462–482
- Publication date
- 1 January 2026