Session-based social and dependency-aware software recommendation

With the increase of complexity modern software, social collaborative coding and reuse open source software packages become more popular, which thus greatly enhances development efficiency quality. However, explosive growth exposes developers to challenge information overload. While this can be addressed by conventional recommender systems, they usually do not consider particular constraints such as influence among dependency relations packages. In paper, we aim model dynamic interests with both constraints, propose Session-based Social Dependency-aware Recommendation (SSDRec) model. This integrates recurrent neural network (RNN) graph attention (GAT) into a unified framework. An RNN is employed short-term in each session two GATs are utilized capture from friends dependent packages, respectively. Extensive experiments conducted on real-world datasets results demonstrate that our significantly outperforms competitive baselines.