Artifact and instructions to generate experimental results for TACAS 2019 paper: Computing the Expected Execution Time of Probabilistic Workflow Nets

This is the artifact accompanying the paper "Computing the Expected Execution Time of Probabilistic Workflow Nets", accepted for TACAS 2019. Timed Probabilistic Workflow Nets (TPWN) are a model for business process modeling extended with time and probabilistic information. In the paper, we have shown that computing the excepted execution time of a TPWN is #P-hard, even for simple net structures. We then developed an exponential time algorithm to compute the expected execution time, and evaluated it on a set of real-word workflow net benchmarks. As a result, we obtained that for most nets, the time can be computed efficiently despite the theoretical hardness.<br><br>This artifact can be used to reproduce the results given in Section 6 "Experimental evaluation" of the paper. It includes the workflow and process mining workbench ProM together with the plugin we implemented to compute the expected execution time. Further, this artifact contains a set of 642 workflow nets from IBM, annotated with 3 sets of random times and probabilities, and scripts to compute the expected execution time of all nets and aggregate the results. This can be used to validate the results in Table 1. Also, the workflow net from the BPI Challenge shown in Fig. 4 of the paper is included, in a deterministic and probabilistic version. This net can be analyzed using ProM to validate the results in Table 2.

本研究工件配套于已被TACAS 2019收录的论文《Computing the Expected Execution Time of Probabilistic Workflow Nets》（《概率工作流网的期望执行时间计算》）。定时概率工作流网（Timed Probabilistic Workflow Nets，TPWN）是一类拓展了时间与概率信息的业务流程建模模型。本论文证明，即便针对简单的网结构，计算TPWN的期望执行时间仍属于#P难问题。随后研究团队提出了一种指数时间算法来计算该期望执行时间，并在一组真实工作流网基准测试集上开展了评估，结果表明尽管该问题存在理论复杂性下界，但多数工作流网的期望执行时间仍可高效求解。

本配套工件可用于复现论文第6章"实验评估"中的实验结果。其中包含了工作流与流程挖掘工作台ProM，以及我们开发的用于计算期望执行时间的插件。此外，本工件还包含了来自IBM的642个工作流网，每个网均标注了3组随机时间与概率参数，同时附带了用于计算所有网的期望执行时间并汇总结果的脚本，可用于验证表1中的实验结果。同时，本工件还收录了论文图4中展示的BPI挑战赛工作流网，分别提供了确定性版本与概率性版本，研究人员可通过ProM对该网进行分析，以验证表2中的实验结果。