.jpg)
In the era of digital transformation, predicting the next step in a business process has become a strategic capability to anticipate delays, exceptions, or service level risks. However, traditional methods assume that event logs correspond to individual cases, while many real processes involve multiple business objects that share events. To address this complexity, the EHHN (Event-driven Heterogeneous Hypergraph Network) model emerges, an architecture based on heterogeneous hypergraphs that allows predicting the next activity in object-centric event logs (OCEL).
The EHHN approach represents each prediction prefix as a hypergraph where event-object hyperedges capture interactions between participants, and a lifecycle hyperedge groups the observed events of the main object. This model employs a dual-flow architecture: a micro-spatial flow models the evolution of the object state driven by events, and a macro-evolutionary flow captures temporal dynamics through retrieved global prototypes. The fusion of both flows allows predicting the next activity with high accuracy, outperforming the best baseline models by up to 12.4 percentage points and reducing GPU memory consumption by up to 24 times.
This innovation has direct implications for companies seeking to optimize their processes through artificial intelligence. Implementing prediction solutions like EHHN requires an integrated approach of AI for businesses that combines data analysis, advanced modeling, and deployment on scalable infrastructure. At Q2BSTUDIO, as a software and technology development company, we help organizations design and implement custom predictive systems, whether through tailored applications that integrate AI engines, or via AWS and Azure cloud services that ensure performance and data security.
Furthermore, the use of heterogeneous hypergraphs opens new possibilities in areas such as cybersecurity, where anomaly detection in shared events can prevent incidents, or in business intelligence, where tools like Power BI can visualize predictions for decision-making. Creating AI agents that automate responses to risk predictions is another line of application we explore in our process automation projects.
Ultimately, EHHN represents a significant advance in activity prediction in complex processes. At Q2BSTUDIO, we combine this type of technology with custom software and business intelligence services to offer solutions that not only anticipate the future but do so efficiently and securely, adapting to the multi-object reality of modern business processes.
