# Neuro-Symbolic Learning for Predictive Process Monitoring via Two-Stage Logic Tensor Networks with Rule Pruning

Source: [Neuro-Symbolic Learning for Predictive Process Monitoring via Two-Stage Logic Tensor Networks with Rule Pruning](<https://www.sota2.com/research/paper/69cb55260636e5bf4afc9fe5>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-04-01T05:00:04.436Z

Authors: Fabrizio De Santis, Gyunam Park, Francesco Zanichelli

Published: 2026-03-27T19:32:49.000Z

## Abstract

Predictive modeling on sequential event data is critical for fraud detection and healthcare monitoring. Existing data-driven approaches learn correlations from historical data but fail to incorporate domain-specific sequential constraints and logical rules governing event relationships, limiting accuracy and regulatory compliance. For example, healthcare procedures must follow specific sequences, and financial transactions must adhere to compliance rules. We present a neuro-symbolic approach integrating domain knowledge as differentiable logical constraints using Logic Networks (LTNs). We formalize control-flow, temporal, and payload knowledge using Linear Temporal Logic and first-order logic. Our key contribution is a two-stage optimization strategy addressing LTNs' tendency to satisfy logical formulas at the expense of predictive accuracy. The approach uses weighted axiom loss during pretraining to prioritize data learning, followed by rule pruning that retains only consistent, contributive axioms based on satisfaction dynamics. Evaluation on four real-world event logs shows that domain knowledge injection significantly improves predictive performance, with the two-stage optimization proving essential knowledge (without it, knowledge can severely degrade performance). The approach excels particularly in compliance-constrained scenarios with limited compliant training examples, achieving superior performance compared to purely data-driven baselines while ensuring adherence to domain constraints.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Predictive process monitoring on Sepsis (temporal split)](<https://www.sota2.com/research/sota/predictive-process-monitoring-on-sepsis-temporal-split.md>): Predictive process monitoring; Sepsis (temporal split); Accuracy: 91.3.

- [Predictive process monitoring on BPIC 2012 (temporal)](<https://www.sota2.com/research/sota/predictive-process-monitoring-on-bpic-2012-temporal.md>): Predictive process monitoring; BPIC 2012 (temporal); Accuracy: 65.27.

- [Predictive process monitoring on BPIC 2017 (temporal)](<https://www.sota2.com/research/sota/predictive-process-monitoring-on-bpic-2017-temporal.md>): Predictive process monitoring; BPIC 2017 (temporal); Accuracy: 72.3.

- [Predictive process monitoring on Traffic fines (temporal split)](<https://www.sota2.com/research/sota/predictive-process-monitoring-on-traffic-fines-temporal-split.md>): Predictive process monitoring; Traffic fines (temporal split); Accuracy: 68.44.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2603.26944>)

[PDF](<https://arxiv.org/pdf/2603.26944.pdf>)
