Relationship between Variants of One-Class Nearest Neighbours and Creating their Accurate Ensembles

TL;DR

This paper explores variants of one-class nearest neighbor classifiers, analyzes their relationships, optimizes their parameters without non-target data, and proposes ensemble methods that improve classification accuracy on various datasets.

Contribution

It provides a theoretical analysis of OCNN variants, introduces a parameter optimization method without non-target data, and develops ensemble approaches that enhance OCNN performance.

Findings

Random-projection ensembles of OCNN perform best

Parameter optimization improves OCNN accuracy

Theoretical relationships among OCNN variants are established

Abstract

In one-class classification problems, only the data for the target class is available, whereas the data for the non-target class may be completely absent. In this paper, we study one-class nearest neighbour (OCNN) classifiers and their different variants. We present a theoretical analysis to show the relationships among different variants of OCNN that may use different neighbours or thresholds to identify unseen examples of the non-target class. We also present a method based on inter-quartile range for optimising parameters used in OCNN in the absence of non-target data during training. Then, we propose two ensemble approaches based on random subspace and random projection methods to create accurate OCNN ensembles. We tested the proposed methods on 15 benchmark and real world domain-specific datasets and show that random-projection ensembles of OCNN perform best.