Recommending Whom to Follow on GitHub

Abstract

Social networking activities on GitHub allow the construction of interesting interaction networks. One such network is `follow-network' which enables an effective information dissemination process. As a result, GitHub users are bombarded with stacks of information which also puts the users at risk of information overload. This motivates us to recommend the relevant user such that developers are only provided with the relevant information. In this work, we present an attributed network embedding based framework to recommend whom to follow on GitHub. This is a challenging task due to the complex social network structure of the developers. In particular, we first construct a developers' `follow-network'. Further, we extract the node embeddings of each node and feed these embeddings to a K-nearest Neighbour classifier. We validate our approach on the developers of three popular programming languages (C++, Python, and Java). We were able to achieve promising results with an F1-score of 72%.