[Objective] Promote exchanges and cooperation among scientific researchers to maximize scientific research efficiency. This paper proposes an improved translation model TransTopic for predictive research in scientific research cooperation in the field of stem cells.
[Methods] TransTopic aims to uniformly map the nodes and edges in the scientific research cooperation network to low-dimensional vectors. First, use the LDA topic model to extract the topic distribution features of the paper description documents on the edges, and then use the deep autoencoder to encode the topic features into edge vectors. The node vector is obtained based on the translation mechanism, and finally the scientific cooperation prediction is achieved through the semantic calculation between the vectors.
[Results] TransTopic's AUC (95.21%) and MeanRank (17.48) indicators for link prediction perform best, and the topic prediction accuracy rate reaches 86.52%.
[Limitations] The collaborative prediction method only considers a one-step translation path, and information such as the author's institution, research interest, and publication level are not fully utilized.
[Conclusions] The prediction method based on the translation model can effectively complete the scientific research cooperation prediction in the field of stem cells.