Energy efficient software development is a requirement to meet software quality standards. A great number of works have been done to enhance the level of information related to software energy consumption (SEC). They are generally focused on raw code data. These data can be profiled to predict SEC trends of future versions of a software. However, SEC works lack energy profiling with powerful predictive models. In this work, a deep learning-based SEC model is proposed. The model is than evaluated with 14 open-source projects. The experiment shows that deep learning performs better in SEC profiling than the alternatives such as random forest. Further, contrary to expectations, the success of the profiler is sensitive for the number of hidden layers of deep neural network.
CITATION STYLE
Öztürk, M. M. (2020). Deep Learning-Based Software Energy Consumption Profiling. In Lecture Notes on Data Engineering and Communications Technologies (Vol. 43, pp. 73–83). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-030-36178-5_7
Mendeley helps you to discover research relevant for your work.