Using the ModelSet dataset to support machine learning in model-driven engineering

Abstract

The availability of curated collections of models is essential for the application of techniques like Machine Learning (ML) and Data Analytics to MDE as well as to boost research activities. However, many applications of ML to address MDE tasks are currently limited to small datasets. In this demo paper, we will present ModelSet, a dataset composed of 5,466 Ecore models and 5,120 UML models which have been manually labelled to support ML tasks (http://modelset.github.io). ModelSet is built upon the models collected by the MAR search engine (http://mar-search.org), which provides more than 500,000 models of different types. We will describe the structure of the dataset and we will explain how to use the associated library to develop ML applications in Python. Finally, we will describe some applications which can be addressed using ModelSet.