Title: Local outlier detection based on information entropy weighting

Authors: Lina Wang; Chao Feng; Yongjun Ren; Jinyue Xia

Addresses: School of Electronics and Information Engineering, Nanjing University of Information Science and Technology, No. 219, Ningliu Road, Nanjing, 210044, China ' School of Electronics and Information Engineering, Nanjing University of Information Science and Technology, No. 219, Ningliu Road, Nanjing, 210044, China ' School of Computer and Software, Nanjing University of Information Science and Technology, No. 219, Ningliu Road, Nanjing, 210044, China ' International Business Machines Corporation (IBM), New York, USA

Abstract: As a key research area in data mining technologies, outlier detection can expose data inconsistent with the majority in the dataset and therefore is applicable in extensive areas. The addition of entropy weighting to the spatial local outlier measure (SLOM) and local distance-based outlier factor (LDOF) algorithms in outlier data mining, i.e., the adoption of entropy in the calculation of weighted distance is taken into consideration, leads to enhanced accuracy of outlier detection and produces more expense of time. The algorithm of entropy-weighted LDOF is more optimised than that of entropy-weighted SLOM in terms of detection accuracy. The superiority of the entropy-weighted algorithm is verified through experimental results.

Keywords: local outlier; detection; information entropy weighting; SLOM; spatial local outlier measure; LDOF; local distance-based outlier factor.

DOI: 10.1504/IJSNET.2019.101239

International Journal of Sensor Networks, 2019 Vol.30 No.4, pp.207 - 217

Received: 26 Feb 2019
Accepted: 07 Mar 2019
Published online: 29 Jul 2019 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article