ProBench:一种评估流程相似性查询算法的基准数据集

发布时间：2018-05-13 05:00

本文选题：基准数据集 + 业务流程　；参考：《计算机集成制造系统》2017年05期

【摘要】：针对目前缺乏评估现有流程相似性算法性能的基准数据集问题,在IBM公开的数据集基础上,用Petri网建模流程模型,提出一种评估流程相似性查询算法的基准数据集。该数据集由100个流程模型组成,其中标记出了10个检索流程与其各自的9个相关流程,以及相关流程的排序顺序。对于每个检索流程,其9个相关流程与该检索流程的相关性排序顺序由一个用户调查的结果确定,将该结果作为一个基准对算法的结果进行评估。选取3个基于结构的和1个基于行为的流程相似性查询算法,对它们在准确率和效率两个方面进行了评估,实验结果展示了这些算法各自的适用场景。所提出的基准数据集和相关的算法代码已经公开发布在网上,可供研究人员下载使用。
[Abstract]:In view of the lack of a benchmark data set for evaluating the performance of existing process similarity algorithms, a benchmark data set for evaluating process similarity query algorithm is proposed based on IBM published data sets and process model modeling using Petri nets. The data set consists of 100 process models, in which 10 retrieval processes and their 9 related processes are marked, as well as the ranking order of the related processes. For each retrieval process, the ranking order of the correlation between the nine related processes and the retrieval process is determined by the result of a user survey, and the results are evaluated as a benchmark. Three structure-based and one behavior-based process similarity query algorithms are selected to evaluate their accuracy and efficiency. The experimental results show their respective applicable scenarios. The proposed benchmark data set and related algorithm codes have been published online and can be downloaded by researchers.
【作者单位】：浙江工业大学计算机科学与技术学院;
【基金】：国家自然科学基金资助项目(61602411,61272308) 浙江省自然科学基金资助项目(LY15F020030) 浙江省重大科技专项重点工业资助项目(2015C01034,2015C01029) 杭州市重大科技创新资助项目(20152011A03)~~
【分类号】：TP301.1;TP391.3

【相似文献】