Abstract

Recent developments in artificial intelligence and data availability are reshaping how we model, understand, and forecast complex systems. This session brings together three perspectives at the intersection of statistics, data science, and AI, with applications ranging from time series forecasting to innovation analysis and network modeling.

The first talk examines the role of generative AI and foundation models in time series forecasting, comparing them with classical statistical models and recurrent neural networks. Using benchmark datasets, it discusses both predictive performance and uncertainty, with a focus on understanding when these newer approaches offer real advantages.

The second talk introduces a new framework to study global innovation dynamics using large-scale patent data. By proposing a measure of temporal distance to the technological frontier, it provides insights into patterns of technological leadership and catch-up across regions and over time.
The third talk focuses on multilayer and tensor networks and proposes a new methodology for detecting structural changes and analyzing complex interconnected systems. Applications include cryptocurrency markets and transportation networks, highlighting how network-based approaches can support decision-making in real-world problems.

This session is particularly relevant for participants of the ISI World Statistics Congress interested in how statistical methods are evolving in response to large-scale data and AI-driven approaches. It provides both methodological contributions and practical insights, highlighting the continued role of statistics in ensuring reliable, interpretable, and useful results across a range of modern applications.