Industrial layout prediction method and device, equipment and storage medium

Abstract

The invention discloses an industrial layout prediction method and device, equipment and a storage medium. The method comprises the steps that a heterogeneous graph of each year is constructed according to industrial data and core technology data of a target industry in different years, nodes in the heterogeneous graph comprise region nodes, industrial nodes and core technology nodes, and edges in the heterogeneous graph are used for representing the incidence relation between every two connected nodes; inputting the heterogeneous graph data of continuous N years into a pre-trained industry layout prediction network, and predicting a recommendation value of an edge between the region and the industry in the future year; wherein in the training sample of the network, the true value of the edge between the region and the industry is obtained according to the potential score of the core technology to which the industry belongs in the year and the investment weight of the region to the industry. The dynamic change of the industry and the technology along with time evolution can be captured by predicting according to the heterogeneous graph for continuous years, the mutual influence between different areas is considered, and the accuracy and the efficiency of industrial layout prediction are improved.

Description

Technical Field

[0001] The present disclosure relates to the technical field of data processing, and in particular, to an industrial layout prediction method, apparatus, device, and storage medium. Background Art

[0002] Future industries are the core carriers for the development of new quality productivity, the long-term focus of global technology and industrial competition, and the high ground for enterprises to win sustainable competitive advantages. However, their development faces many new challenges such as unclear trends, unclear boundaries, unclear mechanisms, and unclear paths, and there is a risk of "rushing in and then scattering".

[0003] Traditional industrial layout decisions are mainly based on static factors such as location advantages, resource endowments, industrial bases, and scientific research conditions. However, these factors do not fully consider the dynamic changes of industries and technologies over time and the mutual influence between different regions. With the rapid development of technology and the intensification of the globalization trend, there is an urgent need for a comprehensive analysis and prediction method to guide more efficient and sustainable industrial layout strategies. Summary of the Invention

[0004] The present disclosure provides an industrial layout prediction method, apparatus, device, and storage medium.

[0005] According to a first aspect of the present disclosure, there is provided an industrial layout prediction method, including:

[0006] Construct an heterogeneous graph for each year based on the industrial data and core technology data of the target industry in different years. The nodes in the heterogeneous graph include region nodes, industry nodes, and core technology nodes, and the edges in the heterogeneous graph are used to represent the association relationship between two connected nodes;

[0007] Input the heterogeneous graph data for consecutive N years into a pre-trained industrial layout prediction network to predict the recommended value of the edge between regions and industries in future years;

[0008] Wherein, in the training samples of the network, the true value of the edge between regions and industries is obtained based on the potential score of the core technology to which the industry belongs in the corresponding year and the investment weight of the region in the industry.

[0009] In some embodiments, the method further includes:

[0010] Adopt a sliding time window method to obtain a set of time series graphs based on the heterogeneous graphs for consecutive M years within each time window, so that the heterogeneous graph data for consecutive N years includes (N - M + 1) sets of time series graphs, where M < N, and the industrial layout prediction network is used to predict the recommended value of the edge between regions and industries in the next (M - 1) years.

[0011] In some embodiments, obtaining a set of time series graphs according to heterogeneous graphs of M consecutive years in each time window includes:

[0012] For each year's heterogeneous graph, nodes of a set type are made to obtain messages from neighboring nodes, and after aggregating the received messages, the node representation is updated using the aggregation results;

[0013] Obtaining an information sequence according to information of each node in the heterogeneous graph;

[0014] The information sequences corresponding to each year in M ​​consecutive years are spliced ​​together to obtain a set of time series diagrams.

[0015] In some embodiments, each node and each edge in the heterogeneous graph has attribute information of a set number of dimensions, and each edge also has a unified embedding representation formed by mapping attributes of edges of the same type to a low-dimensional space.

[0016] In some embodiments, the industrial layout prediction network is a long short-term memory network.

[0017] In some embodiments, constructing a heterogeneous graph for each year based on the industry data and core technology data of the target industry in different years includes:

[0018] According to the industry data of each year, establish the industry nodes in the heterogeneous graph of that year;

[0019] According to the core technology data of the year, establish the core technology nodes in the heterogeneous graph of the year;

[0020] By pairing each region with each industry and each core technology, edges between region nodes and industry nodes, and edges between region nodes and core technology nodes are established;

[0021] By identifying the core technologies related to each industry, edges between industry nodes and core technology nodes are established.

[0022] In some embodiments, the step of establishing an edge between an industry node and a core technology node by identifying the core technology associated with each industry includes:

[0023] Based on the patent application data of the target industry within one year, establish patent nodes and edges between patent nodes and regional nodes;

[0024] According to the international patent classification number and abstract content of the patent application, establish the edge between the patent node and the core technology node;

[0025] According to the international patent classification number of the patent application, the core technology is mapped to the industrial node, and the edges between the core technology node and the industrial node are established.

[0026] In some embodiments, when the training samples further include core technologies labeled with whether they are potential technologies, the industrial layout prediction network is further configured to predict whether a core technology is a potential technology.

[0027] In some embodiments, the method further includes: obtaining a list of regions recommended for the development of each industry and a list of industries recommended for the development of each region according to the recommended values of the edges between the predicted regions and industries.

[0028] According to a second aspect of the present disclosure, there is provided an industrial layout prediction device, including:

[0029] A construction unit configured to construct a heterogeneous graph for each year according to the industrial data and core technology data of the target industry in different years, where the nodes in the heterogeneous graph include region nodes, industry nodes, and core technology nodes, and the edges in the heterogeneous graph are used to represent the association relationship between two connected nodes;

[0030] A prediction unit configured to input the heterogeneous graph data for continuous N years into a pre-trained industrial layout prediction network to predict the recommended values of the edges between regions and industries in future years;

[0031] Wherein, in the training samples of the network, the true values of the edges between regions and industries are obtained according to the potential scores of the core technologies to which the industries belong in the corresponding years and the investment weights of the regions in the industries.

[0032] In some embodiments, the device further includes a time-series graph conversion unit configured to:

[0033] Adopt a sliding time window method to obtain a set of time-series graphs according to the heterogeneous graphs for continuous M years within each time window, so that the heterogeneous graph data for continuous N years includes (N - M + 1) sets of time-series graphs, where M < N, and the industrial layout prediction network is used to predict the recommended values of the edges between regions and industries in the next (M - 1) years.

[0034] In some embodiments, when the time-series graph conversion unit is configured to obtain a set of time-series graphs according to the heterogeneous graphs for continuous M years within each time window, it is specifically configured to:

[0035] For each year's heterogeneous graph, enable nodes of a set type to obtain messages from neighbor nodes, aggregate the received messages, and then update the node representations using the aggregation results;

[0036] Obtain an information sequence according to the information of each node in the heterogeneous graph;

[0037] Concatenate the information sequences corresponding to each year in the continuous M years to obtain a set of time-series graphs.

[0038] In some embodiments, each node and each edge in the heterogeneous graph has attribute information of a set number of dimensions, and each edge also has a unified embedding representation formed by mapping attributes of edges of the same type to a low-dimensional space.

[0039] In some embodiments, the industrial layout prediction network is a long short-term memory network.

[0040] In some embodiments, when the construction unit is used to construct a heterogeneous graph for each year according to the industry data and core technology data of the target industry in different years, it is specifically used to:

[0041] According to the industry data of each year, establish the industry nodes in the heterogeneous graph of that year;

[0042] According to the core technology data of the year, establish the core technology nodes in the heterogeneous graph of the year;

[0043] By pairing each region with each industry and each core technology, edges between region nodes and industry nodes, and edges between region nodes and core technology nodes are established;

[0044] By identifying the core technologies related to each industry, edges between industry nodes and core technology nodes are established.

[0045] In some embodiments, when the construction unit is used to establish an edge between an industry node and a core technology node by identifying the core technology related to each industry, it is specifically used to:

[0046] Based on the patent application data of the target industry within one year, establish patent nodes and edges between patent nodes and regional nodes;

[0047] According to the international patent classification number and abstract content of the patent application, establish the edge between the patent node and the core technology node;

[0048] According to the international patent classification number of the patent application, the core technology is mapped to the industrial node, and the edges between the core technology node and the industrial node are established.

[0049] In some embodiments, when the training samples also include core technologies marked as potential technologies, the industrial layout prediction network is also used to predict whether the core technology is a potential technology.

[0050] In some embodiments, the device further includes a recommendation unit for obtaining a list of regions recommended for development of each industry and a list of industries recommended for development of each region based on the predicted recommendation values ​​of the edges between regions and industries.

[0051] According to a third aspect of the present disclosure, an electronic device is proposed, the device comprising a memory and a processor, the memory being used to store computer instructions executable on the processor, the processor being used to implement any of the methods described above when executing the computer instructions.

[0052] According to a fourth aspect of the present disclosure, a computer-readable storage medium is provided, on which a computer program is stored, and when the program is executed by a processor, the method described in any one of the above items is implemented.

[0053] In the disclosed embodiment, a heterogeneous graph including regional nodes, industrial nodes, core technology nodes and edges between nodes is constructed, and the recommended values ​​of the edges between regions and industries in future years are predicted based on heterogeneous graph data for N consecutive years, wherein the true values ​​of the edges between regions and industries in the training samples of the network are obtained based on the potential scores of the core technologies of the industries in the corresponding years and the investment weights of the regions in the industries. By predicting based on heterogeneous graphs for consecutive years, the dynamic changes of industries and technologies over time can be captured, and the mutual influence between different regions can be considered, thereby improving the accuracy and efficiency of industrial layout prediction.

[0054] The industrial layout prediction method proposed in the embodiments of the present disclosure can integrate technologies such as large language models, deep neural networks, attention mechanisms, spatiotemporal graph algorithms, and adaptive recommendations to develop regional recommendation algorithms for the layout of future industries in accordance with local conditions, build a high-quality database that integrates machine intelligence and subjective insights, and a new paradigm for decision support. On the basis of fully considering factors such as resource endowment, industrial foundation, scientific research conditions, talent structure, and location advantages, the evolution and development laws of industry and technology are introduced from the time dimension, and the layout synergy effects between different regions are introduced from the spatial dimension, providing governments at all levels with optional tools that can realize dynamic tracking, real-time updating, seamless monitoring, and global optimization of future technologies and industries, thereby improving decision-making accuracy and efficiency, and reducing decision-making costs and risks.

[0055] Other features and advantages of the present disclosure are described in detail below. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the specification and, together with the description, serve to explain the principles of the specification.

[0057] Figure 1 A flow chart of an industrial layout prediction method according to an embodiment of the present disclosure is shown.

[0058] Figure 2 An exemplary industrial chain diagram of the ICT industry is shown.

[0059] Figure 3A schematic diagram of a timing diagram conversion method according to an embodiment of the present disclosure is shown.

[0060] Figure 4 A schematic diagram of a heterogeneous graph data processing process and a prediction process according to an embodiment of the present disclosure is shown.

[0061] Figure 5 A block diagram of an industrial layout prediction device according to an embodiment of the present disclosure is shown.

[0062] Figure 6 It is a schematic diagram of the structure of an electronic device shown in at least one embodiment of the present disclosure. DETAILED DESCRIPTION

[0063] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0064] The disclosed embodiments may be applied to a computer system / server that may operate with numerous other general or special computing system environments or configurations. Examples of well-known computing systems, environments, and / or configurations suitable for use with a computer system / server include, but are not limited to, personal computer systems, server computer systems, thin clients, thick clients, handheld or laptop devices, microprocessor-based systems, set-top boxes, programmable consumer electronics, networked personal computers, minicomputer systems, mainframe computer systems, and distributed cloud computing technology environments including any of the above systems, and the like.

[0065] Computer systems / servers may be described in the general context of computer system executable instructions (such as program modules) executed by a computer system. In general, program modules may include routines, programs, object programs, components, logic, data structures, etc., which perform specific tasks or implement specific abstract data types. Computer systems / servers may be implemented in a distributed cloud computing environment where tasks are performed by remote processing devices linked through a communication network. In a distributed cloud computing environment, program modules may be located on local or remote computing system storage media including storage devices.

[0066] Figure 1 FIG. 1 shows an industrial layout prediction method according to an embodiment of the present disclosure. Figure 1 As shown, the method includes steps 101 and 102.

[0067] Step 101, construct a heterogeneous graph for each year based on the industry data and core technology data of the target industry in different years.

[0068] In the embodiment of the present disclosure, the nodes in the heterogeneous graph include regional nodes, industry nodes and core technology nodes, and the edges in the heterogeneous graph are used to represent the association relationship between two connected nodes.

[0069] Among them, the regional nodes represent specific geographical areas, such as provinces, cities, districts, county-level cities, etc.

[0070] In some embodiments, the regional nodes in the heterogeneous graph can be selected from the top 100 counties in the country, for example, the top 25.

[0071] Each regional node in the heterogeneous graph can include a series of attribute information (also called basic information), such as annual GDP, population changes, natural resources, policy environment, infrastructure conditions, etc. Based on this attribute information, the resource endowment of the region can be evaluated, and the location advantages of the region can also be determined to provide a basis for industrial layout prediction and regional development planning.

[0072] Industrial nodes are the basic units in the target industry. The attribute information of industrial nodes can include industrial scale, growth rate, technology maturity, market saturation, industrial chain position and added value, etc. Industrial nodes are the core of industrial layout analysis, and their development status and trends are the key content of industrial layout prediction.

[0073] Taking the ICT (Information and Communication Technology) industry as an example, the industry chain relationship constructed based on expert opinions is as follows: Figure 2 As shown in the figure, based on this figure, the industry nodes included in the heterogeneous graph constructed for the ICT industry can be determined.

[0074] Core technology nodes represent key technologies or innovations that drive industrial development, such as artificial intelligence, new energy, and biotechnology. The attributes of core technology nodes may include technology maturity, innovation capability, application scope, R&D investment, number of talents, number of scientific research platforms, scientific research infrastructure, number of patents, and changes in patent citations, etc. Core technology is the driving force for industrial development, and changes in core technology will also indicate adjustments to the industrial layout.

[0075] Edges are used to connect different nodes in a heterogeneous graph, indicating the association relationship between nodes. The heterogeneous graph in the embodiment of the present disclosure includes edges between regions and industries (abbreviated as region-industry edges), edges between industries and core technologies (abbreviated as industry-core technology edges), and edges between regions and core technologies (abbreviated as region-core technology edges).

[0076] In some embodiments, based on the industry data of the target industry in a certain year, an industry node in the heterogeneous graph of that year can be established. The edge between the region node and the industry node can be established by pairing each region with each industry indicated in the industry data, that is, the region-industry edge can be formed by constructing a Cartesian product relationship between the region and the industry. The edge can represent the association between the region and the industry. In the embodiment of the present disclosure, the edge can be set to represent the recommendation probability of the region for the industry, and the weight value of the edge represents the recommendation probability value.

[0077] In some embodiments, based on the core technology data of the target industry in a certain year, a core technology node in the heterogeneous graph of that year can be established. By pairing each region with each core technology indicated in the core technology data, an edge between the region node and the core technology node can be established, that is, a Cartesian product relationship between the region and the core technology can be constructed to form a region-core technology edge. The edge can represent the region's R&D investment, policy support, talent aggregation, etc. for the core technology.

[0078] In some embodiments, by identifying the core technology related to each industry, an edge between the industry node and the core technology node can be established. The edge can represent the dependency relationship between the industry and the core technology, such as the improvement of industry efficiency and the reduction of cost by the technology.

[0079] Step 102, inputting N consecutive years of heterogeneous graph data into a pre-trained industrial layout prediction network to predict the recommended values ​​of the edges between regions and industries in future years.

[0080] In the disclosed embodiment, the industrial layout prediction network may be a deep neural network, which is pre-trained using training samples, in which the true value of the edge between a region and an industry (the label of the recommended probability) is obtained based on the potential score of the core technology of the industry in the corresponding year and the investment weight of the region in the industry.

[0081] In one example, the potential score of a core technology can be obtained by calculating the annual patent increase rate of the core technology. For example, all core technologies of the industry can be ranked according to the corresponding annual patent growth rate, and the potential score of the core technology can be determined according to the specific ranking.

[0082] In one example, the investment weight of a region in an industry can be determined based on the investment amount of the region in the industry, and the investment amount is proportional to the investment weight.

[0083] Input the heterogeneous graph data for consecutive N years into the industrial layout prediction network pre-trained according to the above training samples, and the recommended values of the edges between regions and industries in future years can be predicted. The future year can be one year after the consecutive N years or multiple years after that.

[0084] In the embodiments of the present disclosure, a heterogeneous graph including region nodes, industry nodes, core technology nodes, and edges between the nodes is constructed, and the recommended values of the edges between regions and industries in future years are predicted based on the heterogeneous graph data for consecutive N years. Among them, in the training samples of the network, the true values of the edges between regions and industries are obtained according to the potential scores of the core technologies to which the industries belong in the corresponding years and the investment weights of the regions for the industries. The embodiments of the present disclosure perform predictions based on the heterogeneous graphs for consecutive years, can capture the dynamic changes of industries and technologies over time, and consider the mutual influences between different regions, improving the accuracy and efficiency of industrial layout prediction, providing strong decision-making support for the formulation of industrial policies, resource allocation, technological innovation, and regional development, and having significant technical and practical value.

[0085] In some embodiments, the heterogeneous graph data for consecutive N years can be converted into a time series graph. Specifically, the sliding time window method can be adopted to obtain a set of time series graphs according to the heterogeneous graphs for consecutive M years within each time window, so that the heterogeneous graph data for consecutive N years includes (N - M + 1) sets of time series graphs, where M < N, and the future year is (M - 1) years after the last year of the consecutive N years.

[0086] Taking M = 2 as an example, see Figure 3 the schematic diagram of time series graph conversion shown.

[0087] The heterogeneous graphs for consecutive N years are, for example, the heterogeneous graphs for 2012, 2013, 2014, and 2015 respectively. In this example, each time of sliding, the time window contains the heterogeneous graphs for consecutive 2 years. For example, the first time of sliding, it contains the heterogeneous graphs for 2012 and 2013, and by splicing the corresponding data sequences of these two heterogeneous graphs, a set of time series graphs is obtained; the second time of sliding, it contains the heterogeneous graphs for 2013 and 2014, and similarly, a set of time series graphs is obtained by splicing the corresponding data sequences of these two heterogeneous graphs; and so on until the sliding of the time window is completed, and 3 sets of time series graphs are obtained in this example.

[0088] When training the prediction network, since the heterogeneous graphs for consecutive N years are processed in the above manner, the network can learn the process of sequence changes in each set of time series graphs, so the weight values (recommended values) of the region-industry edges for the future (M - 1) years can be predicted. Among them, the future (M - 1) years are (M - 1) years after the last year of the consecutive N years.

[0089] Still taking M=2 as an example, in this case the recommended value for the next year can be predicted, that is, the recommended value for 2016 can be predicted.

[0090] In the embodiment of the present disclosure, a sliding time window can be used to capture the dynamic associations between nodes and the laws of industry and technology evolution across time, and the heterogeneous graph can be converted into a spatiotemporal graph to improve the accuracy of the prediction.

[0091] In some embodiments, the heterogeneous graphs of M consecutive years may be converted into a set of time series graphs in the following manner:

[0092] First, for each year's heterogeneous graph, nodes of a set type obtain messages from neighboring nodes, aggregate the received messages, and use the aggregation results to update the node representation.

[0093] The node of the set type may be a node related to subsequent predictions, such as an industry node.

[0094] See also Figure 4 As shown, taking the heterogeneous graph of the nth year as an example, the target node is the point to be aggregated and updated, which is called the target node here. It receives messages from neighboring nodes, aggregates the received messages, and updates the node representation using the aggregation results. Through the above message passing mechanism, dynamic associations across space can be captured. Next, according to the information of each node in the heterogeneous graph, an information sequence is obtained.

[0095] By splicing the characteristic information of all nodes, the information sequence corresponding to the heterogeneous graph of that year can be obtained. In some cases, all nodes can be sorted by ID first, and then the characteristic information of the sorted nodes can be spliced ​​in sequence to form an information sequence; nodes can also be grouped by type first, the node information in the group can be spliced, and then the information sequence corresponding to each group can be spliced ​​to obtain the information sequence corresponding to the heterogeneous graph of that year; characteristic information of each node can also be spliced ​​according to the topological structure of the heterogeneous graph.

[0096] Finally, the information sequences corresponding to each of the M consecutive years are spliced ​​together to obtain a set of time series diagrams.

[0097] In some embodiments, each node and each edge in the heterogeneous graph has attribute information of a first set number of dimensions, and each edge also has an embedded representation of a second set number of dimensions formed by mapping the attributes of the same type of edges to a low-dimensional space. Specifically, for all edges of the same type in the graph, they are assigned a shared embedding vector, which represents the common characteristics or functions of such edges.

[0098] If the number of attribute information set for a node or edge does not reach the set number, the insufficient part will be filled with 0.

[0099] Assume that 10 dimensions of attribute information (basic information) are divided for each node and each edge. For example, the attribute information of a regional node includes annual GDP, population change, natural resources, policy environment, and infrastructure status, and the missing 5 dimensions are filled with 0.

[0100] For the edges in the heterogeneous graph, in addition to the basic information of the 10 dimensions for which they are divided, they also have an embedded vector of, for example, 118 dimensions. In this case, the total feature length of the edge is the sum of the attribute 10 dimensions and the embedding 118 dimensions, that is, 128 dimensions.

[0101] By assigning a fixed dimension to each node and creating a structured feature representation for the node, it is more helpful for the model to process and compare node information; and by mapping the same edges into a unified embedding representation, the expressiveness and generalization capabilities of the model can be improved.

[0102] In some embodiments, a long short-term memory network (LSTM) can be used to predict the region-industry recommendation value for the next (M-1) years based on the heterogeneous graph data of N consecutive years processed as described above.

[0103] See also Figure 4 The heterogeneous graph data processing and prediction process shown in the figure. Assume that the heterogeneous graph data of the nth year and the n+1th year in the heterogeneous graphs for N consecutive years are spliced ​​into a set of time series graphs after node aggregation and update, and LSTM is used to predict the recommendation probability value of the region-industry edge based on the time series graph.

[0104] In some embodiments, the edge between the industry node and the core technology node can be established in the following ways:

[0105] First, obtain the patent application data of the target industry in a certain year, including the number of patent applications, patent applicants, and the regions to which the patent applicants belong. Based on the patent application data, a patent node can be created for each patent, and the patent node can be connected to the corresponding regional node according to the region to which the patent applicant belongs, forming an edge between the patent node and the regional node.

[0106] Next, according to the International Patent Classification (IPC) number and abstract content of the patent application, the core technology field to which the patent belongs is identified, and the edge between the patent node and the core technology node is formed by connecting the patent node with the corresponding core technology node.

[0107] Finally, according to the international patent classification number of the patent application, the core technology is mapped to the unique industry node, and the edge between the core technology node and the industry node is established.

[0108] Specifically, by analyzing the IPC classification number and patent description of the patent, it is possible to determine which core technology fields are closely related to which industries, thereby achieving a mapping between core technology nodes and industry nodes.

[0109] In the embodiments of the present disclosure, the layout of core technologies is reflected by patent data, so that the mapping relationship between core technologies and industries can be accurately achieved.

[0110] In some embodiments, the network may input two prediction tasks simultaneously, one is predicting the region-industry recommendation probability as described above; the other is predicting whether the core technology is a potential technology.

[0111] In order to enable the model to output the above two prediction tasks, the training samples of the network also include core technologies with labels, which are used to indicate whether the core technology is a potential technology.

[0112] Among them, whether a core technology is a potential technology can be judged in the following way: first, the annual patent growth rate of the core technology is calculated, and the core technology with the highest growth rate of a set percentage is selected as a potential node, that is, the label of the node is set to 1; conversely, if a node does not meet the above requirements, the label of the node is set to 0. In an example, the set percentage can be, for example, 10%.

[0113] When the recommended values ​​of the predicted region-industry edges are obtained, on the one hand, a list of regions recommended for development of each industry can be obtained, and on the other hand, a list of industries recommended for development of each region can be obtained.

[0114] For each industry, the recommendation value (recommendation probability) of each region is sorted. The region with a high recommendation value indicates that the industry has great development potential in the region. By setting a threshold or selecting the first L regions with the highest recommendation values, these regions are included in the list of regions recommended for the development of the industry.

[0115] For each region, the recommended values ​​of each industry can be sorted. Industries with high recommended values ​​indicate high suitability for development in the region. By setting a threshold or selecting the first Q industries with the highest recommended values, these industries can be included in the recommended industry list for the region.

[0116] It can be seen that the industrial layout prediction method proposed in the embodiments of the present disclosure can integrate technologies such as large language models, deep neural networks, attention mechanisms, spatio-temporal graph algorithms, and adaptive recommendations, develop a regional recommendation algorithm for the layout of future industries according to local conditions, construct a high-quality database integrating machine intelligence and subjective insights, and a new paradigm for decision-making support. On the basis of fully considering factors such as resource endowment, industrial foundation, scientific research conditions, talent structure, and location advantages, the evolution and development laws of industries and technologies are introduced from the time dimension, and the layout synergy effect between different regions is introduced from the space dimension, providing an optional tool for governments at all levels to achieve dynamic tracking, real-time updating, non-intrusive monitoring, and global optimization of future technologies and industries, improving decision-making accuracy and efficiency, and reducing decision-making costs and risks.

[0117] As Figure 5 shown, the embodiments of the present disclosure also propose an industrial layout prediction device, including:

[0118] A construction unit 501, configured to construct a heterogeneous graph for each year according to the industrial data and core technology data of the target industry in different years. The nodes in the heterogeneous graph include region nodes, industrial nodes, and core technology nodes, and the edges in the heterogeneous graph are used to represent the association relationship between two connected nodes;

[0119] A prediction unit 502, configured to input the heterogeneous graph data of continuous N years into a pre-trained industrial layout prediction network to predict the recommended value of the edge between the region and the industry in the future year;

[0120] Wherein, in the training samples of the network, the true value of the edge between the region and the industry is obtained according to the potential score of the core technology to which the industry belongs in the corresponding year and the investment weight of the region in the industry.

[0121] In some embodiments, the device further includes a time-series graph conversion unit, configured to:

[0122] Adopt a sliding time window method to obtain a set of time-series graphs according to the heterogeneous graphs of continuous M years in each time window, so that the heterogeneous graph data of continuous N years includes (N - M + 1) sets of time-series graphs, where M < N, and the industrial layout prediction network is used to predict the recommended value of the edge between the region and the industry in the future (M - 1) years.

[0123] In some embodiments, when the time-series graph conversion unit is used to obtain a set of time-series graphs according to the heterogeneous graphs of continuous M years in each time window, it is specifically configured to:

[0124] For the heterogeneous graph of each year, enable nodes of a set type to obtain messages from neighbor nodes, aggregate the received messages, and then update the node representation using the aggregation result;

[0125] Obtaining an information sequence according to information of each node in the heterogeneous graph;

[0126] The information sequences corresponding to each year in M ​​consecutive years are spliced ​​together to obtain a set of time series diagrams.

[0127] In some embodiments, each node and each edge in the heterogeneous graph has attribute information of a set number of dimensions, and each edge also has a unified embedding representation formed by mapping attributes of edges of the same type to a low-dimensional space.

[0128] In some embodiments, the industrial layout prediction network is a long short-term memory network.

[0129] In some embodiments, when the construction unit is used to construct a heterogeneous graph for each year according to the industry data and core technology data of the target industry in different years, it is specifically used to:

[0130] According to the industry data of each year, establish the industry nodes in the heterogeneous graph of that year;

[0131] According to the core technology data of the year, establish the core technology nodes in the heterogeneous graph of the year;

[0132] By pairing each region with each industry and each core technology, edges between region nodes and industry nodes, and edges between region nodes and core technology nodes are established;

[0133] By identifying the core technologies related to each industry, edges between industry nodes and core technology nodes are established.

[0134] In some embodiments, when the construction unit is used to establish an edge between an industry node and a core technology node by identifying the core technology related to each industry, it is specifically used to:

[0135] Based on the patent application data of the target industry within one year, establish patent nodes and edges between patent nodes and regional nodes;

[0136] According to the international patent classification number and abstract content of the patent application, establish the edge between the patent node and the core technology node;

[0137] According to the international patent classification number of the patent application, the core technology is mapped to the industrial node, and the edges between the core technology node and the industrial node are established.

[0138] In some embodiments, when the training samples also include core technologies marked as potential technologies, the industrial layout prediction network is also used to predict whether the core technology is a potential technology.

[0139] In some embodiments, the device further includes a recommendation unit for obtaining a list of regions recommended for development of each industry and a list of industries recommended for development of each region based on the predicted recommendation values ​​of the edges between regions and industries.

[0140] For other details and beneficial effects of this embodiment, please refer to the above related introduction, which will not be repeated here.

[0141] Figure 6 An electronic device provided for at least one embodiment of the present disclosure includes a memory and a processor, wherein the memory is used to store computer instructions executable on the processor, and the processor is used to implement the industrial layout prediction method described in any embodiment or implementation of the present disclosure when executing the computer instructions.

[0142] At least one embodiment of the present disclosure further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the industrial layout prediction method described in any embodiment or implementation of the present disclosure.

[0143] It should be understood by those skilled in the art that one or more embodiments of the present specification may be provided as a method, system or computer program product. Therefore, one or more embodiments of the present specification may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, one or more embodiments of the present specification may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0144] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the data processing device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0145] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0146] The embodiments of the subject matter and functional operations described in this specification may be implemented in the following: digital electronic circuits, tangibly embodied computer software or firmware, computer hardware including the structures disclosed in this specification and their structural equivalents, or a combination of one or more of them. The embodiments of the subject matter described in this specification may be implemented as one or more computer programs, i.e., one or more modules in computer program instructions encoded on a tangible non-temporary program carrier to be executed by a data processing device or to control the operation of the data processing device. Alternatively or additionally, the program instructions may be encoded on an artificially generated propagation signal, such as a machine-generated electrical, optical or electromagnetic signal, which is generated to encode information and transmit it to a suitable receiver device for execution by a data processing device. The computer storage medium may be a machine-readable storage device, a machine-readable storage substrate, a random or serial access memory device, or a combination of one or more of them.

[0147] The processes and logic flows described in this specification can be performed by one or more programmable computers executing one or more computer programs to perform corresponding functions by operating on input data and generating output. The processes and logic flows can also be performed by special purpose logic circuits, such as FPGAs (field programmable gate arrays) or ASICs (application-specific integrated circuits), and the apparatus can also be implemented as special purpose logic circuits.

[0148] Computers suitable for executing computer programs include, for example, general and / or special microprocessors, or any other type of central processing unit. Typically, the central processing unit will receive instructions and data from a read-only memory and / or a random access memory. The basic components of a computer include a central processing unit for implementing or executing instructions and one or more memory devices for storing instructions and data. Typically, the computer will also include one or more large-capacity storage devices for storing data, such as magnetic disks, magneto-optical disks, or optical disks, or the computer will be operably coupled to this large-capacity storage device to receive data from it or to transmit data to it, or both. However, the computer does not necessarily have such a device. In addition, the computer can be embedded in another device, such as a mobile phone, a personal digital assistant (PDA), a mobile audio or video player, a game console, a global positioning system (GPS) receiver, or a portable storage device such as a universal serial bus (USB) flash drive, to name a few.

[0149] Computer readable media suitable for storing computer program instructions and data include all forms of non-volatile memory, media and memory devices, including, for example, semiconductor memory devices (e.g., EPROM, EEPROM and flash memory devices), magnetic disks (e.g., internal hard disks or removable disks), magneto-optical disks, and CD ROM and DVD-ROM disks. The processor and memory may be supplemented by, or incorporated in, special purpose logic circuitry.

[0150] Although this specification includes many specific implementation details, these should not be interpreted as limiting the scope of any invention or the scope of protection claimed, but are mainly used to describe the features of the specific embodiments of specific inventions. Certain features described in multiple embodiments in this specification may also be implemented in combination in a single embodiment. On the other hand, the various features described in a single embodiment may also be implemented separately in multiple embodiments or in any suitable sub-combination. In addition, although features may work in certain combinations as described above and even initially claim protection, one or more features from the claimed combination may be removed from the combination in some cases, and the claimed combination may point to a sub-combination or a variation of a sub-combination.

[0151] Similarly, although operations are depicted in a particular order in the accompanying drawings, this should not be understood as requiring that these operations be performed in the particular order shown or performed sequentially, or requiring that all illustrated operations be performed to achieve the desired results. In some cases, multitasking and parallel processing may be advantageous. In addition, the separation of various system modules and components in the above-described embodiments should not be understood as requiring such separation in all embodiments, and it should be understood that the described program components and systems can generally be integrated together in a single software product, or packaged into multiple software products.

[0152] Thus, specific embodiments of the subject matter have been described. Other embodiments are within the scope of the appended claims. In some cases, the actions recited in the claims can be performed in a different order and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the particular order or sequential order shown to achieve the desired results. In some implementations, multitasking and parallel processing may be advantageous.

[0153] The above description is merely a preferred embodiment of one or more embodiments of the present specification and is not intended to limit one or more embodiments of the present specification. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of one or more embodiments of the present specification shall be included in the scope of protection of one or more embodiments of the present specification.

Claims

1. A method for predicting industrial layout, characterized in that: Including: Construct an heterogeneous graph for each year based on the industrial data and core technology data of the target industry in different years. The nodes in the heterogeneous graph include regional nodes, industrial nodes, and core technology nodes, and the edges in the heterogeneous graph are used to represent the association relationship between two connected nodes. Input the heterogeneous graph data for consecutive N years into a pre-trained industrial layout prediction network to predict the recommended value of the edge between regions and industries in future years. Among them, in the training samples of the network, the true value of the edge between regions and industries is obtained based on the potential score of the core technology to which the industry belongs in the corresponding year and the investment weight of the region in the industry.

2. The method according to claim 1, characterized in that The method further includes: Adopt a sliding time window method to obtain a set of time series graphs based on the heterogeneous graphs for consecutive M years within each time window, so that the heterogeneous graph data for consecutive N years includes (N - M + 1) sets of time series graphs, where M < N, and the industrial layout prediction network is used to predict the recommended value of the edge between regions and industries in the next (M - 1) years.

3. The method according to claim 2, characterized in that The obtaining of a set of time series graphs based on the heterogeneous graphs for consecutive M years within each time window includes: For the heterogeneous graph of each year, enable nodes of a set type to obtain messages from neighbor nodes, aggregate the received messages, and then update the node representation using the aggregation result. Obtain an information sequence according to the information of each node in the heterogeneous graph. Concatenate the information sequences corresponding to each year in consecutive M years to obtain a set of time series graphs.

4. The method according to claim 1, characterized in that: Each node and each edge in the heterogeneous graph respectively have attribute information with a first set number of dimensions, and each edge also has an embedding representation with a second set number of dimensions formed by mapping the attributes of the same type of edges into a low-dimensional space.

5. The method according to any one of claims 2 to 4, characterized in that: The industrial layout prediction network is a long short-term memory network.

6. The method according to claim 1, characterized in that The constructing of the heterogeneous graph for each year based on the industrial data and core technology data of the target industry in different years includes: Based on the industrial data of each year, establish industrial nodes in the heterogeneous graph of that year. Based on the core technology data of that year, establish core technology nodes in the heterogeneous graph of that year. By pairing each region with each industry and each core technology respectively, establish edges between regional nodes and industrial nodes, and edges between regional nodes and core technology nodes. By identifying the core technologies related to each industry, establish edges between industrial nodes and core technology nodes.

7. The method according to claim 6, characterized in that The establishing of the edges between industrial nodes and core technology nodes by identifying the core technologies related to each industry includes: Based on the patent application data of the target industry in one year, establish patent nodes and edges between patent nodes and regional nodes. Based on the international patent classification number of the patent application and the abstract content of the patent, establish edges between patent nodes and core technology nodes. Based on the international patent classification number of the patent application, map the core technology to industrial nodes and establish edges between core technology nodes and industrial nodes.

8. The method according to claim 1, characterized in that In the case where the training samples also include core technologies labeled with whether they are potential technologies, the industrial layout prediction network is also used to predict whether a core technology is a potential technology.

9. The method according to claim 1, characterized in that: The method further comprises: obtaining a list of regions recommended for development of each industry and a list of industries recommended for development of each region according to the predicted recommendation values ​​of the edges between the regions and the industries.

10. An industrial layout prediction device, characterized in that: include: A construction unit, used to construct a heterogeneous graph for each year according to the industry data and core technology data of the target industry in different years, wherein the nodes in the heterogeneous graph include regional nodes, industry nodes and core technology nodes, and the edges in the heterogeneous graph are used to represent the association relationship between two connected nodes; The prediction unit is used to input N consecutive years of heterogeneous graph data into the pre-trained industrial layout prediction network to predict the recommended values ​​of the edges between regions and industries in future years; Among them, in the training samples of the network, the true value of the edge between the region and the industry is obtained according to the potential score of the core technology of the industry in the corresponding year and the investment weight of the region in the industry.

11. An electronic device, characterized in that: The device comprises a memory and a processor, wherein the memory is used to store computer instructions executable on the processor, and the processor is used to implement the method according to any one of claims 1 to 9 when executing the computer instructions.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 9 is implemented.

Images