Transition-driven time prediction for business processes with cycles
作者：
Highlights：
• The encodings of prefix and the corresponding suffix for traces are proposed.
• The transition divisions consider loop structures using reachability graph.
• The autoencoder for each transition division reduces dimensionality.
• Deep transfer learning is performed on different prediction models.
• Synthetic and real-life event logs are collected for experimental evaluation.
摘要
•The encodings of prefix and the corresponding suffix for traces are proposed.•The transition divisions consider loop structures using reachability graph.•The autoencoder for each transition division reduces dimensionality.•Deep transfer learning is performed on different prediction models.•Synthetic and real-life event logs are collected for experimental evaluation.
论文关键词：Predictive process monitoring,Time prediction,Business process management,Petri net,Deep neural network
论文评审过程：Received 13 September 2021, Revised 5 July 2022, Accepted 18 July 2022, Available online 26 July 2022, Version of Record 5 August 2022.
论文官网地址：https://doi.org/10.1016/j.eswa.2022.118238
