Research on Innovation Network Features of Patent-Intensive Industry Clusters and Their Evolution

Abstract

In the contemporary economic landscape shaped by globalization and digital transformation, patent-intensive industries have emerged as critical engines for enhancing national competitiveness. This study analyzed 98,464 collaborative patent application records (2012–2023) from listed companies in patent-intensive sectors, sourced from the China National Intellectual Property Administration (CNIPA) database. Through kernel density estimation, social network analysis, and community detection techniques, we examined the evolutionary trajectories of innovation networks and spatial patterns within these industrial clusters. Our findings indicate a notable spatial agglomeration trend in patent-intensive industries, exhibiting a prominent “core-periphery” structural feature. The core nodes of this cluster network closely align with economically developed regions, and the network structure has gradually shifted from a triangular framework supported by Beijing-Tianjin-Hebei, Yangtze River Delta, and Pearl River Delta to a diversified multilateral framework. Moreover, the community structure of the collaborative network within China’s patent-intensive industrial clusters exhibits distinct characteristics driven by technological relevance and strategic synergy, rather than strictly adhering to the principle of geographical proximity. These discoveries not only enrich the application of innovation network theory in the specific context of China, but also provide valuable guidance for cluster enterprises in selecting partners and achieving collaborative innovation.

1. Introduction

In the contemporary economic landscape shaped by intertwined globalization and digitalization, patent-intensive industries have emerged as pivotal engines for innovation-driven growth and economic diversification, becoming the focal point of next-generation industrial cluster development. The 2023 Special Action Plan for Patent Commercialization (2023–2025) explicitly calls for accelerating patent conversion and industrialization of high-value patents, emphasizing quality enhancement to provide technological underpinnings for clustered development. Building on this framework, The 2024 Implementation Plan for Building an Intellectual Property Powerhouse emphasizes accelerating high-value patent commercialization, refining IP frameworks in emerging sectors, and advancing patent-intensive industrial clusters through integrated strategies. The 2024 Annual Guidelines for Promoting High-Quality Intellectual Property Development adds specificity by advocating lifecycle optimization of IP creation, utilization, protection, and management. This approach seeks to unlock innovation potential and deepen IP-industry synergy, thereby propelling high-quality cluster development. Collectively, these policy initiatives underscore how patent-intensive industrial clusters, leveraging their unique strengths, have become indispensable drivers of economic advancement [1]. The alignment of these industrial clusters with innovation-driven development trends holds critical significance for facilitating economic growth pattern transformation and enhancing the operational efficiency of national innovation systems [2].

Existing research predominantly examines the connotations and characteristics of patent-intensive industries [3], with limited scholarly attention directed toward their clustered formations. In the knowledge economy era, these sectors symbolize innovation vitality and developmental potential due to their concentrated innovation activities. Beyond technological excellence, they drive holistic industrial development through resource aggregation, structural optimization, and value chain upgrading, serving as robust engines for regional prosperity and competitiveness enhancement. Therefore, deepening understanding of their formation mechanisms and evolutionary trajectories is crucial for effective policy formulation and high-quality economic development. Amid escalating tech competition, knowledge-driven emerging industrial clusters undergo spatial and organizational reconfiguration across multiple scales [4]. Spatial arrangements and organizational patterns of patent-intensive clusters are also evolving, manifesting in inter-regional collaboration dynamics, value chain integration, and innovation resource diffusion. Inadequate comprehension of these evolutionary patterns may impede regional economic coordination and constrain innovation capacities. For instance, fragmented inter-regional collaboration mechanisms lead to redundant resource allocation, while ambiguous competitive boundaries hinder technological exchange [5]. Incomplete value chains limit cluster expansion, and uneven innovation resource distribution creates regional capacity gaps [6]. Thus, Urgent research is warranted to elucidate evolutionary dynamics, providing empirical foundations for coordinated regional development and innovation capacity enhancement. This study addresses critical knowledge gaps to inform evidence-based policymaking in the rapidly evolving landscape of innovation-driven industrial clusters.

The current economic development paradigm increasingly emphasizes shifting from “factor-driven” to “innovation-driven” growth. Simultaneously, “location-based space” is evolving into “flow space.” This evolution dissolves regional boundaries in technological innovation while transforming innovation models from closed linear approaches to open network-based frameworks [7]. In this context, establishing and sustaining efficient innovation networks has become pivotal for innovation success. As open innovation deepens, patent-intensive industrial clusters emerge as both knowledge innovation hubs and critical nodes in innovation networks. These networks extend beyond internal R&D, production, and marketing processes to encompass upstream and downstream enterprises and institutions across value chains, forming complex multi-tiered innovation ecosystems [8]. Through network linkages, clusters significantly accelerate knowledge, technology, information flows, optimize resource allocation, and expedite innovation commercialization. Conversely, network development strengthens clusters’ innovation capacity and competitive advantages, elevating their position in global value chains. Studying these innovation networks offers critical insights into evolutionary patterns under globalization and open innovation, while providing practical implications for regional economic coordination and overall innovation enhancement. This research addresses key knowledge gaps to inform evidence-based strategies in the rapidly evolving innovation landscape.

Given this context, the study utilizes co-invention patent data (2012–2023) from the China National Intellectual Property Administration to investigate cluster innovation networks. It analyzes the connection characteristics within industrial cluster innovation networks and further reveals the evolutionary trajectories of their spatial patterns. The research seeks to address three questions: (1) What evolutionary characteristics define the innovation networks and spatial patterns of industrial clusters? (2) How do innovation linkages manifest at local versus cross-regional levels within cluster networks? (3) What are the changing characteristics of community structures in patent-intensive industrial cluster collaboration networks? Based on this, the present study offers three innovative aspects and contributions. Firstly, while existing research has largely focused on macro-level industrial distribution or firm-level innovation behavior, there is a lack of systematic investigation into innovation networks in patent-intensive industries at the cluster scale. This study adopts a cluster perspective, starting from the spatial agglomeration patterns of patent-intensive industries and delving into the internal innovation networks and community structures within clusters. This approach allows for a multi-level analysis from macro to micro scales. It not only broadens the theoretical scope of innovation network research but also offers a systematic foundation for scientific planning and precise policy-making in patent-intensive industries. Secondly, previous studies on innovation collaboration have often overlooked structural differences across spatial scales. By integrating patent data from listed companies with spatial and network analysis, this study accurately identifies distinct patterns of local versus cross-regional innovation collaboration at different spatial scales. This contributes not only to refining innovation network theory within the Chinese context but also offers practical insights for cluster firms in selecting partners and achieving synergistic innovation. Thirdly, unlike traditional research that overemphasizes geographical proximity, this study applies a community detection algorithm to examine patent-intensive industrial clusters in China. From both structural and temporal evolutionary perspectives, it reveals a shift in the innovation network—from geographically clustered communities to those dominated by technological linkages. This analysis enhances the understanding of innovation network evolution mechanisms and offers a new pathway for interpreting cluster-based innovation ecosystems.

2. Connotation Differentiation and Literature Review

2.1. Connotation Differentiation of Patent-Intensive Industrial Clusters

Current research on patent-intensive industrial clusters primarily focuses on related fields or expands to broader domains such as strategic emerging industries and innovation-driven sectors [9,10]. Liu et al. (2024) argues that innovative industrial clusters, as a specific organizational form, prioritize strengthening innovation collaboration and cooperative mechanisms among affiliated enterprises within industrial chains [11]. The cluster model aims to foster cross-regional leadership and development through business expansion. Besides, a limited number of studies focus on niche sectors of specific industries, such as semiconductors and biomedicine [12,13]. For instance, Fu et al. (2022) define pharmaceutical industry clusters as geographically concentrated industrial networks [14]. The networks consist of diverse entities including pharmaceutical firms, research institutions, universities, service intermediaries, and local governments. Through collaborative efforts, these actors form close ties and establish efficient interactive networks via innovation systems. Lu and Zhong (2024) define the digital industry cluster as an urban agglomeration guided by China’s new development philosophy [15]. It integrates core digital economy enterprises specializing in manufacturing, services, tech applications and data-driven sectors, along with cooperating enterprises of various scales and supporting institutions. These components form a tightly connected network through interdependent relationships, collectively creating a dynamic urban cluster. Such studies often emphasize technological innovation, industrial upgrading, and intra-cluster knowledge spillover effects, while overlooking the more distinctive features and demands of patent-intensive industrial clusters—namely, the high density of patent activities within clusters and the competitive market advantages they confer on member enterprises through intellectual property engagement.

To address gaps in existing research, this paper defines the connotation of patent-intensive industrial clusters based on the uniqueness of patents as follows. Patent-intensive industrial clusters refer to geographic concentrations of innovative enterprises that rely on invention patents as their core competitive advantage and engage in market competition through intellectual property rights. These enterprises, aligned with innovation-driven development paradigms, serve as core actors driving technological breakthroughs and economic growth within the clusters. Possessing high technological intensity, innovative spirit, and value-added capacity, they construct robust industrial and value chains through continuous technological innovation, intellectual property protection, and close industrial collaboration. Firms within the clusters form tight patent linkages, facilitating investment, technological exchange, and supply chain cooperation. Such linkages endow patent-intensive industrial clusters with exceptional cross-boundary collaboration potential and network-building capacity, enabling the formation of cluster networks that transcend geographic boundaries. Meanwhile, the clusters themselves provide spatial carriers and innovative ecosystems for collaborative development among enterprises.

2.2. Research on Innovation Networks of Patent-Intensive Industrial Clusters

2.2.1. Research Progress on Patent-Intensive Industrial Clusters

In the research evolution of patent-intensive industrial clusters, academic focus has gradually shifted from traditional industrial clusters to innovation clusters, and further deepened to patent-intensive industrial clusters centered on patent output. Since the early 20th century, industrial clusters have become a topic of common interest in geography, economics, and new economic geography. Scholars have developed cluster theories from perspectives such as neoclassical location theory, new institutional economics, and evolutionary economic geography, yielding abundant research outcomes. In 1890, Marshall first proposed the concept of “industrial districts.” He observed that within these districts, geographically concentrated groups of firms interacted through various forms of collaboration, such as subcontracting and joint ventures [16]. Through these interactions, they achieved external economies of scale. He identified external economies and scale economies as the economic drivers of industrial agglomeration. Becattini et al. (2004) applied Marshall’s industrial district theory and the concept of “flexible specialization” to thoroughly investigate the spatial agglomeration of small and medium-sized enterprises based on specialization and technological innovation [17]. This work significantly contributed to the revival of industrial district theory. In 1990, Porter introduced the concept of “industrial clusters.” He defined them as geographically concentrated groups of interconnected companies, suppliers, and institutions within a specific field [18]. These entities are linked through shared characteristics and complementary relationships in their domain. From a geographical perspective, industrial clusters represent a form of spatial agglomeration. This phenomenon is characterized not only by the geographic concentration of firms, but also by the organic integration of upstream and downstream enterprises within the industrial chain, along with related service agencies and support systems. From an economic standpoint, the formation and development of industrial clusters are the result of optimized resource allocation under market economy conditions. Various economic theories and models have been used to explain the drivers and effects of industrial clusters, such as external economies and internal economies, and transaction matching across different levels. Economic geography, from a static perspective, emphasizes the positive effects of clusters on enhancing firm productivity, accelerating knowledge spillovers, and expanding employment. Related research touches on themes such as “beyond clusters [19],” “local buzz-global pipelines [20],” and industrial chain resilience [21]. Furthermore, evolutionary economic geography has introduced a new approach for dynamic analysis in industrial cluster research. It emphasizes that the positive externalities arising from clusters may be evolutionary outcomes rather than merely initial causes [22]. Related studies, based on perspectives such as firm spin-offs [23], firm diversification, labor mobility [24], and social networks [25], have delineated the formation and evolution processes of industrial clusters across various sectors globally. This comprehensive examination covers the life cycle evolution, development trajectories, internal and external drivers, and support systems of clusters.

With the continuous expansion of regional economies, key resources such as land and labor have become scarce, leading to rising costs. These issues have pushed clusters to their saturation points, gradually eroding their competitive advantages and resulting in declining output efficiency. Based on extensive research by scholars domestically and internationally, innovative upgrading of industrial clusters is regarded as a crucial pathway to sustain cluster development and extend their life cycle [26]. Since China’s reform and opening-up, guided by the planned economy system, eastern China’s pioneering regions have actively explored a distinctive “cluster-based development” model. This initiative aims to achieve a strategic transition from “production clusters” to “innovation clusters” with Chinese characteristics. Compared to traditional clusters, innovation clusters exhibit significant embeddedness in knowledge networks and effects of innovative knowledge spillover, placing greater emphasis on intensive collaborative linkages and technological functional connections [27]. The formation of innovation clusters is a goal-oriented selection and organization process driven by innovation actors aiming to reduce uncertainties and enhance returns on innovation. Their dynamic mechanisms encompass self-organization systems, externally-driven systems, and hybrid-driven models. Academic research on innovation clusters can generally be categorized into two types. Firstly, studies discussing the definition, characteristics, and classification of innovation clusters, focusing on the development pathways and policy support for indigenous innovation clusters in China. Relevant research includes pilot industrial parks [28], comparisons of domestic and international development models, and the construction of innovation platforms [29]. Secondly, analyses systematically examining the internal logic and operational mechanisms of innovation cluster [30] development from perspectives such as industry-university-research collaboration, industrial chain integration, and firm motivations. Related studies involve formation mechanisms, learning mechanisms, and coordination mechanisms.

Amid rapid technological iteration and evolving cognitive paradigms in innovation, researchers and practitioners have categorized innovation clusters according to their distinct innovation capabilities and market performance. The World Intellectual Property Organization (WIPO) documents geographic regions with the highest density of inventors and scientific authors worldwide through patent applications and scientific publications [31]. It ranks the top 100 science and technology clusters. Ahoura et al. (2025) selected the six industries with the highest R&D intensity and examined the density of high-tech activities in regions based on employment and number of enterprises [32]. This approach aimed to explore the micro-locations of high-tech industrial clusters in the United States. Meanwhile, significant changes in the global competitive landscape have profoundly reshaped the industrial structures of major developed countries through networked and globalized economic activities. Developed economies such as Germany, the United States, and Japan have actively formulated national innovation strategies. They are committed to strengthening their global competitiveness by building world-class industrial clusters. China’s 19th National Congress of the Communist Party also introduced the policy concept of developing world-class advanced manufacturing clusters. Patent-intensive industries, with their unique development models and significant economic value, have become key carriers for building such clusters. Moreover, innovation activities in patent-intensive industries are traceable and quantifiable. These characteristics provide support for analyzing their cluster-based development models and upgrading pathways. A primary concern in this research is cluster identification. Scholars widely adopt methods such as location quotient, industry concentration ratio, kernel density analysis, and standard deviational ellipse to measure the spatial distribution of industrial clusters [33,34]. These methods serve as basis for identification. Previous literature on patent-intensive industrial clusters has focused largely on specific fields. Examples include semiconductors, integrated circuits, and biomedicine [35,36]. Research themes cover geographical distribution, functional mechanisms, and the evolution of spatial organizational structures. Additionally, some studies focus on industries that highly overlap with patent-intensive sectors. These include strategic emerging industries and high-tech industries. Research topics encompass ecological chain synergy, dynamic evaluation, and network effects. These research findings are equally valuable and can effectively complement studies on the cluster development of patent-intensive industries.

2.2.2. Research Progress on Innovation Networks

Innovation networks originated in the field of sociology, with their conceptual foundation established by anthropologists such as Nadel and Barnes in the mid-20th century. They systematically expanded the concept of networks, viewing them as connected graphs composed of vector segments linking nodes at different levels to depict relationships, state changes, and development trends among behavioral entities. By the 1970s, the concept of networks transcended disciplinary boundaries and was widely applied across various research fields. In the 1980s, scholars in economics and management began replacing individual nodes in social networks with enterprises to represent the complex relational networks among firms. In this context, Imai and Baba (1989) were the first to propose the concept of innovation networks [37]. They pointed out that network organizations represent a fundamental form of integration between organizations and markets, serving as an institutional arrangement to address systemic innovation challenges. Freeman (1991) built upon existing research, further citing and deepening the concept [38]. He explicitly defined innovation networks as a fundamental institutional framework for coping with systemic innovation, emphasizing the network architecture formed through collaborative innovation relationships among enterprises. Early innovation network theory focused on interactive learning and innovation linkages among entities within small-scale regions. It centered on micro-level actors such as enterprises and research institutions, highlighting the importance of local networks as core carriers of innovation activities and key elements of regional development. Within this theoretical framework, innovation resources such as knowledge, information, and technology are optimally allocated and efficiently circulated within the network. This promotes significant improvements in production efficiency and generates widespread synergistic effects. Through interaction and collaboration, innovation entities continuously grow, enhancing their knowledge levels and technological capabilities to achieve mutual progress.

Since the 1980s, economic geographers have begun using a network perspective to examine the spatial agglomeration and dispersion of economic activities. This approach has given rise to theoretical branches such as New Regionalism, Global Production Networks, and Relational Economic Geography. The relationships within innovation networks are complex and variable, comprising formal and informal ties. Formal relationships, bound by clear legal or contractual constraints, serve as stable and reliable links within innovation networks, providing a framework for stable collaboration. Informal relationships, relying on trust, shared values, and social capital among other non-institutional factors, bring flexibility and innovation momentum to the network. With the rise of the network paradigm, academia has placed high importance on leveraging networked advantages to promote the upgrading of innovation clusters. Relevant research covers cluster network organizations, knowledge network construction, multidimensional proximity [39], and network evolution [40]. Research indicates that within patent-intensive industrial clusters, competition and collaboration among innovation entities form a dynamically evolving innovation network. This network features complex structures with spatial diversity and diverse connectivity. Song et al. (2024) analyzed the formation and evolution mechanisms of the collaborative innovation network in China’s new energy vehicle industry [41]. They found that the network’s significant expansion is closely related to strong national policy support and the active participation of innovation actors. Turkina et al. (2016) examined the structure and evolution of formal enterprise networks across 52 aerospace clusters in North America and Europe [42]. They discovered a shift from local structures to geographically transcending, value chain-layered hierarchical structures. Barbosa et al. (2024) conducted an in-depth study on the origin and development of Spain’s innovation network based on historical patent data and social network analysis [43]. The research highlighted isolation and disconnection in the network’s emerging stage, revealing a lack of large connected components and international collaboration. It also pointed out that excessively long patent protection periods may hinder cooperation. Di et al. (2022) focused on the evolution of the global intellectual property network, revealing a distinct “core-periphery” pattern between 2000 and 2019 [44]. The distribution pattern formed a quadrilateral structure with four core regions—the United States, Japan, the European Union, and China—as its vertices. Hu et al. (2021) employed the Incopat global patent database to analyze photovoltaic technology patent data from China over the past two decades [45]. Using Social Network Analysis (SNA) from both single-mode and dual-mode network perspectives, they examined structural characteristics of the innovation network. The findings revealed that leading photovoltaic firms have established relatively stable internal collaboration networks, with the innovation ecosystem demonstrating rapid expansion in scale. Ren et al. (2024) compared China’s new energy vehicle innovation networks between 2015 and 2020, revealing that the network structure evolved from a Shanghai-Suzhou corridor-like backbone into a Shanghai-Suzhou-Hangzhou-Nanjing triangular framework [46]. The network’s topological form evolved from single-core radiating multi-node star-shaped cooperation patterns to complex network structures with robust cores and extensive capabilities.

2.2.3. Research Commentary

Overall, scholarly research on industrial clusters is progressively shifting from broad investigations to more refined analyses. The focus has moved from general industrial clusters to patent-intensive industrial clusters, which better align with the demands of the knowledge economy. The research perspective has significantly broadened. It now extends beyond studying the relationship between industrial clusters and regional economic development to exploring how these clusters interact with regional innovation mechanisms. Furthermore, it increasingly focuses on the dynamic connections between industrial clusters and innovation networks. It is evident that China’s industrial clusters have evolved from their initial forms—either indigenous or processing-oriented “production clusters”—toward “patent-intensive industrial clusters” where patents form the core of development. This transition has led to an extension of their life cycle and an upward shift in the value chain. As inter-firm collaboration deepens, innovation activities among participants within these clusters have gradually evolved into a networked development model. From the research paradigm of innovation networks, the most common approach is to treat patent holders as network nodes and co-authorship relationships as connections within the network. Studies show that developed countries generally build efficient, interconnected, multi-level, and cross-regional networks based on multi-agent collaboration. In contrast, innovation networks in less developed countries often appear sparse and fragmented. In recent years, China’s innovation network structure has evolved from a single-core radiation pattern to a multi-core leadership framework, initially forming a large-scale and relatively complete innovation infrastructure. As the economic contribution of patent-intensive industries becomes increasingly prominent, academic attention has grown regarding their subfields and cluster phenomena. Scholars widely acknowledge the localized, cross-regional, and networked advantages demonstrated by these clusters. Although previous studies have extensively theorized industrial clusters, research specifically focused on patent-intensive industrial clusters remains in its early stages. Significant gaps exist in exploring their conceptual depth, industry identification methods, and enhancement mechanisms. Innovation networks, as critical platforms for connecting and facilitating interaction and collaboration among cluster elements, continue to serve as a breakthrough for understanding the development patterns and upgrading pathways of such clusters. Therefore, this study leverages enterprise-level patent collaboration data to construct an innovation cooperation network within China’s patent-intensive industrial clusters. Through analyzing the current state and evolutionary process of these clusters, it aims to chart their development trajectory.

3. Data Sources and Research Methods

3.1. Data Sources

Since the U.S. Department of Commerce and Patent and Trademark Office released the Intellectual Property and the U.S. Economy: Industry Focus report in 2012, countries worldwide have intensified research and attention on patent-intensive industries. The U.S. Patent and Trademark Office uses patent intensity (the ratio of granted invention patents to employment) as a key metric, defining industries exceeding average patent intensity as patent-intensive [47]. The EU adopts a similar approach but additionally considers relative patent density [48]. China’s National Bureau of Statistics, in its Statistical Classification of Intellectual Property (Patent)-Intensive Industries (2019), defines these industries as collections meeting predetermined standards for invention patent intensity and scale, with intellectual property as their core competitive element and aligning with innovation-driven development [49]. Based on industrial characteristics and patent intensity, China categorizes patent-intensive industries into seven major groups: ICT manufacturing, ICT services, advanced equipment manufacturing, new material manufacturing, pharmaceutical and medical industries, environmental protection, and R&D/design/technical services. This study analyzes industries identified under this classification.

The National Intellectual Property Administration (CNIPA) has conducted systematic statistical surveys on patent-intensive industries since 2012. This study sets the timeframe from 1 January 2012 to 31 December 2023, subdivided into three periods: 2012–2015 (rise amid global recovery), 2016–2019 (deepening under globalization/informatization), and 2020–2023 (industry transformation post-COVID-19). Patent data were sourced from CNIPA’s official website. Data collection followed these steps:

• Aligned with the 7 major categories and 188 sub-sectors in the Statistical Classification (2019), matched National Economic Industry codes with A-share listed companies in the Wind Database, yielding 2454 matched firms by 31 December 2023.
• Cross-referenced branch information for these firms, identifying 3569 branches.
• Used web crawlers to retrieve invention patent data from CNIPA’s Patent Search System (https://pss-system.cponline.cnipa.gov.cn/conventionalSearch, accessed on 3 January 2024), using firm names as search terms. This yielded 455,827 records, including 357,003 independent applications and 98,824 collaborative applications.

Network construction involved:

• Excluding patents with individual applicants.
• Verifying applicant locations (prefecture-level cities) for collaborative patents via Qichacha, Tianyancha, and manual checks, removing foreign-located applicants.
• Using Python 3.10 and geocoding with JavaScript API 3.0 to obtain precise latitude/longitude (9 decimal places) based on applicant addresses.
• Pairing applicants from the filtered 98,464 collaborative patents to form weighted innovation links, constructing adjacency matrices for subsequent analysis.

3.2. Research Methodology

3.2.1. Kernel Density Analysis

Kernel density analysis is a non-parametric statistical method that uses kernel functions to smooth spatial data points, revealing density distribution characteristics across a region [50]. To comprehensively capture multiple dimensions such as spatial agglomeration, inter-industry linkages, and innovation, this study focuses on two core elements—enterprises and innovation output. Through geocoding technology, corporate registration addresses are converted into geographic coordinates for spatial positioning of enterprise points. To mitigate the impact of single-year data fluctuations, the average number of invention patent applications during the study period is used as the weight function indicator for analyzing point density distribution in surrounding environments. Finally, the natural breaks classification method is applied to grade kernel density values of patent-intensive industries, uncovering their spatial density distribution and agglomeration degree. To minimize errors, this study employs a non-parametric estimation model for analyzing the kernel density of patent-intensive industries. This approach does not impose any assumptions on the data distribution, utilizing the Kernel Density tool in ArcGIS 10.8 for the analysis. The kernel density estimation formula adopted in this study is as follows:

$${\widehat{f}}_h(x)={\displaystyle \frac{1}{n}}{\displaystyle \sum _{i=1}^n{K}_h}(x-x_i)={\displaystyle \frac{1}{nh}}{\displaystyle \sum _{i=1}^{n}K}\left({\displaystyle \frac{x-x_i}{h}}\right)$$

where, n represents the number of patent-intensive enterprises, x denotes a spatial point with unknown density, x_i corresponds to the i-th known density spatial point in discrete distribution, K signifies the kernel function based on patent data, and h indicates the distance decay threshold (bandwidth parameter). The value of f(x) reflects the kernel density at point x, with higher values indicating greater spatial concentration of patent-intensive industries.

3.2.2. Social Network Analysis

This study employs social network analysis to model innovation networks within patent-intensive industrial clusters. Social networks comprise nodes (representing social actors) and their relational ties, with this analytical approach quantifying relationships to reveal structural characteristics and influence patterns among actors [51]. To characterize China’s patent-intensive industrial clusters, we construct an undirected weighted network model through sequential steps. Firstly, nodes are defined as cities hosting firms, represented by the vector Vi = [V₁, V₂, …, V_n] where V_i denotes the i-th city. Secondly, inter-city patent collaborations are established as edges, with weights assigned based on collaboration frequency to form the weight matrix W = [W_{i, j}]. Thirdly, an N × N symmetric adjacency matrix A = [A_ij] (where A_ij = Aji) is generated to reflect connection strength. The resulting network model is formally expressed as the triplet C = [V, W, A]. This framework provides unique insights into collaboration patterns among innovation actors, revealing the structural architecture of patent-intensive industrial clusters. Among these, node centrality is used to measure the importance of nodes within the network and can be quantified through multiple indicators. Degree centrality serves as the most direct metric, representing the proportion of a node’s degree (i.e., the number of edges directly connected to it) relative to the degrees of all nodes in the network. Additionally, this study introduces the following three indicators to measure the overall characteristics of the innovation network in patent-intensive industrial clusters: (1) Network density, which refers to the ratio of the actual number of relationships among nodes to the theoretically maximum possible number of relationships, measuring the closeness of connections between urban nodes; (2) Average path length, measuring the average distance between any two urban nodes in the network; (3) Average clustering coefficient, measuring the extent to which nodes in the network cluster together. The measurement methods for the relevant indicators follow Li and Zheng (2024) and Montes et al. (2024) [52,53].

3.2.3. Community Detection

To gain deeper insights into the internal structure and track the dynamic evolution of patent-intensive industrial clusters, we employ community detection to partition the network across three distinct stages. Notably, while kernel density analysis and social network analysis focus on the overall structural and functional characteristics of cluster networks, community detection specifically targets the core components—groups of firms with strong technological, market, or supply chain interdependencies. This approach aims to elucidate the relationship between local network features and the global architecture.

The essence of this method lies in clustering similar nodes to identify subgraph structures characterized by dense internal connections and sparser external linkages, rendering it highly valuable for cluster research. This study utilizes the Louvain community detection algorithm proposed by Blondel et al. (2008) [54], which hierarchically constructs communities by iteratively merging nodes to maximize modularity. Modularity can be interpreted as the difference between the proportion of edges within the same community and the expected value of that proportion under random assignment of edges. It quantifies the density of intra-community connections relative to a random network, serving as a metric to evaluate the quality of community partitioning. The value ranges from −0.5 to 1, with higher values indicating more distinct community structures. The modularity formula is expressed as follows:

$$Q={\displaystyle \frac{1}{2m}}{\displaystyle \sum _{i,j}\left[A_{i,j}-\frac{k_i{k}_j}{2m}\right]\delta (c_i,c_j)}$$

Among them, A_ij represents the edge weight between nodes i and j; k_i and k_j denote the degrees of nodes i and j, respectively; m is the total weight of all edges in the network; c_i and c_j indicate the communities to which nodes i and j belong; and δ is the Kronecker delta function, which equals 1 if i and j are in the same community, and 0 otherwise. In the collaboration network of patent-intensive industrial clusters, the indicator reflecting the closeness of innovation connections should be the frequency of collaboration among innovation entities. Firstly, this study treats each node i as an individual community and adds it to a neighboring node j to calculate the incremental change in modularity (ΔQ). The formula for calculating ΔQ is as follows:

$$\Delta Q=\left[{\displaystyle \frac{{\Sigma }_{in}+k_{i,in}}{2m}}-{\left({\displaystyle \frac{{\Sigma }_{tot}+k_i}{2}}\right)}^2\right]-\left[{\displaystyle \frac{{\Sigma }_{in}}{2m}}-{\left({\displaystyle \frac{{\Sigma }_{tot}}{2m}}\right)}^2-{\left({\displaystyle \frac{k_i}{2m}}\right)}^2\right]$$

In the formula, ∑_in and ∑_tot represent the sum of edge weights within community c and between community c and other communities, respectively, where edge weights denote collaboration frequency. The algorithm maximizes modularity by iteratively reassigning nodes to neighboring communities that yield the maximum modularity gain ΔQ. Subsequently, communities identified in the resulting network are aggregated into super-nodes. The aggregation rule sums up the edge weights between nodes within the original community to construct a new network structure. This aggregated network then serves as input for the next iteration, repeating the above steps. The algorithm terminates when modularity fails to increase further or when a preset iteration limit is reached. The final iteration output is regarded as the optimal partitioning of the innovation collaboration network.

4. Development Trajectory of Patent-Intensive Industries

Overall, the total number of invention patent applications in patent-intensive industries has exhibited a fluctuating rise-then-fall pattern, demonstrating distinct stage-specific characteristics (as shown in Figure 1).

Specifically, from 2012 to 2015, the number of invention patent applications showed a slow growth trend. During this period, patent-intensive industries witnessed a steady increase in independently filed invention patents. Given the pronounced cumulative effect of technological output, a strong correlation exists between patent application volume and existing patent portfolios [55]. Despite enterprises’ progress in technological innovation, the low-end positioning of these industries remained fundamentally unchanged. The period 2016–2021 marked a golden era of vigorous development for patent-intensive industries, with total invention patent applications surging at a high-growth pace. During this phase, China introduced the Catalog of Patent-Intensive Industries (Trial), providing clear industry definitions and development guidance. Adhering to an innovation-driven development strategy, China positioned technological advancement as a core engine for socioeconomic progress [56]. As a critical vehicle for innovation, patent-intensive industries received unprecedented attention, injecting robust momentum into national technological advancement and industrial upgrading. Both independently and jointly filed patents exhibited rapid growth, aligning closely with overall patent application trends. This phenomenon indicates enterprises have reached consensus on pursuing independent R&D and open collaborative innovation, forming a synergistic forward momentum. Post-2021 saw a downturn in invention patent applications within patent-intensive industries. Two primary factors explain this reversal. Firstly, economic pressures from the COVID-19 pandemic compelled numerous enterprises to curtail R&D investments, slowing industry growth [57]. Secondly, intensified global competition for market share and resources triggered trade protectionist measures and technological blockades, elevating barriers to technology acquisition [58]. These challenges not only hindered R&D activities but also severely disrupted patent filing strategies. Notably, while independently filed patents declined significantly, jointly filed patents maintained relative stability. The divergence shows enterprises increasingly prefer resource-pooling collaboration when facing complex innovation challenges. This trend is mainly driven by resource constraints and risk mitigation considerations [59]. Collaborative R&D enables more efficient cost-sharing and risk distribution, thereby enhancing overall R&D efficiency and market competitiveness.

In summary, patent-intensive industries exhibit a robust development trajectory, experiencing only temporary fluctuations while maintaining positive long-term prospects. The evolving patterns in jointly and independently filed invention patent applications reflect a strategic shift within these industries—from traditional closed R&D models toward open collaborative innovation frameworks. Analysis of development trends indicates that patent-intensive industries align with both the inherent logic and external conditions for cluster-based growth, demonstrating potential to evolve into world-class industrial clusters. Since cluster evolution occurs not in isolation but through interconnected dynamics within broader industrial ecosystems, examining patent-intensive industries helps establish the macro-context for cluster evolution, thereby enabling comprehensive understanding of internal competitive and cooperative dynamics as well as interaction mechanisms between clusters and their external environments.

5. Network Characteristics and Community Structure of Patent-Intensive Industrial Clusters

5.1. Spatial Evolution Pattern of Patent-Intensive Industrial Clusters

Currently, the geographical boundaries of China’s patent-intensive industrial clusters are yet to be clearly defined. This study selects prefecture-level cities as the basic analytical units. We construct a spatial density distribution model to examine the agglomeration characteristics of patent-intensive industries across geographic spaces, thereby initiating a preliminary exploration of cluster territorial scope.

Using kernel density analysis, this study reveals the spatial evolution of patent-intensive industrial clusters from 2012 to 2023. Results are shown in Figure 2. Color gradients in the figure represent kernel density values, indicating the agglomeration intensity of patent-intensive industries across regions.

Over time, China’s patent-intensive industries show increasingly pronounced spatial clustering. Innovation entities are predominantly distributed in eastern regions, demonstrating “multi-core agglomeration” and “patch-like” distribution patterns. Notably, Beijing-Tianjin-Hebei, Yangtze River Delta, and Pearl River Delta regions exhibit the most significant agglomeration, with intensity decreasing from central cities to peripheries, forming a distinct “core-periphery” spatial structure. Cities like Xi’an, Zhengzhou, Chengdu, and Changsha also show moderate clustering.

From 2012 to 2015, innovation entities radiated outward from Shenzhen and Beijing, forming two major clusters in Pearl River Delta and Beijing-Tianjin-Hebei. Simultaneously, the Yangtze River Delta centered on Shanghai showed notable secondary agglomeration. Provincial capitals including Xi’an, Chengdu, and Guiyang displayed contiguous block-like distribution patterns.

From 2016 to 2019, enterprise distribution intensified in these regions. Pearl River Delta and Beijing-Tianjin-Hebei maintained high-density distribution with significant kernel density increases. The Yangtze River Delta, as a secondary cluster, expanded westward, developing more low and medium-density zones. Patent activities decreased around Changsha, slightly reducing low-density zones. Other regions continued provincial capital-centered contiguous clustering.

From 2020 to 2023, high-density zones remained concentrated in the aforementioned regions, with some expanding outward or evolving into higher-density zones. Beijing-Tianjin-Hebei recorded the highest kernel density growth, with small-scale density peaks emerging at regional edges. High-density zones in Yangtze River Delta expanded from Shanghai to Suzhou and Jiaxing, with medium-density zones also extending significantly. Enterprises showed contiguous distribution in high-density zones, reflecting typical “Matthew Effect” characteristics. Cities like Wuhan and Zhengzhou also demonstrated notable spatial evolution.

Benefiting from the implementation of the 2014 Beijing-Tianjin-Hebei Collaborative Development Plan Outline, the Beijing-Tianjin-Hebei region established a vertical division system at an early stage characterized by “R&D in Beijing, transformation in Tianjin and Hebei.” Within this framework, Tianjin’s integrated circuit industry and Hebei’s automotive manufacturing sector have formed strong linkages with Beijing’s patent output. Since the Yangtze River Delta integration was elevated to a national strategy in 2018, a synergistic mechanism has been formed between Shanghai’s construction of a global sci-tech innovation center and the G60 Science and Technology Innovation Corridor. This synergy has facilitated the extension of industrial chains from Shanghai’s Zhangjiang High-Tech Park to secondary innovation nodes, including Suzhou Industrial Park and Jiaxing Science and Technology City. Notably, between 2020 and 2023, a vertical collaboration network emerged involving integrated circuit design in Shanghai, semiconductor packaging and testing in Suzhou, and electronic component manufacturing in Jiaxing. This closed-loop industrial chain of design-manufacturing-packaging has driven the outward diffusion of kernel density, demonstrating the efficiency of optimized resource allocation under market mechanisms. Furthermore, collaborations between Shanghai’s Zhangjiang Laboratory and Suzhou BioBay have led to the establishment of joint laboratories. At the same time, digital technology spillovers from Hangzhou’s Future Sci-Tech City—home to the Alibaba ecosystem—have extended to the Jiaxing G60 corridor. These collaborations have effectively reduced innovation costs and increased patent output across the involved cities. In addition, the diffusion effect of talent mobility has also contributed to the increased agglomeration in the Beijing-Tianjin-Hebei, Yangtze River Delta, and Pearl River Delta regions. Graduates from top universities in Beijing, such as Tsinghua University and Peking University, have been systematically relocating to surrounding cities like Tianjin and Shijiazhuang through the Beijing-Tianjin-Hebei Talent Integration Plan. Unlike the policy-driven agglomeration in Beijing-Tianjin-Hebei and the Yangtze River Delta, the patent-intensive industries in the Pearl River Delta rely more on spontaneous collaboration among market entities. Since the establishment of the Shenzhen Special Economic Zone in 1980, the Pearl River Delta has developed a model combining a flexible policy environment with market-oriented exploration. Leading companies such as Huawei, Tencent, and DJI have engaged thousands of SMEs in upstream and downstream industrial chains through market mechanisms like technology outsourcing and joint R&D. These enterprises spontaneously form collaborative relationships based on order demands and technological complementarity, resulting in a regional division characterized by “R&D in Shenzhen, manufacturing in Dongguan, and component production in Foshan.” Meanwhile, the Guangdong-Hong Kong-Macao Greater Bay Area operates under the “one country, two systems” framework, which effectively facilitates cross-border flows of factors and injects global resources into market collaboration. The core advantage of this market-led model lies in its greater flexibility.

5.2. Innovation Network Characteristics of Patent-Intensive Industrial Clusters

5.2.1. Overall Characteristics of Innovation Networks in Patent-Intensive Industrial Clusters

We employ social network analysis (SNA) to examine the overall characteristics of China’s patent-intensive industry innovation network. The results are presented in

Table 1. Statistical Characteristics of China’s Patent-Intensive Industry Innovation Network.

	Network Density	Average Path Length	Average Clustering Coefficient
2012–2015	0.037	2.504	0.457
2016–2019	0.042	2.475	0.503
2020–2023	0.044	2.396	0.493

. During the study period, the innovation network density of China’s patent-intensive industries showed consistent growth, rising from 0.037 in 2012–2015 to 0.044 in 2020–2023. This increase indicates that the number of actual connections among nodes in the network grew relative to the potential number of connections, reflecting higher collaboration frequency and stronger interaction among innovation entities. The average clustering coefficient reflects the degree of agglomeration in the interconnected networks among node cities. A higher clustering coefficient suggests a greater tendency for node cities to form “small-world” structures. Although this indicator peaked at 0.503 in 2016–2019 and slightly declined to 0.493 in 2020–2023, it demonstrated an overall upward trend. This implies the presence of multiple tightly connected local subgroups within the network, especially in the mid-to-later periods. Meanwhile, the average path length in the network continued to decrease across the three periods. This reduction indicates diminished loss and resistance in knowledge flows between city nodes, along with a lower risk of information distortion. The network as a whole exhibits a high degree of organization and cohesion.

5.2.2. Innovation Network Connectivity Characteristics in Patent-Intensive Industrial Clusters

Figure 3 visualizes an undirected weighted network model constructed using social network analysis. Nodes represent cities where patent applicants are located, while connecting lines signify inter-city patent collaborations, revealing innovation linkages within patent-intensive industrial clusters at local and cross-regional levels.

Overall, the core nodes of China’s patent-intensive industrial cluster network align closely with economically advanced regions, showing a spatially uneven innovation pattern where intensity is higher in the east than the west. Over time, collaboration has shifted from few-core dominance to multi-core interconnected cluster networks. The architecture evolved from a triangular structure supported by Beijing-Tianjin-Hebei, Yangtze River Delta, and Pearl River Delta regions to a diversified multilateral framework. At the Intra-regional linkages, eastern clusters demonstrate multi-core leadership with strong synergy and diversified development, while western clusters exhibit isolated leading regions with uneven progress. Regarding interregional linkages, the innovation network’s scale and density continue to grow, yet core nodes remain concentrated in eastern and central regions. These core nodes actively build tightly connected networks, reinforcing stable core cluster groups.

During 2012–2015, patent-intensive industrial clusters exhibited distinct spatial disparities in innovation collaboration at both local and cross-regional scales. Cities like Shenzhen, Beijing, Nanjing, and Shanghai excelled in local innovation cooperation, forming robust networks that enhanced knowledge flow and sharing. Shenzhen stood out for its exceptional synergy in driving intra-regional collaboration. For interregional innovation linkages, Beijing, Shenzhen, Dongguan, and Shanghai emerged as pivotal hubs, radiating influence across broader regions.

Between 2016 and 2019, cities including Beijing, Nanjing, Guangzhou, Shenzhen, and Shanghai established tightly-knit local innovation networks. The number of high-connectivity regions tripled compared to the previous phase, with Beijing surpassing others to become the top city in total local connections. Cross-regional collaboration intensity surged significantly, though disparities in inter-city cooperation frequency widened. Beijing further solidified its dominance in the innovation network, strengthening ties with Nanjing, Shenzhen, and Chengdu.

From 2020 to 2023, multiple Yangtze River Delta cities advanced to the top tier of local innovation connectivity, achieving substantial progress in intra-regional collaboration. Simultaneously, western regions such as Korla and Yili began developing patent-intensive industries and building localized innovation networks. Beijing and Chengdu formed the strongest partnership in the patent-intensive cluster network, while other core nodes expanded and deepened, resulting in an increasingly complex innovation architecture.

The majority of listed companies in patent-intensive industries are concentrated in eastern regions such as the Pearl River Delta, Beijing-Tianjin-Hebei, and the Yangtze River Delta. This geographic concentration has established cities like Beijing, Nanjing, Shanghai, and Shenzhen as primary hubs for both local and cross-regional collaboration, resulting in a “rich-club phenomenon.” The well-developed market mechanisms and abundant resources in eastern regions have intensified competition among enterprises. Against this backdrop, companies have widely heightened their awareness of patent protection, regarding it as a key strategy to enhance core competitiveness. Several cities have established relatively comprehensive industrial chain systems, featuring not only close inter-firm connections but also robust auxiliary institutions such as technical consulting agencies and specialized service providers. This has further deepened the division of labor and collaboration. Innovation activities within and between these core clusters dominate, facilitating the establishment and development of cluster-network-based innovation cooperation models. In western China, cities like Changji and Ürümqi seized early opportunities presented by the Belt and Road Initiative and the Western Development Strategy. They demonstrated strong execution capabilities and foresight in planning patent-intensive industries and implementing supporting measures. Conversely, cities such as Korla and Yili, benefiting from natural resource endowments, initially focused more on traditional sectors like infrastructure construction and resource development. As national emphasis on innovation-driven development strategies increased, these cities began shifting their focus toward patent-intensive industries and deepening local collaboration.

Cities including Yichang, Jinhua, Ürümqi, Hohhot, Baoding, and Yantai have formed preliminary innovation partnerships based on geographic proximity, though cross-regional innovation activities remain relatively limited. Historically, China’s regional spatial development planning prioritized transportation corridors and central cities’ nodal functions but paid insufficient consideration to intra-regional urban functional networks and interactions with peripheral cities. This led to spatially constrained urban development policies. Extensive research indicates that the functional division of labor within metropolitan clusters and urban agglomerations can create economies of scale unattainable by individual cities [60]. Therefore, cities should position themselves within the multi-tiered spatial structure of “urban agglomerations–metropolitan circles–central cities–peripheral cities.” Integrating cross-regional network resources can foster innovative models such as distributed networked manufacturing, service alliances, and platform economies. This thereby promotes value-added regional value chains and enables vertical specialization and horizontal leapfrogging. Less-developed cities like Lishui, Cangzhou, Yuncheng, Yingkou, and Maanshan predominantly occupy marginal positions in cluster networks, aligning with their developmental levels and comprehensive capabilities. Constrained by historical legacies, market maturity, industrial structure, and resource endowments, these cities face challenges such as population outflow and low economic efficiency. Significant developmental gaps exist compared to advanced regions. Blindly imitating or engaging in homogeneous competition during industrial development will hinder breakthroughs. Therefore, the priority is to base strategies on the actual development level and core competitiveness of local industries, implementing tailored development approaches. This means focusing on cultivating distinctive regional products and leading industries that align with local characteristics, industry specifics, and market shifts. Simultaneously, actively pursue collaboration with developed regions and industry leaders. Through measures such as introducing and assimilating advanced technologies, fostering integrated innovation, expanding and optimizing industrial chains, and building cooperative platforms, drive technological upgrades and enhance product quality within local industries.

5.3. Evolution of Community Structure in Patent-Intensive Industrial Clusters

Existing research suggests that an increase in modularity often corresponds to an improvement in community quality [61]. The results obtained using the Louvain community detection method show that the modularity Q-values for all three periods remained at high levels: 0.53 (2012–2015), 0.49 (2016–2019), and 0.61 (2020–2023). It indicates that patent-intensive industrial clusters exhibited significant community structural characteristics at each development stage. Notably, the Q-value in the third period showed a clear upward trend. This not only reveals a concentration trend of patent collaboration within specific groups but also reflects that the community division during this period better aligns with the actual structure of patent-intensive industrial clusters. The analysis reveals that the community structure within the collaboration network of patent-intensive industrial clusters demonstrates a clear orientation toward technology-driven linkages and strategic synergy, rather than strictly adhering to the principle of geographical proximity. This indicates that connections among innovation entities are primarily driven by technological complementarity, shared R&D risks, and the need for strategic resource integration.

During 2012–2015, the patent-intensive industrial cluster, based on 191 nodes, formed 6 closely connected communities. During this period, four major communities emerged, with their core hub nodes highly concentrated in first-tier cities such as Beijing, Shanghai, Guangzhou, and Shenzhen. These communities collectively covered 90% of the total urban units. The formation of large communities during this phase was primarily driven by the strategic leadership of leading enterprises in core cities. National high-tech enterprises located in parks such as Zhongguancun in Beijing and Zhangjiang in Shanghai leveraged their technological monopolies and resource integration capabilities to proactively build innovation ecosystems centered around themselves. This was achieved by establishing collaborative relationships with local supporting enterprises and external technologically complementary partners. Among them, the innovation community with Beijing as its primary hub exhibited the largest scale, encompassing 85 nodes, highlighting its strong cohesion and extensive radiating effect. In contrast, cities such as Jixi and Quanzhou demonstrated relatively low node patent output, with collaboration networks characterized by strong localization. This outcome reflects insufficient willingness and capacity for external technological cooperation among innovation entities in these cities—most of which are small and medium-sized application-oriented enterprises. Their development relies more heavily on localized knowledge spillovers, indicating a need for improvement in knowledge flow and cross-regional collaboration. As a result, these urban nodes struggle to integrate into broader innovation networks. Some cities, such as Suizhou and Tangshan, despite having relatively weak local collaboration, engaged in frequent cross-border collaborative activities. In these cities, certain leading enterprises or “boundary-spanner” firms adopted an outward-oriented collaboration strategy to overcome technological bottlenecks and access key markets. They proactively established targeted alliances with industry leaders within core innovation communities. Such strategic initiatives have facilitated the integration of their respective urban nodes into these central networks. This phenomenon validates Bathelt’s theoretical perspective that strengthening cross-regional innovation linkages can break the path dependence of local development and maximize the benefits of innovation resource integration [62].

From 2016 to 2019, both the number of nodes and communities in the patent-intensive industrial cluster showed significant growth, forming 8 larger communities around 235 core nodes. Compared to the previous period, the innovation network became denser, reflecting not only the rapid expansion of China’s patent-intensive industries but also signaling a positive momentum towards higher-level and broader development of industrial clusters. During this stage, the community centered around Shanghai as its core hub developed into the largest community system. Its influence extended not only to some cities in the Yangtze River Delta region but also reached distant cities like Changsha and Guiyang. On one hand, leading enterprises in innovation hubs have actively established branches in other regions or formed joint R&D alliances with local firms to attract regional talent and gain proximity to local markets. On the other hand, companies from other regions have also sought to integrate into high-end technology value chains by moving closer to innovation-rich areas. This bidirectional, strategic pursuit of resources has become a core driver behind the cross-regional expansion of innovation communities. Similarly, communities centered around Beijing and Chongqing have extended their collaboration networks to western cities such as Hulunbuir and Hami, further confirming this trend. The strong demand for cutting-edge knowledge and complex technologies in patent-intensive industries dictates that collaboration must be prioritarily driven by technological and strategic considerations rather than geographical proximity. The primary criterion for partner selection is no longer physical distance, but rather the ability to access critical technological complementarities, share high R&D risks, and jointly capture future market opportunities. This decision-making mechanism, based on capability matching and strategic synergy, motivates innovation actors to transcend administrative boundaries and proactively seek partners with specialized technical expertise or scarce innovation resources. As a result, cross-regional collaboration networks centered around technology chains have emerged. For example, a Shenzhen company specializing in AI algorithms may establish a tripartite patent collaboration with a top-tier hospital in Beijing and a precision manufacturer in Shanghai, driven by the potential application of its technology in medical image recognition. The formation of such partnerships stems from the high degree of complementarity in technical capabilities and market channels among the involved parties. Moreover, advancements in search and communication technologies in the digital era have further diminished the limiting effect of geographical distance. The collective collaboration behaviors of enterprises—guided by technological and strategic needs—have shaped the unique structure of innovation networks in patent-intensive industries. These networks break geographical constraints and follow an endogenous technology-driven logic, which fundamentally explains the highly dynamic and cross-regional nature of their community evolution. This study finds that the cluster development characteristics of patent-intensive industries align with Boschma’s (2005) theoretical perspective [63]. Geographical proximity is not a necessary prerequisite or decisive factor for innovation. Rather, it functions as an enhancer of other proximity effects.

The period 2020–2023 witnessed robust development momentum in patent-intensive industrial clusters. The number of nodes covered by these clusters showed significant growth, reaching 273. Concurrently, the number of communities expanded to 10. This expansion reflects growth in the clusters’ geographic scope, number of participating entities, and breadth of collaboration. It also reveals increasing complexity and diversification within their internal cooperative structures. Advances in transportation and information technologies have further enabled innovation to overcome geographical constraints. Consequently, innovation communities have achieved broader and more flexible expansion in their topological structure. Patent pools, built upon industrial clusters and technology alliances, are now widely recognized for their positive role in overcoming patent crises and facilitating cluster restructuring [64]. Firms merging or joining larger communities can leverage patent pool mechanisms. Internally, this enables cross-licensing to build technological barriers. Externally, it allows unified licensing systems to manage scattered patent resources. Within the same community, innovators often possess related and complementary technologies. This creates concentrated technological advantages that can influence industry direction. Optimized collaboration networks and stronger patent pools enhance the “coupling” effect between innovation supply and demand. This fosters a stable mechanism for interactive regional development [65]. Notably, the largest community during this period comprised 132 nodes, the biggest size observed across the three time periods analyzed. This further solidified the dominant position of mega-communities in patent collaboration. In the self-organization of innovation networks, entities typically choose their optimal technological development path based on core competencies and technical strengths. If a highly concentrated community structure exists, new entrants can often integrate more easily. Simultaneously, the number of small communities (under 15 nodes) peaked at seven. Their flexible organizational structures and decision-making processes allow rapid adaptation to market changes and technological innovations, enabling them to build distinctive advantages. The rapid development of frontier technologies like AI and smart manufacturing continuously generates new collaboration fields, technological pathways, and market demands. These emerging areas often possess unique innovation needs and collaboration potential, driving the formation and growth of specialized small communities with specific strengths.

6. Conclusions and Implications

Leveraging innovation networks to implement cross-organizational knowledge integration and resource sharing is essential for deepening collaboration between local and cross-border innovators, accelerating the development, transfer, and application of patented technologies, and ultimately achieving industrial upgrading and regional innovation. In the knowledge economy, patent-intensive industries have garnered significant academic attention due to their strong innovation capabilities and intellectual property advantages. This study analyzed the characteristics and evolution of innovation networks within patent-intensive industrial clusters using joint invention patent data from listed companies (2012–2023). It focused on the structural features of China’s patent-intensive industry innovation network, their changes over the past decade, how these changes reflect local and cross-regional innovation linkages, and the patterns of community division within these clusters. This study yields the following key findings.

Firstly, patent-intensive industries achieved significant progress from 2012 to 2023. Concurrently, corporate strategies shifted from large-scale independent R&D towards actively embracing open collaborative innovation. Amidst global economic integration and deep value chain restructuring, these industries secured high positions due to superior resource integration and allocation efficiency. Strengthened policy support and shifting market demand created unprecedented opportunities. However, intensifying competition and emerging technologies rendered closed-door innovation strategies ineffective for maintaining advantages in markets prioritizing cost and speed. Consequently, firms urgently need to break traditional innovation boundaries and seek external inspiration to generate more competitive outcomes. The rapid growth in joint inventions strongly evidences this trend, showing firms accelerating innovation through cross-border collaboration, knowledge sharing, and technological complementarity to find new growth points. This open model facilitates knowledge exchange, technology integration, reduces R&D risks and costs, and provides broad market prospects and lasting competitive advantages.

Secondly, China’s patent-intensive industries exhibit significant spatial agglomeration with a distinct core-periphery structure. The “east-high, west-low” geographic distribution reveals regional economic disparities and points to the need for policy adjustments and optimized resource allocation. High-density areas act as core nodes, forming a multi-lateral cross-border network. Eastern and Central China, benefiting from early economic opening and industrial upgrading, accumulated significant innovation capabilities and talent. Consequently, most core nodes are located there, possessing the cumulative technological characteristics needed to lead cluster development. The value of core nodes extends beyond their own advancement; through tight network interactions, they effectively drive development in surrounding and peripheral areas. Technology spillovers, knowledge flows, and deep industrial chain collaboration create efficient and dynamic interactions between core and periphery, promoting cluster upgrading.

Thirdly, the community structure of the collaborative network within China’s patent-intensive industrial clusters exhibits distinct characteristics driven by technological relevance and strategic synergy, rather than strictly adhering to the principle of geographical proximity. The innovation network evolution is gradual, continuous, and stable. Leveraging cross-cluster characteristics and technology spillover capabilities, these clusters continuously expand and deepen their geographic reach, creating new growth points. Some large or highly specialized communities merged to form mega-communities after adjustment and optimization. Simultaneously, emerging fields led to a gradual increase in the number of small communities. This evolution reflects market selection mechanisms and the clusters’ inherent self-adjustment and optimization, helping maintain robust operation, promote efficient resource utilization, and drive sustainable development.

Based on the above conclusions, this study proposes the following policy implications:

(1)

Promote Open Collaborative Innovation and Strengthen Industry-University-Research Integration

According to the research findings, the traditional closed R&D model no longer meets the needs of patent-intensive industrial development. Therefore, policymakers should actively promote the implementation of open collaborative innovation within the industry. During the initial cultivation stage, policies should focus on establishing connections among enterprises by setting up cross-border collaboration guidance funds targeted at small and medium-sized enterprises, universities, and research institutes. Support should be provided for enterprises to jointly build collaborative R&D platforms with universities and research institutions in core hub cities, with particular encouragement for directed cooperation aimed at breaking through “bottleneck” technologies. In the expansion and integration phase, policymakers should support strategic cross-regional R&D alliances and encourage leading enterprises to form innovation consortiums based on technological complementarity. Such cross-regional consortiums should be granted expedited project approval and provided with matching funds. Most importantly, open innovation platforms should be established based on industrial chains and industrial clusters. This will facilitate the creation of diverse and adaptive interaction mechanisms among upstream and downstream players in the industrial chain, enterprises, and other innovation entities. By building cross-industry and cross-regional innovation networks, deep integration of technological innovation chains and industrial chains can be achieved. This will gradually establish a dynamic innovation ecosystem, stimulating the innovative vitality of patent-intensive industries.

(2)

Implementing Differentiated Regional Development Strategies for Efficient Resource Allocation

China’s patent-intensive industries exhibit a distinct spatial agglomeration pattern, characterized by higher concentrations in the east and lower densities in the west. This “core-periphery” structure highlights regional development imbalances but also reveals potential opportunities. High-density areas function as core nodes within the cluster network. Their interconnected, cross-regional network further demonstrates the importance of regional connectivity for the industry’s comprehensive development. Such spatial linkages play a vital role in allocating innovation resources, facilitating interactions among innovation entities, and diffusing innovation outcomes. Policymakers must fully recognize the significance of this network structure. They should account for technological disparities across different regions. By implementing precise regional development strategies, policymakers can promote positive interactions between core nodes and peripheral areas. This approach will enable efficient resource allocation and balanced industrial growth. Core hub cities must enhance their radiating and driving functions. This involves supporting the construction of high-level scientific and technological infrastructure with global influence, as well as open pilot platforms. Additionally, these cities should encourage their research institutions and leading enterprises to provide technical services and knowledge spillovers to the entire country. Meanwhile, these cities should be guided to focus on breakthroughs in original innovation and foundational common technologies, avoiding excessive duplication in low-end segments. For peripheral and outer-tier cities, the approach of developing full industrial chains should be abandoned in favor of implementing a precise embeddedness strategy. They should be supported to leverage their local specialized industrial foundations, proactively connect with resources from core cities, and strive to develop singular strengths in niche technological areas. By establishing initiatives such as satellite R&D centers and cross-regional project collaborations, these cities can embed themselves into the national innovation network, gradually shifting from passive acceptance to active collaboration.

(3)

Clarifying Industrial Positioning to Guide Collaborative Cluster Development

Given that the division of communities within patent-intensive industrial clusters is not solely based on geographical proximity, policymakers should regularly conduct market research. This research aims to comprehensively understand and identify the core competitiveness and potential advantages of the region’s patent-intensive industries. This will help precisely position the direction of industrial development. On this basis, policymakers should fully leverage policy guidance and financial support, among other measures, to actively guide and encourage enterprises to engage in in-depth collaboration around shared goals of technological innovation and market expansion. This ensures that enterprises with interconnected technological development directions can achieve cooperation. Support should be provided for communities with strong technological correlations and intensive innovation collaboration to establish patent pools and technical standard alliances, thereby reducing the costs and risks of collaborative innovation. Furthermore, policymakers must closely monitor dynamic changes in market demand and guide the development of industrial clusters to closely align with these demands. They can regularly organize industry exchange events and market demand seminars. These initiatives provide platforms for information sharing and market insights among enterprises. This enables businesses to promptly adjust their product development directions and market strategies, allowing them to adapt to evolving market conditions and trends.