Recommending Component by Citation: A Semi-supervised Approach for Determination

Reusing existing components can help developers improve the development productivity as well as reduce the cost. Reuse repositories in this scenario act as a fundamental facility for acquiring needed components. While retrieving components in reuse repositories, developers often face the problem of choosing components from candidates that provide similar functionalities. To address the problem, this paper proposes a semi-supervised method to recommend developers components in reuse repositories. Different from existing rating based recommendation approaches that often suffer from the lack of user ratings, our approach calculates the recommendation probabilities of components based on their citations on the Internet. The citations are acquired through the websites (called host in this paper) that are associated with the components. Using a random walk algorithm, the associations between components and hosts are explored with recommendable components identified. We implemented our approach in a prototyping system based on which we conducted an experimental study to evaluate our approach. The experimental results demonstrate that our approach can accurately recommend components and thus has the potential to assist developers in reuse. Keywords-software reuse;component recommendation;reuse repository