Evaluation, Model Selection and Unsupervised Outlier Detection in Subspaces

• Grant number: 17/04161-0
• Support Opportunities: Scholarships abroad - Research Internship - Doctorate
• Start date: September 01, 2017
• End date: August 31, 2018
• Field of knowledge: Physical Sciences and Mathematics - Computer Science
• Principal Investigator: José Fernando Rodrigues Júnior
• Grantee: Henrique Oliveira Marques
• Supervisor: Joerg Sander
• Host Institution: Instituto de Ciências Matemáticas e de Computação (ICMC). Universidade de São Paulo (USP). São Carlos , SP, Brazil
• Institution abroad: University of Alberta, Canada
• Associated to the scholarship: 15/06019-0 - Evaluation, model selection and unsupervised outlier detection in data spaces and subspaces, BP.DR

Abstract

Although there is a growing literature that tackles the unsupervised outlier detection problem, the unsupervised evaluation of outlier detection results has been notably overlooked. In contrast to the unsupervised cluster analysis, where indexes for internal evaluation and validation of clustering solutions have been conceived and shown to be very useful, in the outlier detection domain only recently an index for internal evaluation of top-n (binary) outlier detection results, called IREOS, was firstly proposed by the candidate during his Masters. The index, based on the separability given by a maximum margin classifier (e.g. SVM and KLR), can evaluate and compare different candidate labelings based solely on the data and the assessed solutions themselves. And, consequently, allowing the selection of the most promising solutions corresponding to the more suitable models (algorithms, parameters). In this project, we propose a step towards bridging the gap related to the internal evaluation of subspace outlier detection results. Moreover, we also intend to use some of the principles used by IREOS to give rise a new unsupervised subspace outlier detection algorithm.

Scientific publications

(References retrieved automatically from Web of Science and SciELO through information on FAPESP grants and their corresponding numbers as mentioned in the publications by the authors)

MARQUES, HENRIQUE O.; CAMPELLO, RICARDO J. G. B.; SANDER, JORG; ZIMEK, ARTHUR. Internal Evaluation of Unsupervised Outlier Detection. ACM TRANSACTIONS ON KNOWLEDGE DISCOVERY FROM DATA, v. 14, n. 4, JUL 2020. (17/04161-0, 15/06019-0)