Abstract: Temporal network can better describe the dynamic evolution properties of topology construction among nodes in complex network. Taking into account the mutual influence of nodes in different time layers, and being inspired by inter-layer temporal correlation coupling relationship in multilayer network, this paper proposes a kind of temporal network with coupling based on vector autoregressive (VAR) model. How to construct temporal network is given and the empirical application in four stock markets including NDX100, S\&P 500, ZSSE100 and SSE180 is explored. Compared with the literature such as 15 and 16, the study shows that the proposed new network exhibits more obvious advantages on resolution power of nodes (stocks) importance ranking and the in-sample and out-sample performance of portfolio optimization. Meanwhile, this paper also discusses how to determine ``peripheral'' stocks on the basis of the nodes importance sequence. It is apparent that our study could further enrich temporal network theory and provide new technical tools for financial market research.