Visualizing Uncertainty in Probabilistic Graphs with Network Hypothetical Outcome Plots (NetHOPs)

Area of Science:

• Computer Science
• Data Visualization
• Network Analysis

Background:

• Traditional node-link diagrams struggle to represent probabilistic graphs, hindering estimation of network statistics under uncertainty.
• Visual encodings like edge width or fuzziness are insufficient for static network visualizations.
• Estimating network properties like density, path lengths, or clustering becomes challenging with edge probabilities.

Purpose of the Study:

• Introduce Network Hypothetical Outcome Plots (NetHOPs) as a novel visualization technique for probabilistic graphs.
• Evaluate the effectiveness of NetHOPs in enabling network analysts to reason about network properties under uncertainty.
• Provide design recommendations for animated visualizations of probabilistic networks.

Main Methods:

• Developed NetHOPs, a technique animating sequential network realizations from a probabilistic edge distribution.
• Employed aggregation and anchoring algorithms for layout stability and uncertainty estimation.
• Implemented a community matching algorithm to visualize uncertainty in cluster membership and community occurrence.

Main Results:

• Network experts using NetHOPs estimated network statistics within 11% of ground truth on average.
• Participants demonstrated improved articulation of network statistic distributions when manipulating layout anchoring and animation speed.
• The study involved 51 network experts performing common visual analysis tasks.

Conclusions:

• NetHOPs offer a viable approach for network analysts to reason about multiple network properties under uncertainty.
• Interactive control over layout anchoring and animation speed enhances the perception of network statistics.
• Design recommendations are synthesized for future animated probabilistic network visualizations.