Papers
Topics
Authors
Recent
Search
2000 character limit reached

Neuro-Symbolic Predictive Process Monitoring

Published 31 Aug 2025 in cs.AI, cs.FL, cs.LG, and cs.LO | (2509.00834v1)

Abstract: This paper addresses the problem of suffix prediction in Business Process Management (BPM) by proposing a Neuro-Symbolic Predictive Process Monitoring (PPM) approach that integrates data-driven learning with temporal logic-based prior knowledge. While recent approaches leverage deep learning models for suffix prediction, they often fail to satisfy even basic logical constraints due to the absence of explicit integration of domain knowledge during training. We propose a novel method to incorporate Linear Temporal Logic over finite traces (LTLf) into the training process of autoregressive sequence predictors. Our approach introduces a differentiable logical loss function, defined using a soft approximation of LTLf semantics and the Gumbel-Softmax trick, which can be combined with standard predictive losses. This ensures the model learns to generate suffixes that are both accurate and logically consistent. Experimental evaluation on three real-world datasets shows that our method improves suffix prediction accuracy and compliance with temporal constraints. We also introduce two variants of the logic loss (local and global) and demonstrate their effectiveness under noisy and realistic settings. While developed in the context of BPM, our framework is applicable to any symbolic sequence generation task and contributes toward advancing Neuro-Symbolic AI.

Summary

  • The paper presents a novel integration of temporal logic into an autoregressive prediction framework for enhanced suffix prediction in BPM.
  • It introduces differentiable logic losses—local and global variants—using the Gumbel-Softmax trick to enforce logical constraints.
  • Experimental evaluations on real-world BPM datasets confirm significant improvements in both prediction accuracy and logical compliance under varied conditions.

Neuro-Symbolic Predictive Process Monitoring

Introduction

The paper "Neuro-Symbolic Predictive Process Monitoring" introduces a novel approach to enhancing suffix prediction within the field of Business Process Management (BPM). This work addresses the limitations of deep learning models which, while effective in data-driven learning systems, often neglect the logical constraints intrinsic to BPM environments. To bridge this gap, the proposed approach integrates neuro-symbolic methods, specifically incorporating Linear Temporal Logic (LTL) into the training of sequence predictors, thereby facilitating predictions that are both accurate and logically consistent.

Methodology

The methodology leverages an autoregressive sequence prediction framework, integrating temporal logic through the use of a differentiable logical loss function. This function is designed using a soft approximation of LTL semantics, enhanced by the Gumbel-Softmax trick to maintain differentiability. This loss is seamlessly combined with more traditional predictive losses, enabling the model to inherently respect logical constraints during the generation of suffixes. Two variants of logic loss are introduced: local and global, which respectively address prediction adherence at different granularities within the sequence.

Experimental Evaluation

The efficacy of the proposed method is evaluated on three real-world BPM datasets. The results indicate significant improvements in suffix prediction accuracy and compliance with temporal constraints, underscoring the practical applicability of the neuro-symbolic approach. The experiments are conducted under both noiseless and noisy conditions to replicate realistic settings, with the findings affirming the robustness of incorporating logical constraints into predictive models. The proposed local and global logic loss variants exhibit distinct advantages depending on the conditions, suggesting a versatile applicability across different BPM scenarios.

Implications and Future Directions

This research contributes to the advancement of Neuro-Symbolic AI by demonstrating how temporal logic can be effectively integrated into predictive models to enhance logical compliance without sacrificing accuracy. Although the framework is developed in the context of BPM, its principles are applicable to a wide range of symbolic sequence generation tasks. Future work may explore the extension of this methodology to other domains, potentially generalizing beyond BPM and refining the logic loss functions to accommodate broader applications. This paves the way for more holistic AI systems that seamlessly merge data-driven learning with domain-specific logical reasoning.

Conclusion

The paper presents a significant step forward in predictive process monitoring by fusing deep learning with symbolic logic to address the traditional shortcomings of purely data-driven approaches. The integration of LTL into the predictive framework not only enhances accuracy but ensures logical consistency, a core requirement in BPM. This work lays the groundwork for further exploration of neuro-symbolic approaches, underscoring their potential in creating AI systems that are both accurate and accountable to domain constraints.

Paper to Video (Beta)

No one has generated a video about this paper yet.

Whiteboard

No one has generated a whiteboard explanation for this paper yet.

Open Problems

We're still in the process of identifying open problems mentioned in this paper. Please check back in a few minutes.

Collections

Sign up for free to add this paper to one or more collections.