The ability to continuously detect the interesting subgraph patterns over dynamic graph-structured data in real time is crucial to many applications. Generally, the edges of such subgraphs have temporal relation. In this paper, the stream model is used to describe the dynamic graph, in which the edges are processed in chronological order. In order to reduce the intermediate results, we propose a topological sorting algorithm that considers both the probabilistic information and the temporal relation of the query graph. Then we present the join-cache tree based query algorithm which can implement the match of topology structure and the verification of temporal relation simultaneously. The experiments performed on real-world dataset demonstrate that the proposed algorithm has 10x speedups over the previous approaches.