Taking MIT as the research object, this paper studies the evolution characteristics of the network density of Industry-University-Research cooperation innovation network in different time windows since 1998. The relationship between innovation network density and innovation performance is analyzed and the MRQAP model is used to make further empirical analysis. The results show that the density of innovation network is increasing to the threshold first and then slowly decreasing; the density of innovation network and innovation performance show the characteristics of "anti C type", that is, after the density reaches the threshold, the innovation network will show obvious redundancy, which will cause unreasonable resource allocation, repeated information acquisition, increased communication cost and reduced network activity. It is suggested that the micro organizations of cooperation in Industry-University-Research should scientifically identify and actively optimize the cooperative innovation network relationship in order to optimize the innovation performance. This paper is of great significance to how to optimize and improve the innovation performance of the cooperative innovation network of production, learning and research in micro organizations.