Node Overlap Removal Algorithms: an Extended Comparative Study

Abstract

In the context of graph layout, many algorithms have been designed to remove node overlapping, and many quality criteria and associated metrics have been proposed to evaluate those algorithms. Unfortunately, a complete comparison of the algorithms based on some metrics that evaluate their quality has never been provided and it is thus difficult for a visualisation designer to select the algorithm that best suits their needs. In this paper, we review 22 metrics available in the literature, classify them according to the quality criteria they try to capture, and select a representative one for each class. Based on the selected metrics,we compare 9 node overlap removal algorithms. Our experiment involves 854 synthetic and real-world graphs. Finally, we propose a JavaScript library containing both the algorithms and the criteria, and we provide a Web platform, AGORA, in which one can upload graphs, apply the algorithms and compare/download the results.

Guaranteed Visibility in Scatterplots with Tolerance

“…A review of the Nodes Dispersion literature was presented in Chen et al's survey [3,4]. PFS [29], PFS' [14], FTA [16], RWordle-L [33] are orthogonal iterative algorithms that rely on scan-line [7] to identify overlaps in a layout and process them.…”

“…The quality metrics considered in this evaluation are taken or inspired from Li et al [24] and Chen et al [3,4]. We made adjustments to some selected metrics to better reflect their purpose.…”

“…Normalized node movements This metric inspired from Chen et al [3,4] quantifies the relative node movements made from the original layout to the adjusted one. Before computing movement, a scale and shift operation superimposes the adjusted layout on the original one and not the opposite as proposed in [3,4]). It enables the movement quantity to be unsensitive to offsets and movements induced by upscaling effects.…”

VRGrid: Efficient Transformation of 2D Data into Pixel Grid Layout

“…Relative Positioning Preservation (RPP) is computed using the Orthogonal Ordering metric primarily designed to evaluate shape preservation of overlap solvers in graph drawing [22]. We use this metric to evaluate how well the position of each element relative to others is preserved before and after computing the grid arrangement.…”

“…Global Positioning Preservation (GPP) is being computed using the Node Movements metric. It evaluates how coordinates moved from the original projection to the resulting arrangement [22]. While coordinate movements are unavoidable if the shape of the projection is altered, we assume the faithful grid arrangement of 2D data involves minimal movements.…”