基于大模型和图重编程机制的多目标业务流程监控方法
Multi-Objective Business Process Monitoring Method Based on Large Language Models and Graph Reprogramming Mechanisms
隋新 1吴步丹1
作者信息
- 1. 北京邮电大学计算机学院（国家示范性软件学院），北京，100876
- 折叠
摘要
随着企业数字化转型的不断深入，基于事件日志的业务流程预测技术在优化资源配置和风险管控方面发挥着越来越重要的作用。现有的序列化深度学习方法往往难以充分表征业务流程中复杂的非线性拓扑结构，而新兴的大语言模型在面对非自然语言的流程数据时，存在显著的模态差异与语义鸿沟。为了充分利用大语言模型的语义理解与预测能力，本文提出了一种基于大模型与图重编程机制的多目标业务流程监控方法。在引入领域语义词汇表与提示工程策略的同时，实现了对流程下一活动与剩余时间的预测。通过在多个公开真实数据集上进行测试，对本文提出的方法在预测准确性及可解释性上的有效性进行了证明。
Abstract
As enterprises deepen their digital transformation, event-log-based business process prediction technologies play an increasingly vital role in optimising resource allocation and risk management. Existing serialized deep learning approaches often struggle to adequately represent the complex non-linear topological structures inherent in business processes, while emerging large language models exhibit significant modal differences and semantic gaps when processing non-natural language process data. To fully leverage the semantic comprehension and predictive capabilities of large language models, this paper proposes a multi-objective business process monitoring method based on large models and graph reprogramming mechanisms. By incorporating domain-specific semantic vocabularies and prompt engineering strategies, it achieves prediction of the next activity and remaining time for processes. Testing across multiple public real-world datasets demonstrates the proposed method\'s effectiveness in predictive accuracy and interpretability.关键词
流程挖掘/大语言模型/流程活动预测/流程图建模/注意力机制Key words
Process Mining/Large Language Model/Process Activity Prediction/Process Graph Modeling/Attention Mechanism引用本文复制引用
隋新,吴步丹.基于大模型和图重编程机制的多目标业务流程监控方法[EB/OL].(2026-02-12)[2026-02-14].http://www.paper.edu.cn/releasepaper/content/202602-75.学科分类
计算技术、计算机技术/自动化技术、自动化技术设备
评论