Automatic adaptation for clustering data streams

• Grant number: 11/19459-8
• Support Opportunities: Scholarships in Brazil - Post-Doctoral
• Start date: May 01, 2012
• End date: February 28, 2013
• Field of knowledge: Physical Sciences and Mathematics - Computer Science
• Principal Investigator: Rodrigo Fernandes de Mello
• Grantee: Marcelo Keese Albertini
• Host Institution: Instituto de Ciências Matemáticas e de Computação (ICMC). Universidade de São Paulo (USP). São Carlos , SP, Brazil

Abstract

The development of methods of computational analysis in machine learning has facilitated the understanding of complex phenomena.The main method used in the exploratory analysis of phenomena is the data clustering, whose goal is to find and distinguish relevant trends from the assessment of the data similarities.However, the planning and execution of data clustering is a complex task that involves several decisions.Currently, such decisions are made by experts and by the application of iterative methods in which one seeks to optimize the performance assessed in the validation step. However, this approach can have high costs and be impractical in its application to phenomena that require the rapid collection and processing of large volumes of data, i.e., data streams.Recently, researchers have sought alternatives to automatically adapt decisions taken by experts.In order to meet this need, the proponent of this project, in the context of his thesis, has developed an approach to adapt the parameters of clustering algorithms.As a result, there was improvement in the performance of clustering and the possibility of extending this approach for the adaptation of distance functions and the strategies of defining clusters. In order to investigate these possibilities and assist the effective understanding of data streams, this research plan proposes studies to develop approaches to the automatic adjustment of clustering algorithms for data streams.

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)

ALBERTINI, MARCELO KEESE; DE MELLO, RODRIGO FERNANDES. Energy-based function to evaluate data stream clustering. Advances in Data Analysis and Classification, v. 7, n. 4, p. 435-464, DEC 2013. (11/19459-8)