Abstract: The global patent cooperation network helps to reveal the law of cooperation. Existing research stays at the stage of describing the general characteristics of the overall network and how the network structure affects individual innovation behavior, and does not involve the influencing mechanism of the network structure on the overall network evolution. Based on OECD (Organization for Economic Co-operation and Development) patent data, the Temporal Exponential Random Graph Models (TERGMs) is used to explore the influencing mechanism of network structure and proximity on dynamic evolution of the global patent cooperation network. The results show that the global patent cooperation network presents weak agglomeration and decentralization in spatial situation. Between-centric countries are concentrated in North America, Western Europe and Oceania, presenting an unbalanced pattern of “three pillars”. The patent cooperation network presents a core-peripheral structure, with the core shifting from “US-Europe” to “US-Europe-Asia”. The global patent network shows transitive characteristics and star-shaped structures. Technological proximity facilitates the absorption and reorganization of heterogeneous knowledge among countries, and cultural proximity avoids risks caused by differences in patent contracts and legal standards between the two parties, which has an impact on the global patent cooperation network.

Key words: global patent innovation network, network structure, multi-dimensional proximity, TERGMs