Compliance-Aware Predictive Process Monitoring: A Neuro-Symbolic Approach
Summary: arXiv:2603.26948v1 Announce Type: new
Abstract
Existing approaches for predictive process monitoring are sub-symbolic, meaning that they learn correlations between descriptive features and a target feature fully based on data, e.g., predicting the surgical needs of a patient based on historical events and biometrics. However, such approaches fail to incorporate domain-specific process constraints (knowledge), e.g., surgery can only be planned if the patient was released more than a week ago, limiting the adherence to compliance and providing less accurate predictions.
In this paper, we present a neuro-symbolic approach for predictive process monitoring, leveraging Logic Tensor Networks (LTNs) to inject process knowledge into predictive models. The proposed approach follows a structured pipeline consisting of four key stages:
- Feature extraction
- Rule extraction
- Knowledge base creation
- Knowledge injection
Our evaluation shows that, in addition to learning the process constraints, the neuro-symbolic model also achieves better performance, demonstrating higher compliance and improved accuracy compared to baseline approaches across all compliance-aware experiments.
Introduction
Predictive process monitoring has gained significant attention in recent years due to its potential to enhance decision-making in various fields, including healthcare and manufacturing. Traditional models rely heavily on historical data, often leading to predictions that overlook critical domain-specific constraints.
Methodology
The neuro-symbolic approach we propose integrates symbolic reasoning with neural networks, allowing for a more robust understanding of process constraints. Here’s how each stage of our pipeline contributes to better predictive modeling:
- Feature Extraction: This initial stage involves gathering relevant data that reflects the operational context of the process.
- Rule Extraction: In this stage, we identify rules from the data that dictate how various features interact with one another.
- Knowledge Base Creation: We construct a knowledge base that encapsulates the identified rules and constraints, ensuring that our model respects these during training.
- Knowledge Injection: Finally, we incorporate the knowledge base into our predictive model, enhancing its ability to make compliant predictions.
Results
Our experiments demonstrate that our neuro-symbolic model significantly outperforms traditional sub-symbolic methods. By incorporating process knowledge, we achieved:
- Higher Compliance Rates: Models that adhered to the process constraints showed improved compliance in predictions.
- Improved Accuracy: The predictive accuracy of surgical needs was significantly enhanced when compared to baseline methods.
Conclusion
The integration of domain knowledge through a neuro-symbolic approach provides a promising avenue for improving predictive process monitoring. By ensuring compliance with process constraints, our model not only enhances prediction accuracy but also aligns more closely with real-world operational practices. Future work will focus on expanding this approach to other domains where compliance is critical.