Combining Visual Analytics and Content Based Data Retrieval Technology for Efficient Data Analysis

Abstract

One of the most useful techniques to help visual data analysis systems is interactive filtering (brushing). However, visualization techniques often suffer from overlap of graphical items and multiple attributes complexity, making visual selection inefficient. In these situations, the benefits of data visualization are not fully observable because the graphical items do not pop up as comprehensive patterns. In this work we propose the use of content-based data retrieval technology combined with visual analytics. The idea is to use the similarity query functionalities provided by metric space systems in order to select regions of the data domain according to user-guidance and interests. After that, the data found in such regions feed multiple visualization workspaces so that the user can inspect the correspondent datasets. Our experiments showed that the methodology can break the visual analysis process into smaller problems (views) and that the views hold the expectations of the analyst according to his/her similarity query selection, improving data perception and analytical possibilities. Our contribution introduces a principle that can be used in all sorts of visualization techniques and systems, this principle can be extended with different kinds of integration visualization-metric-space, and with different metrics, expanding the possibilities of visual data analysis in aspects such as semantics and scalability.