A Meta-Learning Approach Applied to Data Clustering

• Grant number: 11/21816-3
• Support Opportunities: Scholarships in Brazil - Doctorate
• Effective date (Start): March 01, 2012
• Effective date (End): February 28, 2014
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal Investigator: Leandro Nunes de Castro Silva
• Grantee: Daniel Gomes Ferrari
• Host Institution: Escola de Engenharia (EE). Universidade Presbiteriana Mackenzie (UPM). Instituto Presbiteriano Mackenzie. São Paulo , SP, Brazil

Abstract

There is a huge amount of information available as databases. Researchers have dedicated to the development of methods for the knowledge extraction from databases, known altogether as data mining techniques. The two main tasks within data mining are classification and clustering.Currently, there are several data mining techniques capable of solving various problems, but the field still lacks some general guidelines for the choice of a specific algorithm to solve each problem. Meta-learning deals with the problem of finding which features of a problem contribute to a better performance of an algorithm over another and recommend the most adequate algorithm for solving a given problem.Although the literature on meta-learning for classification is vast, the research on meta-learning for clustering is still naive. Added to the importance of the clustering task within data mining and practical applications, this project aims at bringing together meta-learning and clustering in a deep and systematic form, so as to identify which clustering techniques should be employed for each type of problem.Therefore, this project contributes substantially to the scientific research in data clustering, and the graduation of a researcher in data mining. It is worth mentioning that this project is relevant from a practical and theoretical perspective, investigating trending topics in the field with applications to numerical data.

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)

FERRARI, DANIEL GOMES; DE CASTRO, LEANDRO NUNES. Clustering algorithm selection by meta-learning systems: A new distance-based problem characterization and ranking combination methods. INFORMATION SCIENCES, v. 301, p. 181-194, APR 20 2015. (11/21816-3)