A GitHub Project Recommendation Model Based on Self-Attention Sequence

In this paper, we propose a new GitHub recommendation model based on self-attention mechanism that considers user's historical operation sequence. It includes a project embedding layer, multiple encoder layers and a prediction layer. The main idea of our method is to add a position vector to the original project embedding vector to indicate the sequence information of the current project in the user's operation sequence. And considering that the next possible operation project of the user is largely determined by the previous project, model includes a residual connection to the encoder layer. Evaluated our method on a variety of large, real-world datasets, and it shows quantitatively that our outperforms alternative algorithms, especially on sparse datasets. The model can capture personalized dynamics and is able to make meaningful recommendations.