SANE: Visual Analytics for Event-Based Diffusion on Networks

点击翻译按钮获取中文摘要

Wider research context: Visual Analytics (VA) showed its potential in communicating and investigating information diffusion processes over networks. Diffusion processes are highly dynamic and stochastic phenomena. To this end, VA needs to tackle two challenges: representing the progression on the underlying dynamic network structure and capturing the uncertainty of the process. The majority of existing VA approaches approximate the problem by imposing a discrete time structure to the input data and disregarding uncertainties. Objectives: By focusing on the real-time coordinates of the individual propagation events (hence, event-based), we better approximate real-world interactions, which, in turn, improves confidence in the analysis and prediction results. However, this greatly increases the complexity of the problem from both an algorithmic and methodological perspective, and requires a revision of the existing paradigms for the representation of event-based uncertain networks. We call this problem "VA of Event-based Information Diffusion with Uncertainty". We aim to (i) provide a data-task characterization of the information diffusion domain in VA with a typology of the representation of uncertainty on event-based networks, (ii) contribute the current state-of-the-art of temporal graph layout algorithms to improve their accessibility and introduce specific layout strategies to let the underlying diffusion process to shape the final network representation, and (iii) establish a seamless workflow for the visualization of event-based diffusion processes, to serve as a common ground and inspiration for further research in this field. Approach: We base our design methodology on the nested model for visualization design and validation by Munzner et al. And the Design Triangle (Data-Users-Tasks paradigm) by Miksch et al. For the data-task characterization of the domain, we refer to the work by Kerracher et al. About validation and construction of task taxonomies. To validate our findings, we will resort to experimental studies through algorithmic validation, case studies with experts on real-world datasets, and user studies. Innovation: We strive to be among the first to systematically explore the design space of VA for event-based information diffusion with uncertainty. We investigate promising but currently under-represented research trends in VA with innovative solutions and approaches, to contribute to the current body of knowledge in the field and open new and exciting research questions. Primary researchers involved: This proposal puts together two renowned European research groups in Visual Analytics. Alessio Arleo acts as PI and is Post-Doc researcher in VA group at TU Wien; Prof. Silvia Miksch (co-PI) is Prof. in VA at TU Wien. The Co-Applicant from the University of Cologne, under the WEAVE research policy, is Prof. Tatiana Landesberger von Antburg, Prof. in Visualization.