基于Attention-based BiLSTM的可解释性流程规范性监控方法

With the widespread adoption of process information systems in corporate management, there is a growing interest in prescriptive process analysis techniques. However, existing prescriptive analysis approaches often overlook the interpretability of recommended actions, limiting their application in production environments. To address this issue, this paper introduces a novel approach based on an attention-based bidirectional LSTM network and a process transition system. This method can not only provide recommended execution actions for optimizing process instances towards the key performance indicators of the process but also provide relevant explanations. Through testing on real-world datasets, our proposed method demonstrates significant efficacy in both prescriptive process monitoring and providing pertinent explanations.