Novelty detection by machine learning

• Grant number: 15/03355-0
• Support Opportunities: Scholarships in Brazil - Post-Doctorate
• Effective date (Start): August 01, 2015
• Effective date (End): July 31, 2017
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal Investigator: André Carlos Ponce de Leon Ferreira de Carvalho
• Grantee: Isvani Inocencio Frías Blanco
• Host Institution: Instituto de Ciências Matemáticas e de Computação (ICMC). Universidade de São Paulo (USP). São Carlos , SP, Brazil
• Associated research grant: 13/07375-0 - CeMEAI - Center for Mathematical Sciences Applied to Industry, AP.CEPID

Abstract

Traditional machine learning classification techniques are essentially based on the premise that the future distribution of the data remains consistent to the distribution presented in the training set, i.e.,the data distribution is stationary along time. However, over the past few years, more and more new applications need to process not only increasing volumes of data, but also data that is presented continually as a stream. In this scenario the aforementioned assumption is rarely satisfied; future data distribution often changes while data stream processing demands high costs to traditional algorithms, such of processing time and memory. A particular classification problem, which results from such non-stationary datastream, regards identifying previously unknown classes over the stream- a task known as novelty detection. Facing this problem, novelty detection algorithms must conciliate essential, but difficult to gather features, like accuracy and stability at classification performance, processing data instances only once and providing low associated costs with processing time and memory. This work aims at investigating the state of the art methods on novelty detections as well as proposing new ones. The studied and developed methods should be applied to real industrial applications, such as in monitoring and controlling tasks.

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)

FRIAS-BLANCO, ISVANI; DEL CAMPO-AVILA, JOSE; RAMOS-JIMENEZ, GONZALO; CARVALHO, ANDRE C. P. L. F.; ORTIZ-DIAZ, AGUSTIN; MORALES-BUENO, RAFAEL. Online adaptive decision trees based on concentration inequalities. KNOWLEDGE-BASED SYSTEMS, v. 104, p. 179-194, JUL 15 2016. (15/03355-0)

VERDECIA-CABRERA, ALBERTO; BLANCO, ISVANI FRIAS; CARVALHO, ANDRE C. P. L. F.. An online adaptive classifier ensemble for mining non-stationary data streams. Intelligent Data Analysis, v. 22, n. 4, p. 787-806, 2018. (15/03355-0, 13/07375-0)