FuzzStream: Fuzzy data stream clustering based on the online-offline framework

Systems capable of generating data quickly and continuously, known as data streams, are a reality today and tend to increase. Due to the nature of data streams, unsupervised learning, such as clustering algorithms, is appropriate. In addition, techniques derived from fuzzy set theory can be useful a...

Full description

Saved in:
Bibliographic Details
Published in:IEEE International Fuzzy Systems conference proceedings pp. 1 - 6
Main Authors: de Abreu Lopes, Priscilla, de Arruda Camargo, Heloisa
Format: Conference Proceeding
Language:English
Published: IEEE 01.07.2017
Subjects:
Algorithm design and analysis
Clustering algorithms
Integrated circuits
Maintenance engineering
Partitioning algorithms
Proposals
Prototypes
ISSN:1558-4739
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Systems capable of generating data quickly and continuously, known as data streams, are a reality today and tend to increase. Due to the nature of data streams, unsupervised learning, such as clustering algorithms, is appropriate. In addition, techniques derived from fuzzy set theory can be useful and add flexibility to the learning process. Fuzzy clustering algorithms for data streams found in the literature are based on chunks, which require the definition of several parameters besides presenting the drawback of overly reducing the summarization of data. An approach to Data Stream clustering that overpasses some of the limitations of chunk-based algorithms is the one called Online-Offline Framework. This framework comprises two phases: summarization and clustering. To the best of our knowledge, there is not a fuzzy version of this framework. The objective of this work is to propose a fuzzy version for the Online-Offline Framework, called FuzzStream, whose main component is a summarization structure and its corresponding maintenance algorithm to be used in the online phase. The well known Weighted Fuzzy C-Means clustering algorithm is used in the offline phase. Experiments show that our proposal is a promising approach to deal with data streams and presents benefits with relation to the classic version.
ISSN:1558-4739
DOI:10.1109/FUZZ-IEEE.2017.8015698