2016 IEEE Pacific Visualization Symposium (PacificVis) (2016)
April 19, 2016 to April 22, 2016
Fangfang Zhou , School of Information Science and Engineering, Central South University
Juncai Li , School of Software, Central South University
Wei Huang , School of Software, Central South University
Ying Zhao , School of Information Science and Engineering, Central South University
Xiaoru Yuan , Key Laboratory of Machine Perception (Ministry of Education), and School of EECS, Peking University
Xing Liang , School of Computing, Informatics & Decision Systems Engineering, Arizona State University
Yang Shi , School of Information Science and Engineering, Central South University
Subspace-based analysis has increasingly become the preferred method for clustering high-dimensional data. A visually interactive exploration of subspaces and clusters is a cyclic process. Every meaningful discovery will motivate users to re-search subspaces that can provide improved clustering results and reveal the relationships among clusters that can hardly coexist in the original subspaces. However, the combination of dimensions from the original subspaces is not always effective in finding the expected subspaces. In this study, we present an approach that enables users to reconstruct new dimensions from the data projections of subspaces to preserve interesting cluster information. The reconstructed dimensions are included into an analytical workflow with the original dimensions to help users construct target-oriented subspaces which clearly display informative cluster structures. We also provide a visualization tool that assists users in the exploration of subspace clusters by utilizing dimension reconstruction. Several case studies on synthetic and real-world data sets have been performed to prove the effectiveness of our approach. Lastly, further evaluation of the approach has been conducted via expert reviews.
User Interaction, High-Dimensional Data, Subspace Clustering, Visual Clustering
F. Zhou et al., "Dimension reconstruction for visual exploration of subspace clusters in high-dimensional data," 2016 IEEE Pacific Visualization Symposium (PacificVis)(PACIFICVIS), Taipei, Taiwan, 2016, pp. 128-135.