Compact memetic algorithm-based process model matching

Process models can be created for different purposes or in different contexts, and the need for the comparisons of process models foster research on process model matching, which refers to the process of determining the correspondences between semantically identical activities of different process models. Despite the growing number of process model matchers, the process model matching contests reveal that the effectiveness of state-of-the-art process model matchers is still low, i.e., the obtained process model alignment contains many irrelevant correspondences. Thus, how to improve the quality of process model alignment becomes one of the main challenges in the process model matching domain. Being inspired by the success of memetic algorithm (MA) in the alongside related domain such as schema matching and ontology matching, in this work, a compact memetic algorithm-based process model matching technique (CMA-PMM) is proposed to efficiently determine the process model alignment. In particular, a new activity similarity measure is used to determine the similar activities, a optimal model for process model matching problem is constructed, and a CMA is presented to efficiently solve the process model matching problem. CMA can simulate the behavior of population-based MA by employing the probabilistic representation of the population, thus comparing with MA, CMA is able to significantly save the memory consumption without sacrificing the solution's quality. In the experiment, three tracks provided by process model matching contest (PMMC), i.e., university admission processes (UA), birth registration processes and asset management, are utilized to evaluate the performance of CMA-PMM. Comparisons among evolutionary computation-based matchers, PMMC's participants and CMA-PMM show the effectiveness of our proposal.