RBPCP: Visualization on multi-set high-dimensional data

Abstract

Due to the prevalence of Multi-set high-dimensional data in the era of big data, visualization and visual analysis of multiple sets of high-dimensional data are critical to the discovery of data patterns. Parallel Coordinates Plot (PCP) is mainly used for visual analysis of different attributes of the same set. The classic method for visualization on multi-set high-dimensional data is to use a conventional PCP with superimposition design or multiple PCPs with juxtaposition design. However, these methods can't effectively detect patterns. We propose a median-based rearrangement algorithm for bundled Parallel Coordinates Plots. Based on the algorithm, we present Rearranged Bundled Parallel Coordinates Plot (RBPCP) which effectively improves the visual analysis capability of multi-set high-dimensional data and satisfy the aesthetic requirements. In addition, the proposed RBPCP also utilize brushing, mode switch and dynamic axis scaling for the interactive analysis of set relations and hidden patterns.