ATT-BLKAN: A Hybrid Deep Learning Model Combining Attention is Used to Enhance Business Process Prediction
The role of predictive business process tasks in business process management is significant, as they are capable of anticipating potential process events and implementing timely interventions to address discrepancies between the anticipated and actual workflow. Nevertheless, existing deep learning-b...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2025-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10902041/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The role of predictive business process tasks in business process management is significant, as they are capable of anticipating potential process events and implementing timely interventions to address discrepancies between the anticipated and actual workflow. Nevertheless, existing deep learning-based predictive methods are unable to adequately address the current problem due to shortcomings in the training data, the model itself, or the architectures employed. In this paper, we propose a novel training framework for business process prediction based on improved BiLSTM-KAN, which addresses the issue of adaptability to continuous time data. This is achieved by enhancing the BiLSTM model’s ability to capture long-term dependencies through the addition of Agent Attention, while utilising KAN in place of the traditional Multi-Layer Perceptron (MLP) to improve prediction performance and mechanism interpretability. The results demonstrate that the proposed method outperforms all baseline methods in terms of prediction accuracy. This is evidenced by experiments conducted on five real publicly available event logs, which yielded improvements in accuracy of 12.4%, 7.16%, 9.77%, 12.27%, and 5.98%, respectively. The proposed method offers novel insights into the domain of predictive business processes and demonstrates the considerable potential of KAN in the field of predictive analytics. |
|---|---|
| ISSN: | 2169-3536 |