Feature and Tendency of Technology Transfer in Z-Park Patent Cooperation Network: From the Perspective of Global Optimal Path

A patent is an important carrier of enterprise technology innovation, and patent cooperation among enterprises promotes technology diffusion and sharing in the region. Patents are an up-to-date and reliable knowledge source of innovative technologies, and therefore patent analysis has been a vital tool for understanding technological trends and formulating technology strategies (Yoon & Kim, 2012). Based on previous studies, patent cooperation among enterprises can fall into three types: joint application, patent transfer, and patent licensing. The frequency of patent cooperation reflects the level of innovation cooperation between enterprises. This article uses the IETTN model to study the strong correlation in patent cooperation, based on the joint patent application data provided by Z-Park Management Committee. The list of enterprises includes state-owned enterprises, branches, subsidiaries, research institutions, and other innovative entities that registered with the State Administration for Industry and Commerce. The following figure shows the number of innovation entities and innovation cooperation in the IETTN model from 2008 to 2017, as well as the regional innovation performance indicators (including gross industrial output value and technical income) of Z-Park. The left axis corresponds to the number of enterprises inside and outside Z-Park, while the right axis shows the number of cooperative patent applications, gross industrial output, and technical income. It can be shown that more and more entities join the innovation cooperation and the number of innovation cooperation is on the rise while the regional innovative capacity is also continuously strengthened.

Figure 2

Because of the replicability of technology, the technology flow in the regional innovation network has a strong externality. We employed the co-citation modeling method in (Lin et al., 2020) to transform the direct relationship of patent—enterprise to the undirected enterprise-to-enterprise innovation cooperation to study technology diffusion between high-tech enterprises, in which the joint application strength of enterprises i and j (i ≠ j) is as follows: (1) Cij=∑k=1Nakiakj {C_{ij}} = \sum\nolimits_{k = 1}^N {{a_{ki}}{a_{kj}}} Where the node k represents a patent applied jointly by enterprises i and j, so the patent k has two in-edges coming from enterprises i and j and a_kia_kj = 1, otherwise a_kia_kj = 0. Considering the inter-enterprise patent cooperation studied in this paper, the number of patents independently applied by enterprises is excluded, so the diagonal elements are specified as zero in this paper. In this way, we build a weighted undirected network model in which nodes represent enterprises and edges represent joint application relationship, whose weight means the frequency of cooperation between enterprises. The higher the edge weight between the two nodes means that the two enterprises carry out frequent patent cooperation and establish a close communication relationship in terms of technology.

The Z-Park’s IETTN model of the years 2008, 2011, 2014, and 2017 are shown in this section (Figure 3). The following graphs show innovation cooperation involving enterprises in Z-Park of the four years. The size and color of the nodes in the figure are related to the degree value, while the thickness of the edges reflects the frequency of patent cooperation. Also, the number on the node represents different enterprises. It can be seen that the number of nodes and edges in the network increased significantly from 2008 to 2017. This indicates that thanks to the support of innovation policies and increased innovation awareness, enhancement is seen not only in the innovation capability of high-tech enterprises but also in the technical communication and exchanges between them, which produces stronger externalities for regional technological and economic development. Besides, the difference of degree value between nodes in the network is becoming more significant. In the IETTN network of 2017, No. 123 (State Grid Corporation of China) is the most significant node in the network, and No. 70 (China Electric Power Research Institute) followed.

Figure 3

Z-Park falls into sixteen parks geographically and the high-tech enterprises in Z-Park are divided into 11 technospheres according to the Administrative Measures for the Determination of Beijing High-tech Enterprises (specific classification are shown in Appendix I). To reflect the cooperation between parks and technospheres from 2008 to 2017, we established the IETTN-PA model (Figure 4) and the IETTN-TH model (Figure 5). In these figures, nodes represent different parks and technospheres within Z-Park, marked with the names of them. The thickness of the edges represents the frequency of cooperation. It should be noted that the colors of the nodes and edges have no special meaning, but just to distinguish the four years.

Figure 4

Figure 5

It is also obvious from the above two figures that the cooperation across the park and the technosphere have been on the rise from 2008 to 2017, and the network has become denser. With cross-park and cross-technosphere patent cooperation being deepened, innovation diffusion is getting rid of regional and industrial barriers and continuously enhances integration and exchanges. When different regions and industries communicate and cooperate in technology, enterprises will constantly enrich their innovation resources, upgrade their innovation capabilities, and easily produce radical innovation. Although we have the patent data from 2008 to 2017. Due to the layout limitation and the clarity of the figure, we draw the figure above to demonstrate the degree distribution and log-log distribution of IETTN for four years. As the legend shows, nodes with different shapes represent different years. As can be seen from Figure 6, the heterogeneity of the network topology structure is significant in these four years, which means that a few enterprises have a lot of patent collaboration in the network while the majority of enterprises have relatively sparse collaborations.

Figure 6

The IETTN model is an undirected but weighted network that takes innovation entities as nodes, and the number of times they cooperate on the application for patent as the weight of edges. The weighting of innovation associations can provide a more detailed description of the interaction structure of enterprises in the innovation network, and the intensity of innovation cooperation between the enterprises will have an important impact on the nature of the innovation network. Studies have shown that the distribution of weights can affect the structure and dynamic behavior of the network. Changes in weights can not only reduce the shortest distance in the network and increase the aggregation coefficient, endowing the network with the small-world effect, but also improve the synchronization ability of the network and affect the cluster structure of the network (Tian, Di, & Yao, 2011).

Our research, therefore, is focused on the nature of weighted networks. According to the different meanings and ways of weighting, weights can be divided into two categories: dissimilarity weight and similarity weight. Similarity weights are commonly seen in social networks. Due to the difference in the fundamental nature of these two types of weights, definitions of the basic statistics for networks with dissimilarity weight and similarity weight are vastly different. From the perspective of the attribute of edge weight, the IETTN model is a typical network of similarity weights. The greater the edge weight of the network, the higher the frequency of patent application cooperation between enterprises. This also means that a closer innovation partnership has been forged between two entities.

Some corresponding concepts have been developed to characterize the structural properties of weighted networks, such as weights and their distribution characteristics, the correlation of weights, the shortest distance, the betweennesses values, and the definition and statistical properties of the aggregation degree on weighted networks, etc. (Wu & Di, 2004). However, most of the indicators are commonly used to characterize the global and local structural properties of networks with dissimilarity weights, resulting in the partial loss of network information. As a result, it is necessary to introduce new geometric quantities to more accurately depict the nature of similarity weight networks from a global perspective based on the inheritance of the accuracy of original indicators.

In this paper, we utilize the Revised Floyd-Warshall Algorithm (RFWA) and Strongest Relevance Path Length (SRPL) (Xing et al., 2019) as the synthetic measurement of technology diffusion. This algorithm was first proposed by Xing and used to solve this operational problem of tracking network flow. Its iterative and convergence algorithm is as follows: (2) SRPLij(k)=maxi,j,k∈{1,2,…,N}{wij(k−1),wik(k−1)wkj(k−1)wik(k−1)+wkj(k−1)} SRPL_{ij}^{\left( k \right)} = {{\rm max} _{i,j,k \in \left\{ {1,2, \ldots ,N} \right\}}}\left\{ {w_{ij}^{\left( {k - 1} \right)},{{w_{ik}^{\left( {k - 1} \right)}w_{kj}^{\left( {k - 1} \right)}} \over {w_{ik}^{\left( {k - 1} \right)} + w_{kj}^{\left( {k - 1} \right)}}}} \right\} where SRPLij(k) SRPL_{ij}^{\left( k \right)} is the SRPL between nodes i and j, representing the propagation path with maximum efficiency and effectiveness. If it is greater than wij(k−1) w_{ij}^{\left( {k - 1} \right)} , we keep the record as it is, otherwise just equalize it to wij(k−1) w_{ij}^{\left( {k - 1} \right)} . When the SRPLij(k) SRPL_{ij}^{\left( k \right)} happens to be equal to wij(k−1) w_{ij}^{\left( {k - 1} \right)} , it means the optimal path is just the most direct one between nodes i and j. In the IETTN, SRPLij(k) SRPL_{ij}^{\left( k \right)} is the most efficient path of technology diffusion between enterprises i and j, providing a tool to measure the effectiveness of technical cooperation and exchange between enterprises through patent cooperation.

We abstract two sorts of matrixes from any given similarity-weight network based on SRPL, in which SRPL′ is a numerical matrix, and SRPL″ is a string matrix: (3) SRPL′=(SRPL11(N)⋯SRPL1N(N)⋮⋱⋮SRPLN1(N)⋯SRPLNN(N)) SRPL' = \left( {\matrix{ \hfill {SRPL_{11}^{\left( N \right)}} & \cdots \hfill & {SRPL_{1N}^{\left( N \right)}} \cr \hfill \vdots \hfill & \ddots \hfill & \vdots \cr \hfill {SRPL_{N1}^{\left( N \right)}} \hfill & \cdots \hfill & {SRPL_{NN}^{\left( N \right)}} \hfill \cr } } \right) (4) SRPL″=(Str11(N)⋯Str1N(N)⋮⋱⋮StrN1(N)⋯StrNN(N)) SRPL'' = \left( {\matrix{ \hfill {Str_{11}^{\left( N \right)}} \hfill & \cdots \hfill & {Str_{1N}^{\left( N \right)}} \cr \hfill \vdots \hfill & \ddots \hfill & \vdots \cr \hfill {Str_{N1}^{\left( N \right)}} & \cdots \hfill & {Str_{NN}^{\left( N \right)}} \cr } } \right) Where SRPLij(N) SRPL_{ij}^{\left( N \right)} is the value of the SRPL path between nodes i and j after global searching, and Strij(N) Str_{ij}^{\left( N \right)} embodies the concrete details of this path in sequence by a string, which begins from the very first character with the name of node i and ends with the very last character with the name of node j.

In the IETTN model, technology diffusion hides in numerous innovation cooperation and needs to be revealed according to the trade-off between the propagation efficiency and effectiveness of information flow. Therefore, SRPL is used to describe this process, and a brand-new network-based indicator will be designed as an analytical tool.

Newman generalized Freeman’s betweenness centrality to edges and defined the Betweenness Centrality of Edge as the number of shortest paths between pairs of nodes that run along with it (Girvan & Newman, 2002), namely: (5) CE(i,j)=∑h,i,j,k∈{1,2,⋯,N}dhk(i,j)dhk(i≠j h≠k) {C_E}\left( {i,j} \right) = \sum\nolimits_{h,i,j,k \in \left\{ {1,2, \cdots ,N} \right\}} {{{{d_{hk}}\left( {i,j} \right)} \over {{d_{hk}}}}} \left( {i \ne j\,h \ne k} \right) where d_hk(i, j) is the number of the shortest paths connecting nodes h and k through e(i, j), and d_hk is the total number of the shortest paths connecting nodes h and k.

To find the most important inter-park and inter-technosphere relations among numerous couples of sectors in IETTN models, weighted betweenness centrality of edge-based on RFWA is derived to measure the pivotability of patent-based relationships: (6) CEREWA(i,j)=∑h,i,j,k∈{1,2,⋯,N}Strhk(N)(i,j) C_E^{REWA}\left( {i,j} \right) = \sum\nolimits_{h,i,j,k \in \left\{ {1,2, \cdots ,N} \right\}} {Str_{hk}^{\left( N \right)}} \left( {i,j} \right) where Strhk(N) Str_{hk}^{\left( N \right)} is the number of SRPLs within the scope of the whole network, which is connecting any pairs of others cross a given couple of enterprises i and j. Self-loop serving as an SRPL will plus 1 to this index of itself.

The following example is to explain how we pick an inter-node SRPL fro m various existing paths and then calculate the CERFWA C_E^{RFWA} .

First of all, we need to know how many SRPL paths exist between any pair of nodes or on every single edge. In the network, as shown in Figure 7(a), we finally find out 8 SRPL paths and mark them with red dashed lines, which incorporate 7 direct paths and 1 indirect path. Except for the path of A → C, the other edges are passed through by at least 1 SRPL path and at most 3 SRPL paths. Then, we can get the matrix of CERFWA C_E^{RFWA} based on the statistics on the SRPL paths. For instance, in Figure 7(b), CERFWA C_E^{RFWA} (D, C) = 3 means 3 SRPL paths are passing through it. We will find an interesting pattern that the edge weight equals its corresponding SRPL value while its betweenness centrality exists.

Figure 7

Besides, it is worth testify whether there is a correlation between the weight of the edge and the CERFWA C_E^{RFWA} . The truth is neither the edges with large weight nor the ones with small weight will certainly get a proportional CERFWA C_E^{RFWA} , as shown in Figure 7(c). Based on the theory of “The Strength of Weak Ties” introduced by Granovetter (1973), we recognize that a weak tie connecting two heterogeneous communities will have a large betweenness centrality because it plays a crucial role in information bridge. If the background is set to a similarity-weight network, then it is likely that such an edge has a small weight but a large CERFWA C_E^{RFWA} , which we call the “Crucial Weak Tie”. The path of A → C is of this type of edge, which is tied for fourth in the weight of edge and second in the CERFWA C_E^{RFWA} .

It is assumed that the importance of a given park or a technosphere in technology diffusion is determined by the overall performance of enterprises within it, so the research focus can be shifted to the park level or the technosphere level by merging all the patent relationships between enterprises in a certain park or technosphere. To be specific, the merging of the IETTN network is based on the following two rules.

The values in the CERFWA C_E^{RFWA} matrix represent the patent partnership of a node pair and our current goal is to obtain the intensity of cooperation between two parks. In the IETTN model with M parks (s, r = 1,2,…,M), the Cross-Park Pivotability Matrix denoted by P^CP is: (7) PCP=(P11⋯P1M⋮Pst⋮PM1⋯PMM) {P^{CP}} = \left( {\matrix{ {{P_{11}}} & \cdots & {{P_{1M}}} \cr \vdots & {{P_{st}}} & \vdots \cr {{P_{M1}}} & \cdots & {{P_{MM}}} \cr } } \right) (8) PstCP=∑i∈rs∑j∈rtCERFWA(i,j) P_{st}^{CP} = \sum\nolimits_{i \in {r_s}} {\sum\nolimits_{j \in {r_t}} {C_E^{RFWA}\left( {i,j} \right)} } where PstCP P_{st}^{CP} is the pivotability between park s and park t, CEREWA C_E^{REWA} (i, j) is that of patent relation between enterprise i and enterprise j, and r_s and r_t are the enterprise set of park s and park t respectively. To observe the heterogeneity of pivotability at the park-level, park-to-park pivotability matrices of 2008, 2011, 2014, and 2017 have been visualized in the following heat maps (see Figure 8). It should be noted that to show the contrast of the pivotability of parks more obviously, we adopt the logarithmic representation for the results (value of pivotability). The depth of color is proportional to the degree of pivotability while the corresponding value can be found in the legend on the right, and the letters on the left and top axes correspond to Table 1. Specifically, the top 10 pairs of cross-park pivotability of the four years are listed in Table 2.

Figure 8

It can be seen from the above figure that all parks have witnessed significant changes in their pivotability, especially in the following three aspects:

First, the technology-related pivotability in Haidian Park was in a leading position in the four years and showed a significant increase in 2017. The Haidian-Haidian pivotability stood at 558 in 2008, 1084, and 1940 in 2011 and 2014, respectively, and skyrocketed to 5136 in 2017. Its edge-betweenness-based heterogeneity of pivotability gets increasingly stronger. With the strongest innovation capability among all parks, Haidian Park pools a large number of high-tech enterprises and acts as a perfect source and pivot in innovation diffusion, but the increasing heterogeneity indicates that high-tech enterprises in Haidian Park can hardly facilitate technology diffusion of Z-Park. Its plentiful innovation resources and outstanding innovation capability have not been given full play in the diffusion of technology and resources, and are yet to enhance the innovation capability of Z-Park at large.

Second, the cross-park technology-related pivotability reached its peak in 2014. As seen from the above figure, from 2008 to 2014, the cross-park pivotability is crawling up. There were more cross-park color blocks in the heat map of 2014. Supported by the policy of comprehensively deepening reform and innovation-driven development, innovation policies and programs are initiated, including the academic system reform, the government’s financial support plans, and the launching of intellectual property courts. Scientific and technological achievements and breakthroughs continue to emerge in the fields of high-speed rail, information technology, and aerospace. Despite the economic downturn in 2014, the improved innovation environment and the transformation from scientific researches to productive forces have removed geographical barriers, and continuously integrate resources and technologies across regions, thus enabling vigorous development in scientific and technological innovation.

Third, Yizhuang-Daxing Park, Chaoyang Park, and Xicheng Park account for a large proportion of the cross-park pivotability. Xicheng District is a financial hub of Beijing, and financial technology is one of the leading industries in Xicheng Park. Daxing-Yizhuang Park, relying on the Daxing Bio-medicine Industry Park, has impressive technological accumulation in the field of bioengineering. As for Chaoyang Park, its vital and internationalized cultural background is an attraction for a large number of multinational corporate headquarters, whose gathering has also brought numerous resources for the innovative development of Chaoyang Park. It is noticeable that, in technology diffusion, such parks with distinctive regional characteristics and unique innovative resources play a pivotal role and are more likely to generate a strong attraction in innovation cooperation, thereby facilitating technology transfer and enabling groundbreaking innovation.

Analogously, supposing there are N technical fields (u, v = 1,2,…,N) in the IETTN model, the formula of Cross-Technosphere Pivotability Matrix P^CF would be: (9) PCT=(P11⋯P1N⋮Puv⋮PN1⋯PNN) {P^{CT}} = \left( {\matrix{ {{P_{11}}} & \cdots & {{P_{1N}}} \cr \vdots & {{P_{uv}}} & \vdots \cr {{P_{N1}}} & \cdots & {{P_{NN}}} \cr } } \right) (10) PuvCT=∑i∈du∑j∈dvCERFWA(i,j) P_{uv}^{CT} = \sum\nolimits_{i \in {d_u}} {\sum\nolimits_{j \in {d_v}} {C_E^{RFWA}\left( {i,j} \right)} } where PuvCT P_{uv}^{CT} is the pivotability from technosphere u to technosphere v, whose enterprise sets are d_u and d_v, respectively. Accordingly, data of the same years were selected to observe the difference of pivotability at the technosphere-level, as shown in Figure 9. For the same reason shown in Figure 8, the pivotability in Figure 9 is treated logarithmically. The top 10 technosphere pairs by cross-technosphere pivotability in the four years are listed in Table 3.

Figure 9

It can be seen from the above figure that the pivotability of each technosphere has changed significantly in the four years, especially in the following two aspects:

First, the intra-field pivotability in the electronic information field, with the largest proportion, shows a drastic increase in 2011 and 2014, rising from 1,098 in 2011 to 4,598 in 2014. Unquestionably, the electronic information field undergirds innovation development, and with the unstoppable upgrading of computer equipment and Internet applications, the electronic and information field has gained explosive growth in recent years. However, it is worth noting that the fundamental role of the electronic and information field is to facilitate the development of other fields, so limiting the development within its field would be a waste of innovation resources. How to maximize cross-field cooperation is currently a major challenge facing Z-Park.

Second, the science and technology service industry plays an increasingly important role in cross-field cooperation, especially the cooperation with new materials, advanced manufacturing, and environmental protection. In 2014, a policy on accelerating the development of the science and technology service industry was introduced, including science and technology services and supporting technologies into the government-supported high-tech fields and giving preferential tax rates to inject new vitality. It can be seen from the statistics that in recent years, this field has indeed promoted technology diffusion and innovation development of the park. In addition to the necessary innovation provided to support services, the industry has also helped intensify invisible technical exchanges.

In addition to intra-park collaboration, every park is inextricably linked with others. The pivotability of a given park in the patent cooperation network can be expressed as the overall performance of its patent cooperation with all other parks, based on which we introduce the definition of Park-Level Pivotablility denoted by P^R: (11) PR(s)=∑Psrcp {P^R}\left( s \right) = \sum {P_{sr}^{cp}} where P^R(s) is named the park-level pivotability of park s, P_sr is the element in the s-th row of matrix P^CP.

Each park has its advantages and disadvantages. The isolation from each other is inconducive to the technology upgrade; it is the openness that truly helps establish a technical cooperation platform and thus intensify cooperation and innovation with advantages of all parties being given full play. In Figure 10, we show the value of the pivotability of 16 parks, in which different colors correspond to different years. The x-coordinate is the park name, and the y-coordinate is the value of pivotability. Therefore, high pivotability means that the park is more active in cooperation, and data of the years 2008, 2011, 2014, and 2017 was selected to observe the pivotal role of each park in patent cooperation.

Figure 10

According to Figure 10, Haidian Park, Chaoyang Park, Yizhuang-Daxing Park, and Xicheng Park boast high Park-Level pivotability. From 2011 to 2014, the pivotability of major parks increased significantly. In particular, Chaoyang Park achieved a leap from 377 to 3,089, surpassing Haidian Park, which has the strongest innovation capability. In 2014, the intra-park pivotability of Chaoyang Park accounted for only 4.5%. As one of the areas with a considerable number of multinational companies and corporate headquarters, Chaoyang Park has always been in a leading position in cooperation across regions and even countries and has a distinctive innovation specialty of international R&D, attracting more than 150 multinational companies from more than ten countries and regions to invest and settle in. The introduction of international innovation resources, despite the barriers in the development of core technologies, has hugely attracted and expanded regional and cross-regional cooperation. The pivotability of Xicheng Park also increased from 272 in 2011 to 1,927 in 2014. Xicheng Park is committed to establishing a demonstration zone for innovation in financial technology and specialized services, which promote and facilitate the innovative development of the entire park. The promotion of specialized service parks is also prominently beneficial to technology diffusion.

As always, Haidian Park tops all the other parks in terms of pivotability. Especially in 2017, significantly enhanced on the original basis, the pivotability of Haidian Park reached 5,626, in which, however, the intra-park pivotability is as high as 91%. With a large number of scientific research institutes and high-tech enterprises, the basic innovation strength of Haidian Park is undoubted, especially in the field of electronics and information. Its early start and profound basic innovation strength led to high industry and geographic agglomeration, and on the other hand, limited its communication and cooperation with other parks. Considering its unquestionably high innovation output, it does not play a matching role in technical cooperation and innovation diffusion.

Accordingly, we defined the technosphere-level pivotability by summarizing the patent partnerships between a technology field and all others, i.e. summing up the elements in each row of the matrix P^CT respectively to thus get the vector P^D: (12) PD(u)=∑PuvCT {P^D}\left( u \right) = \sum {P_{uv}^{CT}} where P^D(u) is named the technosphere-level pivotability of technosphere u, and P_uv represents the elements in the u-th row of matrix P^CT.

The following figure shows the degree of pivotability in the different technosphere. Similarly, different colors represent different years. It should be noted that because the degree of technosphere in the field of electronics and information is much higher than in other fields, we add a double line to the vertical axis to compress to show all the values clearly. Cooperation across technospheres is an effective way to expand cooperation, which is often accompanied by the emergence of new products, new technologies, and new design concepts. Then the variation of technosphere-level pivotability at different times is analyzed to explore the dynamic influence of different technology fields in the patent cooperation network.

According to Figure 11, the electronic and information field boasts high technosphere-level pivotability, exceeding 6,040 in 2014. It continuously contributes to innovative development with its considerable high-tech enterprises and the most extensive regional distribution. Between 2014 and 2015, the concept of “Internet +” gave rise to the emerging development pattern of deep integration of the Internet and traditional industries, which led to the technology diffusion in the electronic and information field and the upgrade of traditional industries. At the same time, the “Made in China 2025” Strategy emphasizes the integrated development of manufacturing industries and the Internet, making advanced manufacturing technology a hub area for close collaboration and exchanges with the electronic information field. Seen from a dialectical point of view, the innovation in this field still focuses more on the innovation of the business model, rather than technological innovation concerning people’s livelihood and national hard power. Therefore, to enhance the overall innovation strength of a region, enterprises in the field need to break individual and industrial restrictions, establish extensive and in-depth cooperation with enterprises in other fields, and play a pivotal role, with more focus on technologies regarding 5G and aerospace.

Figure 11

As can be seen from the above figure, although the pivotability of major technosphere in 2017 has slightly decreased compared with 2014, and an increase has been seen in the pivotability of new energy and high-efficiency energy-saving technology and environmental protection technology, among which that of the new energy technosphere rose from 342 to 626. At the end of 2016, the 13th “Five-Year Plan” for energy development emphasized technological innovation in the energy field and clean and low-carbon development, which provided new impetus for cooperative innovation and technological updates. Since both national strategies and public awareness have higher requirements for environmental protection and low-carbon life, the environmental protection industry is expected to have more and more in-depth cooperation with other fields.