Wavelet-based clustering for data streams.

• Grant number: 10/05062-6
• Support Opportunities: Scholarships in Brazil - Doctorate
• Effective date (Start): April 01, 2011
• Effective date (End): October 31, 2013
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal Investigator: Rodrigo Fernandes de Mello
• Grantee: Cássio Martini Martins Pereira
• Host Institution: Instituto de Ciências Matemáticas e de Computação (ICMC). Universidade de São Paulo (USP). São Carlos , SP, Brazil

Abstract

Recently, technological advances have increased the capacity of generation and capture of data in many areas of society. That data started to constitute a continuous stream, which needs processing for extraction of useful knowledge. This need motivated the creation of the data stream mining area, which differs from others by imposing restrictions in the amount of memory and time available for algorithms, due to the huge volume of data continuously generated. In 2006, the Brazilian Computer Society (SBC) defined five great challenges for computing research in Brazil until 2016. The first challenge deals with the management of information in large volumes of data, pointing out that research in this area might help in e-learning, e-gov and e-science scenarios, besides in processing of information from digital TV, which has been a strong subject of Brazil's government funding. Clustering algorithms have been an attractive approach for data stream clustering due to its ability to generate models without human supervision. However, existing algorithms for data stream clustering consider only the similarity between patterns through a distance function, defined in the metric space the data is in. This work assumes the hypothesis that analyzing the frequencies which compose the observed patterns, through the Wavelet transform, it is possible to better model them. In this sense, this project aims to propose a new clustering algorithm for data streams, which considers not only the distance between patterns, but also the frequencies which compose them. Better data clustering quality is hoped to be achieved in an efficient way.

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)

PEREIRA, CASSIO M. M.; DE MELLO, RODRIGO F.. TS-stream: clustering time series on data streams. JOURNAL OF INTELLIGENT INFORMATION SYSTEMS, v. 42, n. 3, p. 531-566, JUN 2014. (10/05062-6)

Academic Publications

(References retrieved automatically from State of São Paulo Research Institutions)

PEREIRA, Cássio Martini Martins. Time series clustering for data streams. 2013. Doctoral Thesis - Universidade de São Paulo (USP). Instituto de Ciências Matemáticas e de Computação (ICMC/SB) São Carlos.