License recommendation for open source projects in the power industry

Abstract

Context:Establishing secure and appropriate licensing procedures for open-source software is essential in the development of a decentralized renewable energy system within the smart grid industry. Nonetheless, software developers in the power industry encounter obstacles in comprehending and electing licenses on account of factors such as resemblances in terms, intricacies of the law, compatibility of licenses, and the slow development of the open source movement in the power industry. Objective:This paper aims to comprehensively examine the licenses of open source projects in the power industry, which is essential for the completion and popularity of projects. A novel framework consisting of two stages (i.e. data processing and recommendation) is proposed to analyze the current situation of open source license selection in the power industry. Method:By analyzing 274,442 open source repositories related to 40 electricity-related keywords from GitHub, we developed a machine learning-powered license recommendation methodology. We first employed the K-means method to cluster the selected repositories and identified 6 major clusters. Next, we utilized the random forest method to predict licenses for new repositories based on the clustering results. We evaluated the accuracy of the model by testing it on training and testing datasets and achieved 96% accuracy. Results:We found that open source repository clusters in the power industry have distinct licensing preferences reflecting their unique objectives, with MIT being the most popular due to its permissiveness, and GPL-3.0, Apache-2.0, and BSD-3-Clause being favored by clusters valuing copyleft principles, closed-source derivatives protection, and control over software use, respectively. In addition, the study recognizes the content of open source projects as a meaningful indicator for license recommendation. Conclusion:These insights substantially enhance comprehension of the distribution and the selection of open source licenses in the power industry, potentially aiding future research on license recommendation in this field.