Parallel coordinate order for high‐dimensional data

Abstract

AbstractVisualization of high‐dimensional data is counter‐intuitive using conventional graphs. Parallel coordinates are proposed as an alternative to explore multivariate data more effectively. However, it is difficult to extract relevant information through the parallel coordinates when the data are high‐dimensional with thousands of overlapping lines. The order of the axes determines the perception of information on parallel coordinates. Thus, the information between attributes remains hidden if coordinates are improperly ordered. Here we propose a general framework to reorder the coordinates. This framework is general enough to cover a wide range of data visualization objectives. It is also flexible enough to contain many conventional ordering measures. Consequently, we present the coordinate ordering binary optimization problem and enhance it to achieve a computationally efficient greedy approach that suits high‐dimensional data. Our approach is applied to wine data and genetic data. The purpose of dimension reordering of wine data is to highlight attributes' dependence. Genetic data are reordered to enhance cluster detection. The proposed framework shows that it is able to adapt the criteria for the visualization objective.