A scalable dynamic ensemble selection using fuzzy hyperboxes

Reza Davtalab, Rafael M. O. Cruz, Robert Sabourin

Abstract: Highlights•We propose a dynamic ensemble selection framework based on fuzzy hyperboxes.•FH-DES can model the classifier’s misclassification leading to improved selection.•We demonstrate modeling misclassifications leads to a better selection scheme.•We propose and evaluate several strategies for hyperbox contraction and expansion.•FH-DES outperforms existing methods in large-scale tasks and can learn incrementally.