What is it about?
Business processes often involve many interconnected objects, such as orders, products, customers and deliveries. Predicting how these processes will evolve is difficult because the objects influence one another. CICERONE transforms ongoing object-centric process executions into natural-language narratives that preserve these relationships. A Large Language Model then analyses all ongoing executions together and produces predictions for all the involved objects in a single step.
Featured Image
Photo by Alina Grubnyak on Unsplash
Why is it important?
Most predictive process monitoring methods consider individual process traces or objects separately. However, real-world processes are interconnected: a delay, decision or event involving one object may affect several others. CICERONE introduces a global approach that explicitly considers these interactions. To the best of our knowledge, it is the first method to combine Large Language Models and Object-Centric Event Logs specifically for Predictive Process Monitoring. KEY TAKEAWASYS • Natural language can represent complex object-centric process executions and their interactions. • Large Language Models can support global prediction across multiple ongoing process executions. • Considering relationships among objects can improve object-centric predictive monitoring. • Occlusion-based analysis helps explain how object interactions affect the predictions.
Perspectives
This work reflects a research direction that I find particularly stimulating: bringing together object-centric process mining and Large Language Models without losing the structural complexity of real-world processes. I especially value the idea behind CICERONE of moving beyond predictions for isolated cases and considering all ongoing, interconnected process executions from a global perspective. Transforming these executions into natural-language narratives allowed us to explore a new way of making complex object interactions accessible to language models. It was also a pleasure to develop this research with Vincenzo Pasquadibisceglie and Annalisa Appice. I hope the paper will encourage further investigation into how language-based representations and global learning can support more accurate, explainable and realistic predictive process monitoring.
Prof. Donato Malerba
Universita degli Studi di Bari Aldo Moro
Read the Original
This page is a summary of: CICERONE: A natural language-based global approach for object-centric Predictive Process Monitoring, Information Systems, December 2026, Elsevier,
DOI: 10.1016/j.is.2026.102788.
You can read the full text:
Contributors
The following have contributed to this page
