Simulated Event Log: Udon-ya

This dataset is a simulated event log in standard XES format. It contains events pertaining to restaurant operations of a fictional Udon, a Japanese noodle dish, Restaurant. In Japanese:うどん屋 [udon-ya]. There are 16,412 generated cases distributed over the span of one year, containing in total 674,486 events of 36 unique activities. The log makes use of the lifecycle extension, providing start and complete events per event instance (+ schedule for one activity). Timestamps are (semi-)realistically distributed over daily and weekly business patterns. The underlying simulation respects resource availability and capacity and incorporates queueing as its main dynamics-drivers in addition to a specified medium-complexity control flow. There is looping behavior, interleaving concurrency and event attribute-dependent control-flow routing. The event log is comprised of three different restaurant processes: preparation (prep), store opening/closing and service. prep: In the mornings, senior employees prepare ingredients with long cooking/resting times. opening/closing: Restaurant setup, opening, cleanup and closing performed with higher propensity by part-time employees. service: Customer groups arrive, queue for tables, are seated and order one dish per customer. Then, ordered dishes are prepared concurrently as resource availability permits and served all at once. Finally, the customers at one table eat concurrently but pay and leave the restaurant together (synchronized). There is lunch service ~11:30-14:30 and dinner service ~18:00-21:00 with peak customer rush hours at ~13:00-13:15 and ~18:30-19:00 on weekdays. The restaurant is closed on Mondays. The dish preparation subprocess has a 1 to n multiplicity with cases, making the log pseudo-object centric. Dishes have a unique global id which is referenced in all associated events. Organizationally, there are six different employees with varying working hours, capacity and activity assignments, propensities and skill. Activity assignments are based on a division of back of the house (BOH) and front of the house (FOH). There are also additional resources to represent, e.g., the restaurant table capacity. Tenchou-san (Jap. 店長), the shop manager. Up to two times faster than the other cooks. Mostly BOH. Oku-san, the shop manager's wife. Bridge between BOH and FOH. Deshi-san A & B, two skilled cooks. Only BOH. Baito-san A & B, two part-time workers mostly serving customers (FOH). Furthermore, there are five engineered instances of temporary concept drift with varying impact within the overall process span from 2023-04-01 to 2024-03-31. From 2023-06-20 to 2023-07-26: Decreased customer arrival rate and group size, resulting in visibly reduced activity. From 2023-09-01 to 2023-10-22: Slightly increased customer arrival rate and group size. From 2023-10-15 to 2023-11-05: Most important (human) resource becomes unavailable, resulting in increased utilization of compensating resources. From 2023-11-19 to 2023-12-10: Reduced restaurant table capacity, not causing any prominent effects. From 2023-12-10 to 2024-01-15: Increased restaurant table capacity, not causing any prominent effects. The simulation software used is a past version of https://git.rwth-aachen.de/leah.tgu/qprsim, by the same author as this dataset. The repository specifically contains the configuration for this log generation as a Jupyter notebook in sample/udonya.ipynb.

本数据集为采用标准XES格式的仿真事件日志，其内容围绕一家虚构的日式乌冬面餐厅（日语：うどん屋 [udon-ya]）的运营活动展开。本次生成的日志共包含16412个流程实例，时间跨度为一年，累计生成674486个事件，涵盖36种唯一活动。该日志使用了生命周期扩展，为每个事件实例提供了开始与完成事件，其中某一活动还包含调度事件。时间戳分布贴合现实中的每日、每周营业规律。底层仿真模型遵循资源可用性与容量约束，以排队机制作为核心动态驱动因素，同时结合了指定的中等复杂度控制流。日志中包含循环行为、交错并发以及基于事件属性的控制流路由。本事件日志涵盖三类不同的餐厅运营流程：备餐（prep）、门店开闭店流程以及服务流程。备餐流程指每日晨间，资深员工需准备需长时间烹煮或静置的食材；门店开闭店流程包含门店布置、开业、清洁与闭店环节，该类活动更多由兼职员工执行；服务流程为顾客群组到店后排队等待就座，入座后每位顾客点选一份餐品，随后餐品制备环节将根据资源可用情况并发进行，并在完成后统一送至餐桌，最后同桌顾客同时用餐，但会同步完成结账并离店。该餐厅营业时间为午市约11:30-14:30、晚市约18:00-21:00，工作日的用餐高峰分别为13:00-13:15与18:30-19:00，且周一闭店。餐品制备子流程与流程实例间存在1对多的多重性关系，因此该日志属于伪面向对象的事件日志。每份餐品均拥有唯一全局标识符，所有关联事件均会引用该标识符。组织层面，该餐厅共设置6名不同的员工，其工作时长、服务容量、可执行活动、活动倾向与技能水平均存在差异。活动分配基于后厨（back of the house, BOH）与前厅（front of the house, FOH）的职责划分，此外还设置了额外资源以表征餐厅桌位容量等信息。具体岗位包括：店长（Tenchou-san，日语：店長），为餐厅经理，工作效率可达其他厨师的两倍，主要负责后厨工作；老板娘（Oku-san），为店长的妻子，作为后厨与前厅的沟通桥梁；学徒厨师A与B（Deshi-san A & B），为两名熟练厨师，仅负责后厨工作；兼职员工A与B（Baito-san A & B），为两名兼职人员，主要负责前厅服务工作。本次仿真还引入了5种经过人工设计的临时概念漂移场景，在2023-04-01至2024-03-31的整体时间范围内，不同场景的影响程度各异：1. 2023-06-20至2023-07-26：顾客到达率与群组规模下降，导致活动量显著降低；2. 2023-09-01至2023-10-22：顾客到达率与群组规模小幅提升；3. 2023-10-15至2023-11-05：核心人力资源不可用，导致替代资源的使用率上升；4. 2023-11-19至2023-12-10：餐厅桌位容量下降，未产生显著影响；5. 2023-12-10至2024-01-15：餐厅桌位容量提升，未产生显著影响。本次仿真使用的软件为https://git.rwth-aachen.de/leah.tgu/qprsim的旧版本，其开发者与本数据集的作者为同一人。该代码仓库还包含了用于生成本事件日志的配置文件，存储于sample/udonya.ipynb的Jupyter Notebook中。