Graph Deep Learning for Time Series Forecasting

Methods, Challenges, and Practice
Erscheinungsjahr 2027
ISBN: 978-3-032-43626-9
Verlag: Springer

Buch, Englisch, Format (B × H): 155 mm x 235 mm

Neural networks have been used to forecast time series for decades. One of the key elements enabling most of the field's recent achievements is the training of a single neural network on large collections of related time series. This approach, however, often considers each time series independently from the others and, consequently, discards dependencies that might be instrumental for accurate predictions. 

This book's purpose is to aim to address the shortcomings of the state of the art in correlated time series forecasting by relying on graph representations and graph deep learning methods. The author proposes graph-based predictors that model pairwise relationships among time series by conditioning forecasts on a (possibly dynamic) graph spanning the collection. The research introduces a comprehensive methodological framework characterizing the family of these predictive models and provides design principles for graph-based forecasting. Within this context, the book proposes methods to tackle the inherent challenges of the field, i.e., dealing with irregular observations, local effects, and latent relational dependencies. In addition, it addresses the computational scalability of the proposed framework, together with methodologies enabling transfer learning and hierarchical forecasting.

Extensive empirical results validate the introduced methodologies and place graph deep learning methods among the most important tools available in modern forecasting.

This book is a revised version of the PhD dissertation written by the author to receive his PhD from the Università della Svizzera italiana, Lugano, Switzerland. He won the 2025 Informatics Europe Best Dissertation Award, granted to the most outstanding European PhD thesis in the field of computer science. His current research focuses on machine learning methods for time series analysis, graph processing, and uncertainty quantification.

Zielgruppe

Research

Autoren/Hrsg.

Cini, Andrea

Fachgebiete

• Mathematik | Informatik EDV | Informatik Informatik Künstliche Intelligenz Mustererkennung, Biometrik
• Technische Wissenschaften Technik Allgemein Modellierung & Simulation
• Mathematik | Informatik EDV | Informatik Professionelle Anwendung Computersimulation & Modelle, 3-D Graphik
• Mathematik | Informatik EDV | Informatik Informatik Künstliche Intelligenz Maschinelles Lernen
• Mathematik | Informatik Mathematik Stochastik
• Mathematik | Informatik EDV | Informatik Informatik Bildsignalverarbeitung

Inhaltsverzeichnis

.- Introduction.
.- Forecasting correlated time series.
.- Graph deep learning for time series forecasting.
.- Benchmarks and baselines.
.- Local effects.
.- Missing data.
.- Latent graph learning.
.- Computational scalability.
.- Graph-based hierarchical forecasting.
.- Conclusion.
.- Appendixes.

Über Autor(innen)

Andrea Cini is an incoming Assistant Professor at École Polytechnique (FR). Previously he held a Swiss National Science Foundation fellowship, during which he did postdoctoral research at the University of Oxford (UK), at EPFL (CH), and at Mila and McGill University (CA). Andrea earned his PhD from Università della Svizzera italiana (CH), supervised by Prof. Cesare Alippi, and also held a visiting researcher position at Imperial College London (UK). His research focuses on machine learning methods for time series and graph processing. Andrea is one of the creators of Torch Spatiotemporal, and his work has been consistently published in leading machine learning venues, including JMLR, NeurIPS, ICML, and ICLR. He has organized several workshops and tutorials on graph machine learning and related topics at international conferences. His PhD thesis on graph-based forecasting received the Informatics Europe Best Dissertation Award 2025.