期刊
COMPUTERS & GRAPHICS-UK
卷 84, 期 -, 页码 160-172出版社
PERGAMON-ELSEVIER SCIENCE LTD
DOI: 10.1016/j.cag.2019.08.015
关键词
Visual analytics; Data visualization; Attribute space analysis; Feature selection
资金
- Brazilian Federal Research Council (CNPq) [307411/2016-8]
As a consequence of the current capabilities of collecting and storing data, a data set of many attributes frequently reflects more than one phenomenon. Understanding the role of attribute subsets and their impact on the organization and structure of a data set under study is paramount to many exploratory and analytical tasks. Example applications range from medicine to financial markets, whereby one wishes to locate subsets of variables that impact the prediction of target categorical attributes. The user is essential in this context since automated techniques are not currently capable of embedding user knowledge in attribute selections. In this work, we propose an approach to deal with the analysis and selection of attributes in a data set based on three principles: firstly, we center the analysis of the relationships on categorical attributes or labels, because they usually summarize important state variables in the application; secondly, we express the relationship between target attributes and all others in the data set within a single visualization, providing understanding of a large number of correlations in the same visual frame; thirdly, we propose an interactive dual-visual approach whereby changes and selections in attribute space reflect visually on the configuration of data layouts, conceived to support immediate analysis of the impact of selected subsets of attributes in the organization of the data set. We validate our approach by means of a number of case studies, illustrating distinct scenarios of knowledge acquisition and feature selection. (C) 2019 Elsevier Ltd. All rights reserved.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据