Damilya Saduakhas wins Best Paper Award at Spatial Statistics 2025 in The Netherlands

About

Damilya Saduakhas, a KAUST Ph.D. candidate in statistics, has received the Best Paper Award at the Spatial Statistics 2025: At the Dawn of AI conference, held from July 15 to 18 in Noordwijk, The Netherlands. This well-recognized international event, organized by Elsevier, brings together experts in spatial statistics and artificial intelligence from around the world to present the latest research and discuss emerging challenges in the field. The award was presented by Professor Alfred Stein, Conference Chair, who appears with Damilya in the photo during the award ceremony.

In the award-winning paper “Log-Gaussian Cox Processes on Metric Graphs: Applications to Road Accident Risk Analysis”, Damilya Saduakhas, David Bolin, and Alexandre B. Simas present a novel statistical framework that enables the modeling of spatial point processes directly on road networks. By leveraging advanced mathematical tools and efficient inference methods, their approach, implemented in the open-source MetricGraph R package, makes it possible to accurately identify accident hotspots across complex road systems. This work provides valuable insights for data-driven road safety interventions, as demonstrated in a large-scale analysis of road accident data from Al-Ahsa, Saudi Arabia.

Although modeling spatial point processes has advanced considerably, extending these models to road networks remains a challenge. The authors address this by introducing log-Gaussian Cox processes on compact metric graphs, based on Gaussian Whittle–Matérn fields. To ensure scalability and accuracy, they develop a novel likelihood approximation that enables efficient and fully Bayesian inference without directly approximating the Gaussian process. Integrated with R-INLA and implemented in the MetricGraph R package, this framework allows for the analysis of extensive network data. Their approach identifies high-risk road segments using exceedance probabilities and excursion sets, providing detailed and actionable insights into accident risk across complex networks.