Heterogeneous collaboration patterns and radical innovation performance-a data-driven analysis from specialized, refined, differentiated, and innovative enterprises.

With intensifying market competition, the characteristics of knowledge elements and collaboration networks under the deep integration patterns of industry, universities, and research institutes have significant impacts on radical innovation performance, which has profound implications for enterprises striving to establish a sustainable competitive advantage.This study used Specialized, Refined, Differentiated, and Innovative enterprises as the research object, combining machine learning methods such as the classification and regression tree algorithm and Bayesian network analysis to explore the interactive mechanism between the structural characteristics of internal knowledge elements and external collaboration networks of enterprises adopting different collaboration patterns, such as industry–university, industry–research institute, and industry–university–research institute, to provide an in-depth analysis of the multiple pathways for enterprises to improve their radical innovation performance. The research results verified and found the following: (1) Enterprises adopting industry-university-research institute collaboration pattern are more likely to achieve high radical innovation performance than those adopting industry-university or industry-research institute collaboration patterns. (2) The structural characteristics of different knowledge elements and collaboration networks have a complex nonlinear impact on enterprises’ radical innovation performance, with knowledge diversity being the most important factor. (3) The influence of the characteristic variables of enterprises adopting heterogeneous collaboration patterns on radical innovation performance shows significant differences, and appropriately utilizing structural holes can facilitate enterprises to conduct innovative activities. Enterprises of heterogeneous collaboration patterns can identify the optimal development pathways tailored to their specific conditions and strategic goals, as well as strategically adjusting their resource allocation from the perspective of interaction between knowledge elements and collaboration networks. In this way, the goal of enhancing radical innovation performance can ultimately be achieved through “different paths leading to the same destination”.