The invention discloses a link prediction method and system for a clinical temporal knowledge graph. The method comprises first obtaining the clinical temporal knowledge graph, vectorizing an entity and complex semantic relationship, adding context information, performing serialized learning, performing incremental computation on the temporal knowledge graph, extracting a feature vector from the temporal information, establishing a link prediction model for the clinical temporal knowledge graph, wherein the link prediction model includes a triplet vectorization layer, a sequence incremental learning layer, a sequence feature combination layer and an output layer; performing link prediction on the clinical temporal knowledge graph according to the link prediction model. The method, by usingan incremental LSTM model, highlights the semantic and temporal information implied in a clinical fact, mine the dependence information before and after the semantic and temporal information by usingserialized learning, compensating for the low accuracy of time-sensitive knowledge graph prediction of a traditional link prediction model, improve the accuracy of knowledge graph link prediction.