Using Ant Colony Optimisation to Select Features Having Associated Costs

Alfredo Ibias, Luis Llana, Manuel Núñez

Abstract: Software Product Lines (SPLs) strongly facilitate the automation of software development processes. They combine features to create programs (called products) that fulfil certain properties. Testing SPLs is an intensive process where choosing the proper products to include in the testing process can be a critical task. In fact, selecting the best combination of features from an SPL is a complex problem that is frequently addressed in the literature. In this paper we use evolutionary algorithms to find a combination of features with low testing cost that include a target feature, to facilitate the integration testing of such feature. Specifically, we use an Ant Colony Optimisation algorithm to find one of the cheapest (in terms of testing) combination of features that contains a specific feature. Our results show that our framework overcomes the limitations of both brute force and random search algorithms.