Temporal knowledge graph completion is a technique that uses existing knowledge to predict or infer the missing information in the temporal knowledge graph. It combines the technical features of knowledge graph completion and time series analysis to deal with entities and relationships that change over time. The existing temporal knowledge graph completion technology fails to make effective use of the special relationship between relations and time series information, and it is difficult to fully represent the complex relationships existing in the graph. In order to solve the above problems, the model based on time probability box embedding (TPBoxE) was proposed. Firstly, the entities and relationships in the temporal knowledge graph are represented in the vector space by box embedding, so as to complete the static part of the temporal knowledge graph. Secondly, the head and tail entities that exist at the same time in a given time period are selected, and the completed static parts are filtered according to the time information of the entities. Finally, the Bayesian classification method is used to fully mine the time features hidden in the relationship, and the completion results are obtained by combining the confidence scores of the static parts. The link prediction task test of the proposed model on YAGO11k, WIKIdata12k, ICEWS18 and GDELT datasets shows that the proposed model has better performance than the existing excellent models, which proves the effectiveness and advancement of TPBoxE.