Authors:
Ibrahim Louhi
1
;
Lydia Boudjeloud-Assala
2
and
Thomas Tamisier
3
Affiliations:
1
Université de Lorraine and Luxembourg Institute of Science and Technology, France
;
2
Université de Lorraine, France
;
3
Luxembourg Institute of Science and Technology, Luxembourg
Keyword(s):
Data Stream, Subspace Clustering, Visualization.
Related
Ontology
Subjects/Areas/Topics:
Abstract Data Visualization
;
Computer Vision, Visualization and Computer Graphics
;
General Data Visualization
;
Information and Scientific Visualization
;
Interactive Visual Interfaces for Visualization
;
Interface and Interaction Techniques for Visualization
;
Visual Data Analysis and Knowledge Discovery
;
Visual Representation and Interaction
Abstract:
In this paper, we propose a visual subspace clustering approach for data streams, allowing the user to visually
track data stream behavior. Instead of detecting elements changes, the approach shows visually the variables
impact on the stream evolution, by visualizing the subspace clustering at different levels in real time. First we
apply a clustering on the variables set to obtain subspaces, each subspace consists of homogenous variables
subset. Then we cluster the elements within each subspace. The visualization helps to show the approach
originality and its usefulness in data streams processing.