Identifying the structural characteristics of a collaboration network that influence firms' exploratory and exploitative innovation performance can help these firms enhance the output of innovation achievements and their core competitiveness. Based on 14,790 issued patents of 281 firms in the artificial intelligence industry in China, this study explores the complex nonlinear relationship between the structural characteristics of inter-organizational collaboration networks and firms' exploratory and exploitative innovation performance by using clustering algorithms and classifications based on information entropy or the Gini index. The results indicate the following: (1) The four characteristics of degree centrality, closeness centrality, the local clustering coefficient, and structural holes affect the exploratory and exploitative innovation performance of firms. (2) In different firm clusters, there are different characteristic combinations that provide firms with various development strategies to improve this performance. (3) There are different paths that firms can take to improve this performance, which should be comprehensively considered along with the development goals of firms.
Keywords: collaboration network; exploitative innovation performance; exploratory innovation performance; information entropy; machine learning algorithms.