Time sequence knowledge graph reasoning method

The invention provides a time sequence knowledge graph reasoning method, and belongs to the technical field of knowledge graph reasoning. The method comprises the following steps of: designing a new graph convolution model adopting a joint volume combiner, and obtaining embedding of entities and relationships on a knowledge graph by utilizing improved graph convolution at each timestamp, so that structural dependence and potential relationships of concurrent facts are fully captured; at the same timestamp, the event attribute information of time is obtained according to all occurred facts to supplement node embedding, and the contingency of the event is captured; gradually updating nodes and relationships on the time sequence by using a gate cycle component; prediction scores of historical entities and historical relations are increased according to historical data, and meanwhile repeatability of events is captured. According to the method, the problems of insufficient knowledge graph node neighborhood inform