Deakin University
Browse

File(s) under permanent embargo

A New Effective and Efficient Measure for Outlying Aspect Mining

conference contribution
posted on 2020-01-01, 00:00 authored by Durgesh Samariya, Sunil AryalSunil Aryal, Kai Ming Ting, Jiangang Ma
Outlying Aspect Mining (OAM) aims to find the subspaces (a.k.a. aspects) in which a given query is an outlier with respect to a given data set. Existing OAM algorithms use traditional distance/density-based outlier scores to rank subspaces. Because these distance/density-based scores depend on the dimensionality of subspaces, they cannot be compared directly between subspaces of different dimensionality. Z-score normalisation has been used to make them comparable. It requires to compute outlier scores of all instances in each subspace. This adds significant computational overhead on top of already expensive density estimation—making OAM algorithms infeasible to run in large and/or high-dimensional datasets. We also discover that Z-score normalisation is inappropriate for OAM in some cases. In this paper, we introduce a new score called Simple Isolation score using Nearest Neighbor Ensemble (SiNNE), which is independent of the dimensionality of subspaces. This enables the scores in subspaces with different dimensionalities to be compared directly without any additional normalisation. Our experimental results revealed that SiNNE produces better or at least the same results as existing scores; and it significantly improves the runtime of an existing OAM algorithm based on beam search.

History

Volume

12343

Pagination

463-474

Location

Amsterdam, The Netherlands

Start date

2020-10-20

End date

2020-10-24

ISSN

0302-9743

eISSN

1611-3349

ISBN-13

9783030620073

Language

eng

Publication classification

E1 Full written paper - refereed

Editor/Contributor(s)

Huang Z, Beek W, Wang H, Zhou R, Zhang Y

Title of proceedings

WISE 2020 : Proceedings of the 2020 International Conference on Web Information Systems Engineering

Event

Web Information Systems Engineering. Conference (2020 : Amsterdam, The Netherlands)

Publisher

Springer

Place of publication

Berlin, Germany

Series

Lecture Notes in Computer Science

Usage metrics

    Research Publications

    Exports

    RefWorks
    BibTeX
    Ref. manager
    Endnote
    DataCite
    NLM
    DC