Business processes are a key driver of organizational success, which is why business process improvement (BPI) is a central activity of business process management. Despite an abundance of approaches, BPI as a creative task is time-consuming and labour-intensive. Most importantly, its level of computational support is low. The few computational BPI approaches hardly leverage the opportunities brought about by computational creativity, neglect process data, and rely on rather rigid improvement patterns. Given the increasing amount of process data in the form of event logs and the uptake of generative machine learning for automating creative tasks in various domains, there is huge potential for BPI. Hence, following the design science research paradigm, we specified, implemented, and evaluated ProcessGAN, a novel computational BPI approach based on generative adversarial networks that supports the creation of BPI ideas. Our evaluation shows that ProcessGAN improves the creativity of process designers, particularly the originality of BPI ideas, and shapes up useful in real-world settings. Moreover, ProcessGAN is the first approach to combine BPI and computational creativity.

业务流程是组织成功的关键驱动力，这就是为什么业务流程改进 (BPI) 是业务流程管理的核心活动。尽管有很多方法，但 BPI 作为一项创造性任务是耗时耗力的。最重要的是，它的计算支持水平很低。为数不多的计算 BPI 方法几乎没有利用计算创造力带来的机会，忽视了过程数据，并且依赖于相当严格的改进模式。鉴于以事件日志形式出现的过程数据量不断增加，以及在各个领域采用生成式机器学习来自动执行创造性任务，BPI 具有巨大的潜力。因此，我们遵循设计科学研究范式，指定、实施和评估ProcessGAN，一种基于生成对抗网络的新型计算 BPI 方法，支持 BPI 思想的创建。我们的评估表明，ProcessGAN提高了流程设计者的创造力，尤其是 BPI 思想的原创性，并塑造了在现实世界中有用的环境。此外，ProcessGAN是第一个结合 BPI 和计算创造力的方法。