Classification-based Mining of Reusable Components on Software Product Lines

Abstract

Software Product Lines and Component-based systems can be combined to maximize reuse in a predictable and opportunistic manner. When a product line is built for a certain subdomain within a more generic domain, future needs from a closely subdomain may be fulfilled by mining the line's internal components to build a new product line. In this work, we present an approach to classify internal and external (third party) reusable components into a repository, by applying a K-Nearest Neighbors strategy, as a support for building new product lines. Natural language techniques and the WordNet lexical database are also used to process information from software components. We validate the approach with an experiment based in a dataset of external third-party components and reusable components from a product line that we built in the geographic subdomain of marine ecology.