Using the patent data of 171 enterprises in Information Technology (IT) industry from 2007 to 2016, this paper uses negative binomial regression model, empirically tests the influence of the shorter average path length and clustering coefficient of the industry-university-research cooperation network’s small-world on enterprise innovation, and the moderating effect of knowledge base. The results show that the shorter average path length is beneficial to the improvement of enterprise innovation, but the impact of clustering coefficient on the enterprise innovation in an inverted-U curvilinear way. The knowledge base width moderates positively the effect of average path length on the enterprise innovation, but moderates negatively the effect of clustering coefficient on enterprise innovation.