Neuro-Symbolic Predictive Process Monitoring

2025-08-31

AI Summary

To address the critical issue of logical inconsistency in suffix prediction for business process management (BPM), this paper proposes a neuro-symbolic integration method that, for the first time, embeds finite-trace linear temporal logic (LTLf) into an autoregressive sequence prediction framework via a differentiable logic loss function. By combining soft approximations of LTLf semantics with the Gumbel-Softmax reparameterization, we design two differentiable logic loss variants—local and global—that are jointly optimized with standard prediction loss in an end-to-end trainable architecture. Evaluated on three real-world BPM datasets, our approach significantly improves both predictive accuracy and adherence to temporal constraints compared to purely data-driven baselines. Moreover, it demonstrates strong robustness against observational noise and real-world process variations. This work establishes a novel paradigm for compliance-sensitive BPM applications, uniquely balancing prediction fidelity with formal interpretability.

Technology Category

• Constraint Satisfaction and Optimization: Satisfiability Modulo Theories
• Planning, Routing, and Scheduling: Temporal Planning
• Machine Learning: Neuro-Symbolic Learning

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

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.

Problem

• Predicting process suffixes with logical constraints
• Integrating temporal logic into neural network training
• Ensuring generated sequences are accurate and consistent

Innovation

• Neuro-Symbolic approach integrating deep learning with temporal logic
• Differentiable logical loss function using soft LTLf approximation
• Local and global logic loss variants ensuring logical consistency