Towards Feature Location in Models through a Learning to Rank Approach

In this work, we propose a feature location approach to discover software artifacts that implement the feature functionality in a model. Given a model and a feature description, model fragments extracted from the model and the feature description are encoded based on a domain ontology. Then, a Learning to Rank algorithm is used to train a classifier that is based on the model fragments and feature description encoded. Finally, the classifier assesses the similarity between a population of model fragments and the target feature being located to find the set of most suitable feature realizations. We have evaluated the approach with an industrial case study, locating features with mean precision and recall values of around 73.75% and 73.31%, respectively (the sanity check obtains less than 35%).