Demand-driven kNN classification

Jiagang Song

doi:10.1016/j.knosys.2025.114090

Demand-driven kNN classification

2025-07-15

0

PRE

AI

DOI：10.1016/j.knosys.2025.114090

Original

Original request for help

Save

Abstract

En 中文

• A demand-driven kNN framework is proposed to learn the most suitable k per test instance. • The method adapts k values based on user-defined confidence preferences and requirements. • A novel objective function is formulated using a bias term and an anchor graph structure. • Two new metrics, k-entropy and k +-entropy, are introduced for suitable k selection.

Keywords:

demand-driven

kNN

confidence preference

objective function

k-entropy

Demand-driven kNN classification

Abstract

Journal

Organization