Evolutionary algorithms for data stream clustering

• Grant number: 12/10396-6
• Support Opportunities: Scholarships abroad - Research Internship - Doctorate
• Effective date (Start): September 01, 2012
• Effective date (End): February 28, 2013
• Field of knowledge: Physical Sciences and Mathematics - Computer Science
• Principal Investigator: André Carlos Ponce de Leon Ferreira de Carvalho
• Grantee: Jonathan de Andrade Silva
• Supervisor: João Manuel Portela da Gama
• Host Institution: Instituto de Ciências Matemáticas e de Computação (ICMC). Universidade de São Paulo (USP). São Carlos , SP, Brazil
• Research place: Universidade do Porto (UP), Portugal
• Associated to the scholarship: 10/15049-7 - Clustering Data Streams with Automatic Estimation of Number of Clusters, BP.DR

Abstract

Data clustering techniques are commonly used to find clusters. The number of clusters is usually a priori unknown. Such techniques assume that the data set is fixed-size and can be stored entirely into main memory. However, an actual and important challenge involves applying such data clustering techniques into sources of data in which data flows continuously over dynamic environments. These data sources are known as data streams. In many data stream applications, data access operations are restricted to one (or to a small number) of passes over the data with time and memory restrictions. In this sense, some data stream clustering algorithms have been proposed in literature. Many of these techniques are based on the k-means algorithm. However, k-means suffers from several major drawbacks, particularly related to local minima and to the need of specifying the number of clusters in advance. In this context, the main goal of this project involves the development and evaluation of evolutionary algorithms for data stream clustering that estimate automatically the number of clusters from data. (AU)