Business processes may face a variety of problems due to the number of tasks that need to be handled within short time periods, resources' workload and working patterns, as well as bottlenecks. These problems may arise locally and be short-lived, but as the process is forced to operate outside its standard capacity, the effect on the underlying process instances can be costly. We use the term high-level behavior to cover all process behavior which can not be captured in terms of the individual process instances. %Whenever such behavior emerges, we call the cases which are involved in it participating cases. The natural question arises as to how the characteristics of cases relate to the high-level behavior they give rise to. In this work, we first show how to detect and correlate observations of high-level problems, as well as determine the corresponding (non-)participating cases. Then we show how to assess the connection between any case-level characteristic and any given detected sequence of high-level problems. Applying our method on the event data of a real loan application process revealed which specific combinations of delays, batching and busy resources at which particular parts of the process correlate with an application's duration and chance of a positive outcome.

翻译：业务流程可能因短时间内需要处理的任务数量、资源的工作负荷与工作模式以及瓶颈问题而面临各种挑战。这些问题可能局部出现且持续时间短暂，但当流程被迫超出标准容量运行时，其对底层流程实例的影响可能代价高昂。我们用术语“高层行为”来涵盖所有无法通过单个流程实例捕获的流程行为。当此类行为出现时，我们将其涉及的案例称为参与案例。自然产生的问题是：案例特征与其引发的高层行为之间存在何种关联？本研究首先展示了如何检测与关联高层问题的观测结果，并确定相应的（非）参与案例；随后阐明了如何评估任何案例层面特征与任意检测到的高层问题序列之间的关联。将我们的方法应用于真实贷款申请流程的事件数据后，揭示了在流程特定环节中，延迟、批量处理与资源繁忙的特定组合如何与申请持续时长及获得积极结果的可能性存在关联。