构建数字创新网络已成为制造业企业提升数字创新绩效的重要途径，以社会网络理论和结构洞理论为基础，基于1985—2022年国家知识产权局专利数据库中的14631条数字专利信息，采用社区发现算法刻画数字创新网络，运用K-means聚类算法将这些网络划分为不同类型，并利用CART决策树算法挖掘不同类型数字创新网络特征与企业数字创新绩效之间的复杂非对称关系。研究发现：①共有4种不同类型的数字创新网络，即边缘探索型、核心枢纽型、稳健合作型和紧密协作型，不同类型数字创新网络因结构多样化展现出显著的数字创新绩效差异，即核心枢纽型高数字创新绩效占比最高，边缘探索型高数字创新绩效占比最低；②差异化网络特征配置能使企业获得不同水平的数字创新绩效，呈现出“同途异策，成效迥异”特征，但合作强度具有普适性，是影响不同类型数字创新网络形成差异化数字创新绩效的核心因素；③边缘探索型主要由“低合作强度—高中介中心性”特征驱动，核心枢纽型主要以“高合作强度”为核心驱动力，稳健合作型主要由“合作强度适中—高聚集系数”特征驱动，紧密协作型主要由“合作强度适中—度中心性高”特征驱动。研究结论可为我国传统制造业企业根据自身特点优化数字创新网络结构以及达成高数字创新绩效目标提供新路径与实践工具。

The continuous enhancement of enterprise digital innovation performance has become the sole pathway to achieving high-quality economic growth. Collaboration through network construction is increasingly recognized as a vital avenue for traditional manufacturing enterprises to engage in digital innovation activities. However, in practice, not all manufacturing enterprises can attain the anticipated outcomes of digital innovation by merely establishing digital innovation networks. This paper, therefore, delves into how traditional manufacturing enterprises can effectively utilize digital innovation networks to bolster their innovation performance and secure high-quality economic development.Current research tends to focus on the singular impact of a particular network characteristic on enterprise digital innovation performance, with scant exploration of how multiple network factors collectively drive the performance of enterprise digital innovation. Moreover, existing studies often treat innovation networks as homogeneous, neglecting the nuanced topological differences among them. Furthermore, the majority of the literature employs conventional empirical research paradigms, testing the relationship between networks and enterprise digital innovation performance through preconceived theoretical models and regression tests. The use of machine learning methods to investigate the complex interplay of variables is relatively rare.
Given that traditional research methods struggle with identifying heterogeneous digital innovation networks and the asymmetric relationships among multiple variables, this study selects six significant characteristic variables: degree centrality, betweenness centrality, closeness centrality, structural holes, cooperation intensity, and clustering coefficient. Drawing on social network theory and structural hole theory, it analyzes 14 631 digital patent records from the State Intellectual Property Office's patent database spanning from 1985 to 2022. Then the study employs community detection algorithms to delineate digital innovation networks and use the K-means clustering algorithm to categorize these networks into distinct types. The CART decision tree algorithm is then applied to explore the intricate asymmetric relationships between the characteristics of different types of digital innovation networks and digital innovation performance.
This paper reveals several key findings. Firstly, there are four distinct types of digital innovation networks: edge exploration, core hub, robust cooperation, and close collaboration. These network types exhibit significant variations in digital innovation performance due to structural diversification, with the core hub type having the highest proportion of high digital innovation performance and the edge exploration type having the lowest. Secondly, a differentiated configuration of network features enables enterprises to achieve varying levels of digital innovation performance, highlighting the adage “same path, different strategies, different outcomes.” However, cooperation intensity is a universal factor that significantly influences the formation of differentiated digital innovation performance levels across network types. Thirdly, the edge exploration type is primarily driven by the “ low cooperation intensity - high betweenness centrality” feature, the core hub type by “high cooperation intensity”, the robust cooperation type by “moderate cooperation intensity-high clustering coefficient”, and the close collaboration type by “ moderate cooperation intensity-high centrality”. These findings offer a novel perspective and practical tools for Chinese traditional manufacturing enterprises to optimize their digital innovation network structures according to their unique characteristics and achieve high digital innovation performance goals.
The findings offer practical insights for Chinese manufacturing firms' digital innovation efforts. It highlights the importance of core hub-type digital innovation networks for high performance, suggesting that companies should focus on building these networks while the government should support digital infrastructure and encourage cross-industry cooperation. The study also reveals a complex relationship between digital innovation networks and performance, emphasizing the universal impact of cooperation intensity. Enterprises should adjust their roles in different network types so as to enhance their digital innovation performance. Cooperation with partners is crucial, and establishing close relationships can improve digital innovation efficiency. Government departments can facilitate this by creating platforms for information and resource sharing. Lastly, the study indicates that digital innovation networks' openness and dynamism can reduce reliance on structural holes, allowing companies to leverage digital technologies for efficient resource allocation and decision-making.