TimeLighting: Guided Exploration of 2D Temporal Network Projections

TL;DR

TimeLighting presents a guidance-enhanced visual analytics approach for exploring temporal networks directly in the space-time cube, preserving full temporal resolution while guiding users to informative intervals and elements. By projecting node trajectories into a 2D visualization through time-coloring and a movement-based ranking, it enables detailed analysis of dynamics, aging, and relations without relying on fragile timeslices. The system combines a Main View with trajectory density, edge-on-demand rendering, and a movement score alongside timeline- and node-centric guidance to support tasks like overview, event tracking, and relationship investigation. Evaluations with experts on real datasets show strong insight generation and task efficiency, while revealing areas for improving explainability of layout decisions, guidance adaptivity, and scalability. Overall, TimeLighting demonstrates how guided visualization can enhance the interpretation of complex temporal graphs and inform targeted explorations in real-world networks.

Abstract

In temporal ( event-based ) networks, time is a continuous axis, with real-valued time coordinates for each node and edge. Computing a layout for such graphs means embedding the node trajectories and edge surfaces over time in a 2D+t space, known as the space-time cube. Currently, these space-time cube layouts are visualized through animation or by slicing the cube at regular intervals. However, both techniques present problems such as below-average performance on tasks as well as loss of precision and difficulties in selecting timeslice intervals. In this paper, we present TimeLighting , a novel visual analytics approach to visualize and explore temporal graphs embedded in the space-time cube. Our interactive approach highlights node trajectories and their movement over time, visualizes node "aging", and provides guidance to support users during exploration by indicating interesting time intervals ("when") and network elements ("where") are located for a detail-oriented investigation. This combined focus helps to gain deeper insights into the temporal network's underlying behavior. We assess the utility and efficacy of our approach through two case studies and qualitative expert evaluation. The results demonstrate how TimeLighting supports identifying temporal patterns, extracting insights from nodes with high activity, and guiding the exploration and analysis process.

Figures (7)

• Figure 1: The TimeLighting metaphor. Rays of light (dashed lines), travel from $t=0$ through the space-time cube and reach the observer at $t_{max}$. They interact with nodes' trajectories and carry this information to the projection plane.
• Figure 2: TimeLighting overview. The view is comprised of the (A) toolbar and sidebar, (B) main view, (C) event timeline, and (D) the side bar depicting the movement scores. The yellow bar in the timeline shows the absolute age of the hovered node (visible in the top-center area).
• Figure 3: Trajectories: (a) a higher sampling frequency is selected and aging is visible thanks to the change in opacity; (b) sampling is lowered (3 points per segment) and the trajectory is superimposed on mouseovering the yellow node. Orange stroke and size difference indicate coordinates coming from the data compared to interpolated nodes.
• Figure 4: Example of a locked trajectory. The circles in red are fully within the temporal selection of the users, whereas the less saturated ones are outside of the selected interval.
• Figure 5: Illustrations from Case Study 1. We can observe how the movement of nodes changes in the two halves of the season. (A) The first half of the season; (B) The second half of the season. Timeline brushing is used to filter out events. Trajectory sampling is set at 4 points. (C) The two teams' trajectories are visible as guidance intervals in the timeline. The selected one (blue rectangle) relates to the match between the two teams in the first round of the competition. (D) Highlights the final match and the connections between the 2nd best team and other teams of interest in the case study.