Predictive Process Monitoring Using Object-centric Graph Embeddings

2025-07-21

AI Summary

This paper addresses process prediction in object-centric event logs, targeting joint next-activity and next-event-time prediction. We propose an end-to-end method that unifies graph-structured modeling with sequential modeling. Specifically, we construct an object–activity heterogeneous graph and employ Graph Attention Networks (GATs) to capture dynamic inter-object interactions, while leveraging LSTMs to model cross-object temporal dependencies. Object-centric log parsing and joint loss optimization enable holistic modeling of complex, collaborative object behavior. Experiments on one real-world and three synthetic object-centric event logs demonstrate that our approach achieves state-of-the-art performance on both prediction tasks—outperforming conventional process mining techniques and pure sequence-based models. Results validate the effectiveness and generalizability of our framework in modeling object interactions within object-aware processes.

Technology Category

• Planning, Routing, and Scheduling: Activity and Plan Recognition
• Machine Learning: Graph-based Machine Learning
• Data Mining & Knowledge Management: Graph Mining, Social Network Analysis & Community

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Web Mining and Content Analysis: Models for Web evolution

Abstract

Object-centric predictive process monitoring explores and utilizes object-centric event logs to enhance process predictions. The main challenge lies in extracting relevant information and building effective models. In this paper, we propose an end-to-end model that predicts future process behavior, focusing on two tasks: next activity prediction and next event time. The proposed model employs a graph attention network to encode activities and their relationships, combined with an LSTM network to handle temporal dependencies. Evaluated on one reallife and three synthetic event logs, the model demonstrates competitive performance compared to state-of-the-art methods.

Problem

• Predict future process behavior using object-centric event logs
• Extract relevant information for next activity prediction
• Model temporal dependencies for next event time prediction

Innovation

• Uses object-centric event logs
• Employs graph attention network
• Combines LSTM for temporal dependencies