📄 中文摘要
现有的预测过程监控方法主要是亚符号的，完全基于数据学习描述特征与目标特征之间的关联，例如根据历史事件和生物特征预测患者的手术需求。然而，这些方法未能纳入特定领域的过程约束（知识），例如，手术只能在患者出院超过一周后进行，这限制了合规性并降低了预测的准确性。研究提出了一种神经符号方法，通过逻辑张量网络（LTNs）将过程知识注入预测模型中。该方法遵循结构化的过程，旨在提高预测的准确性和合规性。
合规意识的预测过程监控：一种神经符号方法
📄 English Summary
Compliance-Aware Predictive Process Monitoring: A Neuro-Symbolic Approach
Existing predictive process monitoring approaches are sub-symbolic, relying solely on data to learn correlations between descriptive features and a target feature, such as predicting a patient's surgical needs based on historical events and biometrics. However, these methods fail to incorporate domain-specific process constraints, such as the requirement that surgery can only be planned if the patient was discharged more than a week ago, which limits compliance and reduces prediction accuracy. A neuro-symbolic approach is proposed, leveraging Logic Tensor Networks (LTNs) to inject process knowledge into predictive models. This structured approach aims to enhance both the accuracy of predictions and adherence to compliance requirements.