Drug interaction prediction method based on relation perception network

The invention discloses a prediction method for drug interaction based on a relation perception network, which comprises the following steps of: in a multi-relation DDI network, obtaining embedding of drugs through two steps: firstly, capturing multi-relation information among the drugs through TransD, and learning an embedding relation of each entity; secondly, the medicine obtains the embedding of a network structure through a relation perception information spreading mechanism; and the drug inserts obtained in the two steps are connected in series to obtain the final insert. Multi-relation embedding and network structure embedding are combined, so that the extracted medicine information is richer; multi-relation information between drugs can be captured by using a TransD model, and potential embedding of each entity can be learned. And the performance of the model can be effectively improved. 本发明公开了一种基于关系感知网络的药物相互作用的预测方法，包括在多关系DDI网络中，将药物通过两个步骤得到药物的嵌入，第一步通过TransD捕获药物之间的多关系信息，并学习每个实体的嵌入关系。第二步是药物通过关系感知信息传播机制得