A Pairwise Approach for Model Merging

Abstract

There are several software engineering activities that require merging a set of models to produce a single one. In practice, models are often merged in a pairwise way, without considering the order in which models are combined. In this case, the quality of the results is not always guaranteed as it depends on the order of merging. The approach presented in this paper aims to improve the results, by considering the order of merging. It involves an iterative process, which is repeated until merging all models. In each iteration, we first compare the set of input models to measure the similarity degree of each pair of them. Then we combine a subset of these pairs of models, such that the sum of their similarity degrees is maximal.