【摘要】 创新社团是应对绿色创新、提升绿色创新能力的重要组织模式。不同的绿色创新社团绩效存在明显差异，但当前研究对于创新社团这一中观对象缺少深入分析。基于此，构建“结构嵌入-关系嵌入-知识嵌入”的社团创新绩效理论分析框架，从中观视角切入，以2010年至2021年的中国绿色专利发明授权数据作为样本数据，通过Louvain算法对绿色创新网络进行社团识别，运用k均值聚类算法将结构相似的社团归为同一个数据簇，对每一个簇采用XGBoost模型对前因变量的特征重要性进行筛选，运用模糊集定性比较分析方法分别讨论各数据簇内社团创新绩效的组态效应，挖掘结构异质的社团绿色创新绩效的差异化提升路径。结果表明：按照结构异质性绿色创新社团可分为3种类型，即权力集中类社团、紧密合作类社团及复杂结构类社团，不同类型社团的嵌入性特征存在明显差异；权力集中类社团存在3条绿色创新绩效提升路径，包括深度合作驱动下的领域专精路径、深度合作驱动下的共创深耕路径及核心主体驱动下的综合能力路径；紧密合作类社团产生高绿色创新绩效有3条路径，即行业优势驱动下的同质关系型路径、行业优势驱动下的精准合作型路径及规模化多样化异质化并驱型路径；复杂结构类社团产生高绿色创新绩效有5条路径，即行业优势驱动下的多方聚力型路径、行业优势驱动下的多元技术型路径、强关联广合作多元差异化并驱型路径、深度合作驱动下的技术深耕型路径及深度合作驱动下的多方探索型路径。更多还原

【Abstract】 Innovation communities represent a key organizational model for addressing green innovation and enhancing green innovation capabilities. However, there are significant differences in performance among various green innovation communities, and current research lacks in-depth analysis of innovation communities as a meso-level subject. Based on this, this paper constructs a theoretical framework for community innovation performance analysis,incorporating structural embeddedness, relational embeddedness, and knowledge embeddedness, carries out study from a mesoscopic perspective, taking China’s green patent invention authorization data from 2010 to 2021 as sample. First,the Louvain algorithm is utilized to identify the associations within the green innovation network. Subsequently, the k-mean clustering algorithm is applied to group structurally similar associations into the same data cluster. Next, the XGBoost model is used to filter the importance of the characteristics of antecedent variables in each cluster. Lastly, the fuzzy-set qualitative comparative analysis method is employed to separately discuss the configurational effects of the innovation performance of communities within each data cluster, and to explore the differentiated improvement paths of the green innovation performance of communities with structurally heterogeneous characteristics. The findings reveal that green innovation communities can be categorized into three types based on structural heterogeneity: powercentralized communities, close-collaboration communities, and complex-structure communities—each characterized by distinct patterns of embeddedness. Power-centralized communities have three performance enhancement pathways:a domain-specialization path driven by deep collaboration, a co-creation and deep-cultivation path driven by deep collaboration, and a comprehensive capability path driven by core entities. Close-collaboration communities have three pathways: a homogeneous relationship path driven by industry advantages, a precise collaboration path driven by industry advantages, and a parallel scale-diversification-heterogeneity path. Complex-structure communities have five pathways: a multi-party synergy path driven by industry advantages, a multi-technology path driven by industry advantages, a strong association-broad collaboration-multi-dimensional differentiation parallel path, a technology deep-cultivation path driven by deep collaboration, and a multi-party exploration path driven by deep collaboration.更多还原