EHHN: An Event-driven Heterogeneous Hypergraph Network for Object-Centric Next Activity Prediction

2026-07-02 · Source: Machine Learning · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Expert, quick

Summary

EHHN, an Event-driven Heterogeneous Hypergraph Network, is proposed for object-centric next activity prediction in service-oriented processes. This model addresses limitations of existing methods that struggle with real-world service processes involving events shared by multiple typed business objects, as captured in Object-centric Event Logs (OCELs). EHHN represents each prediction prefix as a heterogeneous hypergraph, utilizing event--object hyperedges for co-participating objects and a lifecycle hyperedge for the primary object's observed events. Its dual-stream architecture features a micro-spatial stream for event-driven object-state evolution and a macro-evolution stream for temporal dynamics using global prototypes. These streams are fused to predict the next activity. Experiments on four public OCEL benchmarks demonstrate EHHN's superior accuracy and macro F1-score, achieving improvements of up to 8.1 and 12.4 percentage points over the strongest baselines. EHHN also reduces peak GPU memory by up to 24 times compared to the strongest OCEL-native graph baseline. The code was published on 2026-07-02.

Key takeaway

For Machine Learning Engineers developing process prediction systems, EHHN offers a significant advancement for object-centric event logs. If you are struggling with current methods that lose cross-object context or consume excessive GPU memory, you should evaluate EHHN. Its hypergraph representation and dual-stream architecture provide superior accuracy and macro F1-scores, reducing peak GPU memory by up to 24 times. Consider integrating this approach to improve the anticipation of upcoming steps and mitigate service-level risks in complex service processes.

Key insights

EHHN uses a dual-stream heterogeneous hypergraph to model object-centric event logs for superior next activity prediction.

Principles

• Hypergraphs effectively model multi-object event interactions.
• Dual-stream architectures can capture both micro-spatial and macro-temporal dynamics.
• Jointly modeling object state changes and global patterns improves prediction.

Method

EHHN constructs a heterogeneous hypergraph from prediction prefixes, then processes it via a micro-spatial stream for object state and a macro-evolution stream for temporal dynamics, fusing outputs for prediction.

In practice

• Apply hypergraph representations to complex multi-object event logs.
• Implement dual-stream networks for combined local and global context.
• Consider EHHN for process mining tasks requiring high accuracy.

Topics

• Object-centric Event Logs
• Next Activity Prediction
• Hypergraph Networks
• Process Mining
• Event-driven Systems
• GPU Memory Optimization

Code references

• chenkaitao1112/EHHN

Best for: Research Scientist, AI Scientist, Machine Learning Engineer

Related on AIssential

• Schema-Agnostic Process Trace Construction: From Raw Tables to Execution Behavior — Automated process trace reconstruction from raw, schema-poor relational data is achievable via statistical signals and Temporal Convolutional Networks.
• WHAR Arena: Benchmarking the State of the Art in Efficient Wearable Human Activity Recognition — Benchmarking reveals WHAR predictive performance has plateaued, shifting focus to deployment efficiency and domain adaptation.
• Promoting Simple Agents: Ensemble Methods for Event-Log Prediction — N-gram models offer competitive accuracy for event-log prediction with significantly fewer resources than neural networks.
• Build a Domain-Specific Embedding Model in Under a Day — This article introduces a pipeline for building domain-specific embedding models in under a day using a single GPU, significantly improving Retrieval-Augmented Generation (RAG) system performance with…
• PHINN: Persistent Homology Inspired Neural Network for Rare-Event Time Series Generation — Rare events in time series possess stable topological fingerprints, enabling more effective generative modeling than statistical moments.
• A Reproducible Log-Driven AutoML Framework for Interpretable Pipeline Optimization in Healthcare Risk Prediction — Reproducible, log-driven AutoML reveals healthcare risk prediction pipeline component importance and search space redundancy.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine Learning.