Maintaining Third-Party Libraries through Domain-Specific Category Recommendations

Proper maintenance of third-party libraries contributes toward sustaining a healthy project, mitigating the risk it becoming outdated and obsolete. In this paper, we propose domain-specific categories (i.e., grouping of libraries that perform similar functionality) in library recommendations that aids in library maintenance. Our empirical study covers 2,511 GitHub projects and 150 domain-specific categories of Java libraries. Our results show that a system uses up to six different categories in their dependencies. Furthermore, recommending domain-specific categories is practical (i.e., with an accuracy between 66% to 81% for multiple categories) and its suggestion of libraries within that domain is comparable to existing techniques.