A technology transfer supply chain management method and system employing a dynamic matching algorithm

Abstract

This application provides a technology transfer supply chain management method and system employing a dynamic matching algorithm, relating to the field of supply chain management technology. The method includes: constructing feature vectors for technology suppliers, demanders, and intermediary nodes to obtain a node feature vector set; obtaining an initial transfer path based on the node feature vector set and a directed acyclic graph method; collecting real-time information to construct a multi-level graph containing multiple transfer nodes and their collaborative relationships; obtaining transfer prediction results by combining historical transfer data and node attribute information; optimizing the path using a multi-objective dynamic matching algorithm based on the transfer prediction results and dynamic information from upstream and downstream of the supply chain to obtain the target technology transfer path; and performing technology transfer supply chain management based on the target technology transfer path. This solves the technical problems of prominent capability bottlenecks and low conversion success rates in existing technology transfer supply chain management technologies.

Description

Technical Field

[0001] This invention relates to the field of supply chain management, and in particular to a technology transfer supply chain management method and system that employs a dynamic matching algorithm. Background Technology

[0002] As the core link connecting technological innovation and industrial application, the efficient management of the technology transfer industry chain is directly related to the efficiency of transforming technological achievements into real productivity. It is of key significance for optimizing the allocation of innovation resources and improving the overall innovation efficiency of the industry chain.

[0003] However, traditional technology transfer industry chain management often adopts a static model, which suffers from problems such as information asymmetry and low matching efficiency between technology suppliers, demanders and intermediary nodes. Furthermore, it ignores the dynamic changes in the upstream and downstream of the industry chain, such as resource fluctuations, demand adjustments and policy changes, resulting in a disconnect between the technology transfer path and the actual scenario, which in turn leads to prominent capability bottlenecks and a low success rate of transformation.

[0004] Therefore, there is an urgent need for a management method that can perform dynamic matching to overcome the limitations of traditional static management. Summary of the Invention

[0005] This invention addresses the technical problems of prominent capability bottlenecks and low conversion success rates in existing technology transfer supply chain management by providing a technology transfer supply chain management method and system using a dynamic matching algorithm.

[0006] The technical solution of the present invention to solve the above-mentioned technical problems is as follows:

[0007] In a first aspect, the present invention provides a technology transfer supply chain management method employing a dynamic matching algorithm, comprising:

[0008] Construct feature vectors for technology suppliers, demanders, and intermediary nodes, and obtain node feature vector sets;

[0009] Based on the node feature vector set, the initial transition path is obtained by combining the directed acyclic graph method;

[0010] Real-time information is collected based on the initial transfer path, and a multi-level graph containing multiple transfer nodes and collaborative relationships is constructed based on the real-time information collection results.

[0011] By combining historical transfer data and node attribute information, the multi-level graph is traversed to predict capacity bottlenecks and conversion success rates, and the transfer prediction results are obtained.

[0012] Based on the transfer prediction results and combined with the dynamic information of the upstream and downstream of the industry chain, a multi-objective dynamic matching algorithm is used to optimize the path, obtain the target technology transfer path, and perform technology transfer industry chain management based on the target technology transfer path.

[0013] Secondly, this invention provides a technology transfer supply chain management system employing a dynamic matching algorithm, comprising:

[0014] The vector construction module is used to construct feature vectors for technology suppliers, demanders, and intermediary nodes, and to obtain the node feature vector set.

[0015] The initial path acquisition module is used to acquire the initial transition path based on the node feature vector set and combined with the directed acyclic graph method.

[0016] The graph construction module is used to collect real-time information based on the initial transfer path and construct a multi-level graph containing multiple transfer nodes and cooperative relationships based on the real-time information collection results.

[0017] The transfer prediction module is used to combine historical transfer data and node attribute information to traverse the multi-level graph to predict capacity bottlenecks and conversion success rates, and obtain transfer prediction results.

[0018] The optimized output module is used to optimize the path based on the transfer prediction results and the dynamic information of the upstream and downstream of the industrial chain, using a multi-objective dynamic matching algorithm to obtain the target technology transfer path, and to perform technology transfer industrial chain management based on the target technology transfer path.

[0019] The beneficial effects of this invention are:

[0020] Compared to existing technologies, this application first constructs feature vectors for technology suppliers, demanders, and intermediary nodes, obtaining a set of node feature vectors. This provides a unified data foundation for subsequent path calculation based on directed acyclic graphs (DAGs) and optimization of dynamic matching algorithms, ensuring quantitative analysis and accurate decision-making throughout the technology transfer process. Secondly, based on the node feature vector set, and combined with the DAG method, the initial transfer path is obtained, clarifying the core flow nodes and hierarchical relationships from supply to demand. This transforms the technology transfer path from qualitative judgment to quantitative calculation, providing a structured foundation for more refined management. Thirdly, real-time information is collected based on the initial transfer path, and a multi-level graph containing multiple transfer nodes and collaborative relationships is constructed based on the real-time information collection results. This refines the initial transfer path into a structured network containing specific execution units and collaborative relationships, providing more realistic underlying data support for subsequent accurate prediction and optimization. Furthermore, by combining historical transfer data and node attribute information, multi-level graphs are traversed to predict capacity bottlenecks and conversion success rates, obtaining transfer prediction results. This provides accurate capacity bottleneck and conversion success rate assessments for each candidate path, laying a data-driven decision-making foundation for subsequent optimization. Finally, based on the transfer prediction results and combined with dynamic information from upstream and downstream of the industry chain, a multi-objective dynamic matching algorithm is used for path optimization to obtain target technology transfer paths. Technology transfer industry chain management is then implemented based on these target paths, enabling the technology transfer paths to adapt to changes in the industry chain while achieving a multi-objective balance between minimizing capacity bottlenecks and maximizing conversion success rates. This avoids the failure issues caused by environmental changes during the execution of static paths, ultimately achieving a dual improvement in technology transfer efficiency and stability.

[0021] Through the aforementioned technical solutions, this application addresses the problem of inefficient matching caused by the ambiguity of main features by quantifying the multidimensional attributes of suppliers, demanders, and intermediary nodes using feature vector quantification technology. It also refines the collaborative relationships between nodes through multi-level graphs, overcoming the limitations of unclear node connections in traditional coarse frameworks. Furthermore, by combining a transfer prediction model with a multi-objective dynamic matching algorithm, it rapidly adjusts paths based on dynamic changes in the upstream and downstream of the industry chain, effectively avoiding the risk of path disconnection under static management. This improves the matching accuracy, process efficiency, and conversion success rate of technology transfer industry chain management, accelerating the transformation of technological achievements into industrial applications. Attached Figure Description

[0022] Figure 1 A flowchart illustrating a technology transfer supply chain management method employing a dynamic matching algorithm, provided by this invention;

[0023] Figure 2 This invention provides a schematic diagram of a technology transfer supply chain management system employing a dynamic matching algorithm.

[0024] In the attached diagram, the components represented by each number are as follows:

[0025] Vector construction module 11, initial path acquisition module 12, graph construction module 13, transition prediction module 14, and optimized output module 15. Detailed Implementation

[0026] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] In the description of this invention, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0028] In the description of this invention, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this invention is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed herein.

[0029] Example 1, as Figure 1 As shown, this embodiment of the invention provides a technology transfer supply chain management method using a dynamic matching algorithm, including:

[0030] S10: Construct feature vectors for technology suppliers, demanders, and intermediary nodes, and obtain node feature vector sets.

[0031] The technology transfer industry chain involves technology suppliers, demanders, and intermediary nodes. Traditional methods suffer from problems such as information asymmetry and low matching efficiency, resulting in insufficient accuracy in matching technology supply and demand, and lagging planning of transfer paths. This can easily lead to problems such as resource misallocation and hindered industry chain collaboration.

[0032] To address the aforementioned issues, this application constructs feature vectors for technology suppliers, demanders, and intermediary nodes to obtain a set of node feature vectors.

[0033] Specifically, step S10 in the method includes:

[0034] The vector dimensions of the feature vector include technical attributes, subject attributes, and service attributes;

[0035] The technical attributes include at least the technology type, technology maturity, technical indicators, technological innovation points, and technological relevance;

[0036] The subject attributes include at least the subject size, subject financial status, subject credit rating, and subject cooperation history;

[0037] The service attributes include at least the service scope, service success rate, service efficiency, and service resources.

[0038] In this embodiment, the vector dimension of the feature vector includes technical attributes, subject attributes, and service attributes, which can comprehensively characterize the node features and provide objective and comprehensive quantitative support for subsequent path planning and matching calculation.

[0039] Among these, technological attributes reflect the core characteristics of the technology itself, including at least technology type, technology maturity, technological indicators, technological innovation points, and technological relevance: Technology type refers to the field to which the technology belongs, such as artificial intelligence, biomedicine, and new materials, used to accurately match the industry direction of the demand side; technology maturity is used to quantify the development stage of the technology, such as the laboratory stage or the commercialization stage; technological indicators include quantitative standards such as core performance parameters and cost thresholds; technological innovation points reflect the core advantages of the technology, such as breakthroughs in core technologies and performance leaps; and technological relevance reflects the degree of compatibility with upstream and downstream technologies. These attributes together accurately depict the essential characteristics of the technology and are the core basis for determining the degree of technology matching.

[0040] Among these, the entity attributes reflect the node's own capabilities and reputation level, and at least include the entity's size, financial status, creditworthiness, and cooperation history: entity size includes the number of employees and institutional qualification level; entity financial status includes revenue and cash reserves; entity creditworthiness includes historical default rate and partner evaluations; and entity cooperation history includes the success rate and duration of past technology transfer projects. These attributes comprehensively reflect the node's execution capabilities and cooperation reliability, and are key criteria for assessing the feasibility of cooperation.

[0041] Among these, service attributes are used to measure the node's ability to provide supporting services, including at least service scope, service success rate, service efficiency, and service resources: service scope refers to the geographical area or technical field covered by the service; service success rate refers to the proportion of successful conversions of historical service projects; service efficiency refers to the service response speed and completion cycle; service resources include the available resource pool, such as the number of testing equipment, the size of the expert database, and policy access channels. These attributes collectively reflect the node's service support capabilities in technology transfer, providing an important reference for ensuring the feasibility of technology transfer implementation.

[0042] Specifically, the phrase "constructing feature vectors for technology suppliers, demanders, and intermediary nodes, and obtaining a set of node feature vectors" includes:

[0043] Based on the supply capacity data of technology suppliers, we analyze and extract the technical attributes, subject attributes, and service attributes of technology supply nodes to form a feature vector of technology supply nodes.

[0044] Based on the demand data from the demand side, analyze and extract the technical attributes, subject attributes, and service attributes of the demand nodes to form a feature vector of the demand nodes;

[0045] The intermediary nodes are clustered to obtain a set of intermediary node categories. Based on big data, statistical analysis is performed on the set of intermediary node categories to extract typical capability parameters of each intermediary node category and generate an intermediary node feature vector.

[0046] The feature vectors of the technology supply nodes, the feature vectors of the demand nodes, and the feature vectors of the intermediary nodes are summarized to form the node feature vector set.

[0047] In this embodiment, firstly, based on the supply capacity data of the technology supplier, the technical attributes, subject attributes, and service attributes of the technology supply nodes are analyzed and extracted to form a feature vector of the technology supply nodes. The supply capacity data includes technical specifications, patent documents, production qualification documents, and company annual reports. For example, based on the supply capacity data of the technology supplier, such as technical specifications, patent documents, production qualification documents, and company annual reports, the technical attributes, subject attributes, and service attributes of the technology supply nodes are extracted to form a feature vector of the technology supply nodes. For example, {Technical attributes: [Technology type = New energy battery, TRL = 7, Technical indicators = Energy density 350Wh / kg, Technological innovation point = Solid electrolyte, Technological relevance = Requires adaptation to BMS system], Subject attributes: [Entity size = Listed company, Enterprise financial status = AAA rating, Enterprise credit rating = 98%, Enterprise cooperation history = 120 times], Service attributes: [Service scope = Global authorization, Service success rate = 95%, Service efficiency = Mass production in 6 months, Service resources = 10 contract manufacturers]}, can reflect the technology supply capacity and supporting service level of the technology supplier.

[0048] Secondly, based on the demander's requirements data, the technical attributes, subject attributes, and service attributes of the demand nodes are analyzed and extracted to form a demand node feature vector. The demand data includes a list of technical requirements, application scenario descriptions, etc. For example, based on the demander's technical requirement list, application scenario descriptions, and other demand data, the technical attributes, subject attributes, and service attributes of the demand nodes are extracted to form a demand node feature vector. For instance, {Technical Attributes: [Technology Type = Autonomous Driving Algorithm, TRL Requirement = 6, Technical Indicator = Latency < 50ms, Technical Innovation Requirement = Multi-Sensor Fusion, Subject Attributes: [Subject Size = Medium-Sized Automaker, Subject Reputation = 91%, Subject Cooperation History = 35 Times], Service Attributes: [Service Scope = Localized Deployment, Service Success Rate Requirement > 85%, Service Efficiency = Delivery within 3 Months]} can reflect the demander's technical requirement standards and cooperation conditions.

[0049] Secondly, intermediary nodes are clustered according to their service scope using a clustering algorithm to obtain a set of intermediary node categories. Then, based on big data, statistical analysis is performed on these categories to extract typical capability parameters for each category, such as average service success rate and resource coverage, generating intermediary node feature vectors. For example, K-means clustering (k=5) is performed on the intermediary nodes according to their service scope to obtain a set of intermediary node categories, such as {A: testing and certification, B: financing matching, ...}. Then, based on big data, statistical analysis is performed on these categories to extract typical capability parameters for each category. For example, if the testing efficiency of intermediary node category A is 10 days, then the feature vector for category A is generated as follows: {Technical attributes: [Service technology field = electronics / mechanical, certification standard coverage = 98%], Subject attributes: [Average size = 50 people, industry reputation rating = AA], Service attributes: [Scope = global, success rate = 99%, efficiency = 10 days]}. This highlights the commonalities of the categories, facilitating efficient matching of intermediary nodes.

[0050] Finally, the feature vectors of the technology supply nodes, demand nodes, and intermediary nodes are summarized to form a node feature vector set, providing a quantitative comparison basis for subsequent DAG path planning.

[0051] In summary, compared to existing technologies, this application constructs feature vectors for technology suppliers, demanders, and intermediary nodes to obtain a set of node feature vectors. Thus, by constructing feature vectors and obtaining a set of node feature vectors, a unified data foundation is provided for subsequent path calculation based on directed acyclic graphs and optimization of dynamic matching algorithms, ensuring quantitative analysis and accurate decision-making throughout the entire technology transfer process.

[0052] S20: Based on the node feature vector set, the initial transition path is obtained by combining the directed acyclic graph method.

[0053] Technology transfer from supply to demand involves multiple nodes. Traditional methods often result in unclear transfer paths, high matching costs, and redundant links due to the lack of quantitative analysis of the relationships between nodes and the assessment of path priorities.

[0054] To address the aforementioned issues, this application uses the node feature vector set and a directed acyclic graph method to obtain the initial transition path.

[0055] Specifically, step S20 in the method includes:

[0056] Based on the node feature vector set, a directed acyclic graph is constructed with the technology supply node as the starting point, multiple intermediary nodes as intermediate nodes, and the demand node as the ending point. The edges of the directed acyclic graph are unidirectional edges with the direction pointing from the technology supply node to the demand node.

[0057] In the directed acyclic graph, based on the principle of minimizing edge weight, the shortest path algorithm is used to calculate the path from the technology supply node to the demand node through at least one of the intermediary nodes to obtain the initial transfer path;

[0058] The initial transfer path is used to define the hierarchical path of technology transfer.

[0059] In this embodiment, a directed acyclic graph (DAG) is first constructed. Specifically, based on the node feature vector set, a DAG is constructed with the technology supply node as the starting point, multiple intermediary nodes as intermediate nodes, and the demand node as the ending point. The edges of the DAG are unidirectional, pointing from the technology supply node to the demand node. For example, the nodes of the DAG are the technology supply node (technology provider), multiple intermediary nodes, and the demand node (technology receiver). The edges are unidirectional, and their directions strictly follow the logic of technology supply node → intermediary node → demand node, meaning they can only point from upstream to downstream, and reverse or cyclical directions are not allowed. This ensures the acyclicity of the graph and avoids logical contradictions in the technology transfer path.

[0060] Secondly, in a directed acyclic graph (DAG), the principle of minimizing edge weights is followed, and the shortest path algorithm is used to calculate the path from the technology supply node to the demand node via at least one intermediary node, obtaining the initial transfer path. This initial transfer path defines the hierarchical path of technology transfer. For example, the edge weights in a DAG reflect the path transfer cost and can be calculated using the following formula: Edge weight = 1 / (Technology matching degree × Node service capacity coefficient × Historical success rate). Here, the technology matching degree is the cosine similarity of the technical attributes of the technology supply node and the demand node; the service capacity coefficient is the weighted value of the service success rate and service efficiency of the intermediary node; and the historical success rate is the past success rate of cooperation between the two nodes. A smaller edge weight represents higher matching degree, higher service capacity, and higher success rate. Therefore, minimizing edge weights is the principle in a DAG. For example, in a directed acyclic graph, a shortest path algorithm, such as Dijkstra's algorithm or Floyd's algorithm, is used to calculate the total edge weight of all possible paths from the technology supply node through at least one intermediary node to the demand node. The path with the smallest total edge weight is selected as the initial transition path. The initial transition path is a simplified hierarchical relationship chain, for example, technology supply node (supplier) → intermediary node A → intermediary node B → demand node (demander), which clarifies the core flow nodes, hierarchical relationships and flow order from supply to demand.

[0061] In summary, compared to existing technologies, this application obtains the initial transfer path based on the aforementioned node feature vector set and combined with the directed acyclic graph method. Thus, by transforming node features into a graphical network and calculating the optimal path, the core flow nodes and hierarchical relationships from supply to demand are clearly identified, realizing the transformation of technology transfer path from qualitative judgment to quantitative calculation, and providing a structured foundation for subsequent more refined management.

[0062] S30: Real-time information is collected based on the initial transfer path, and a multi-level graph containing multiple transfer nodes and collaborative relationships is constructed based on the real-time information collection results.

[0063] The initial transfer path, as a rough framework for technology transfer, only includes the macro-level relationship between the supply side, intermediary nodes, and demand side. In reality, each intermediary node often contains multiple transfer nodes that perform specific functions.

[0064] To address the aforementioned issues, this application collects real-time information based on the initial transfer path and constructs a multi-level graph containing multiple transfer nodes and their collaborative relationships based on the real-time information collection results.

[0065] Specifically, step S30 in the method includes:

[0066] Based on the initial transfer path, extract multiple transfer nodes corresponding to each intermediary node and generate a list;

[0067] Information is retrieved from the list, and then regularized and standardized to obtain the entity feature information of the transfer node;

[0068] Based on the entity feature information, the initial transfer path is expanded accordingly to generate a multi-level graph. Each level of the multi-level graph corresponds to the technology supply node, the demand node, and multiple intermediary nodes. Each intermediary node has multiple transfer nodes under it. Each transfer node represents its capabilities and collaborative relationships in graph form.

[0069] In this embodiment, multiple transfer nodes corresponding to each intermediary node are first extracted based on the initial transfer path to generate a list. Each transfer node is a specific execution unit under the intermediary node, and the intermediary node and transfer nodes are in a parent-child hierarchy, connected through a collaborative relationship. For example, multiple transfer nodes are extracted from intermediary node A in the initial path: technology supply node → intermediary node A → demand node, such as transfer node A1, transfer node A2, and transfer node A3. If intermediary node A is a testing institution, then transfer nodes A1, A2, and A3 may be for material testing, safety assessment, and certification issuance, respectively. Thus, generating a transfer node list based on multiple transfer nodes clarifies the specific objects for which information needs to be collected, defining the scope for subsequent information acquisition and avoiding blind information collection.

[0070] Secondly, information is retrieved from the list and then regularized and standardized to obtain the entity characteristic information of the transfer nodes. This information can be retrieved from internal enterprise systems (such as ERP and CRM) and IoT devices (such as multimodal sensors). For example, operational data of the transfer nodes in the list, such as equipment status, task queues, and personnel size, is retrieved via real-time API. This data is then regularized and standardized. Regularization unifies the information dimensions of different transfer nodes, eliminating incomparability caused by differences in units of measurement. Standardization involves data cleaning to remove extreme values ​​and normalization, reducing noise interference and ensuring data stability and consistency. This transforms the data into structured characteristic information. The entity characteristic information of the transfer nodes reflects their real-time capabilities (such as efficiency and resource reserves) and status (such as load), serving as core data for characterizing the actual operation of the transfer nodes.

[0071] Finally, based on entity feature information, the initial transfer path is expanded to generate a multi-level graph. Each level of the multi-level graph corresponds to a technology supply node, a demand node, and multiple intermediary nodes, with each intermediary node overseeing multiple transfer nodes. Each transfer node's capabilities and collaborative relationships are represented in graph form. For example, based on entity feature information, the initial transfer path is expanded into a multi-level graph. The top layer represents the technology supply node, the bottom layer represents the demand node, and the middle layer represents each intermediary node and its subordinate transfer nodes. An attribute graph model is used to accurately describe the node capabilities (such as efficiency) and collaborative relationships (such as sequential dependency or parallel execution) of each transfer node. Thus, the initial transfer path, initially a coarse framework of technology supply node → intermediary node A → demand node, is refined into a specific operational chain of technology supply node → intermediary node A {transfer node A1, transfer node A2, transfer node A3} → demand node. This clearly demonstrates the participants, capabilities, and collaborative logic of each micro-level link in the technology transfer process, providing a visual analytical framework for subsequent prediction of path bottlenecks and optimization of collaborative efficiency.

[0072] In summary, compared to existing technologies, this application collects information in real time based on the initial transfer path and constructs a multi-level graph containing multiple transfer nodes and their collaborative relationships based on the real-time information collection results. This refines the initial transfer path into a structured network containing specific execution units and their collaborative relationships, providing more realistic underlying data support for subsequent accurate prediction and optimization.

[0073] S40: Combining historical transfer data and node attribute information, traverse the multi-level graph to predict capacity bottlenecks and conversion success rates, and obtain transfer prediction results.

[0074] Capability bottlenecks refer to the obstacles encountered during technology transfer due to insufficient resource allocation, poor process coordination, or capability mismatch at various nodes. These bottlenecks manifest as task delays, process interruptions, or resource waste. Success rate refers to the probability that technology, originating from the supplier, is successfully received and applied by the demand side after collaborative efforts across intermediary nodes. It comprehensively reflects the collaborative efficiency, technological adaptability, and resource support capabilities of each node. In technology transfer supply chain management, capability bottlenecks directly lead to extended transfer cycles, increased costs, and even path interruptions. Success rate determines the quality of technological achievement commercialization; a low success rate results in resource misallocation and investment losses, both of which are influenced by the dynamic state of nodes and exhibit significant uncertainty. Therefore, it is necessary to predict these bottlenecks in advance using transfer forecasting models to provide quantitative basis for path optimization.

[0075] To address the aforementioned issues, this application combines historical transfer data and node attribute information to traverse the multi-level graph to predict capacity bottlenecks and conversion success rates, thereby obtaining transfer prediction results.

[0076] Specifically, step S40 in the method includes:

[0077] Based on the multi-level graph, multiple levels are traversed from top to bottom to randomly select multiple transfer nodes to obtain alternative technology transfer paths, and a set of alternative technology transfer paths is obtained iteratively.

[0078] Clustering is performed based on the node attribute information of the multiple transfer nodes obtained, and the clustering results are output as a list of multiple node clusters;

[0079] Based on the node cluster list, the historical transfer data is homomorphically partitioned to form multiple node historical transfer data clusters;

[0080] Using the historical transfer data clusters of the aforementioned nodes as sample data, multiple transfer prediction models are constructed and supervised training is performed.

[0081] The entity feature information of multiple transfer nodes is input into multiple transfer prediction models to predict the capacity bottleneck and conversion success rate. The prediction results of capacity bottleneck and conversion success rate are superimposed and fused according to the set of alternative technology transfer paths to obtain the path capacity bottleneck and path conversion success rate of each alternative technology transfer path, and the output is the transfer prediction result.

[0082] In this embodiment, firstly, based on a multi-level graph, multiple levels are traversed from top to bottom to randomly select multiple transfer nodes. That is, starting from the technology supply node, the process passes through the intermediate nodes of the middle level and their subordinate transfer nodes in sequence, and finally reaches the demand node. Different combinations of transfer nodes are randomly selected to obtain alternative technology transfer paths, and the set of alternative technology transfer paths is obtained iteratively. Random selection is used to avoid getting trapped in local optima and to fully explore the potential feasible paths in the multi-level graph. For example, from the multi-level graph: technology supply node → intermediary node A {transfer node A1, transfer node A2, transfer node A3} → demand node, multiple levels are traversed from top to bottom to randomly select multiple transfer nodes to obtain candidate technology transfer paths. For example, technology supply node → transfer node A1 → demand node, technology supply node → transfer node A2 → demand node, technology supply node → transfer node A3 → demand node. By iterating and repeating the above selection process multiple times, a set of candidate technology transfer paths containing multiple different combinations of transfer nodes is generated to ensure that enough potential paths are covered and to avoid the optimal solution being missed due to insufficient path enumeration.

[0083] Secondly, clustering is performed based on the node attribute information of the acquired multiple transfer nodes, and the clustering results are output as a list of multiple node clusters. For example, based on the node attribute information of the acquired multiple transfer nodes, such as service range, service success rate, service efficiency, service resources, etc., clustering algorithms (such as K-means, hierarchical clustering) are used to cluster the multiple transfer nodes into multiple node clusters. For example, material detection transfer nodes with the same service range are clustered into one node cluster. Clustering is performed in the same way, and finally a list of multiple node clusters is formed, clarifying the transfer nodes contained in each cluster and their common characteristics. In this way, transfer nodes with similar attributes are grouped into one category, reducing prediction complexity and improving prediction efficiency and accuracy.

[0084] Secondly, the historical transfer data is homomorphically partitioned based on the node cluster list, forming multiple historical transfer data clusters. Homomorphic partitioning means that the clustering method of the historical transfer data is consistent with the classification logic of the node clusters; that is, the historical transfer data is partitioned according to the node cluster labels, ensuring that the training data and the prediction target are distributed in the same way. For example, based on the node cluster list, the nodes involved in the historical transfer data are classified according to their respective clusters. For instance, the historical transfer data is homomorphically partitioned based on the material detection node cluster in the node cluster list, forming a material detection node historical transfer data cluster. This homomorphic partitioning is repeated in the same way, ultimately forming multiple historical transfer data clusters, ensuring the consistency between the training data and the prediction target.

[0085] Furthermore, using historical migration data clusters from multiple nodes as sample data, multiple migration prediction models are constructed and supervised training is performed. For example, the migration prediction model can be constructed using the following technical path: 1. Data preparation: Extract historical data from the node historical migration data clusters, and extract node attribute features from the historical data, such as load rate and number of personnel in a particular migration, as sample input features; and correspondingly collect historical capacity bottlenecks and historical conversion success rates from the historical data, manually labeling the sample input features with historical capacity bottleneck and historical conversion success rate labels, and dividing them into a training set (fitting the model) and a validation set (adjusting parameters) in a 7:3 ratio. 2. Model Construction: A neural network model architecture can be adopted. The transfer prediction model can use a multi-input dual-output adaptive structure, mainly composed of an input layer, a feature processing layer, a feature fusion layer, and an output layer. The input layer receives entity feature information from the transfer nodes, mainly including numerical features such as directly quantifiable indicators like efficiency (day / order), load rate (%), and success rate (%), as well as categorical features such as service type and coverage area, which require encoding. The input layer performs preliminary separation of the two types of features to prepare for subsequent processing. The feature processing layer transforms discrete categories into low-dimensional dense vectors for categorical features through an embedding layer, capturing the potential correlation between categories. For numerical features, a standardization layer (such as Z-Score normalization) eliminates dimensional differences and ensures balanced feature weights. The processed data is then integrated into a feature vector of uniform dimension through a concatenation layer. The feature fusion layer uses a fully connected network for deep feature extraction, containing two fully connected network layers: the first layer has 32 neurons (ReLU activation function), and the second layer has 16 neurons (ReLU activation function). A Dropout layer is used to further enhance the feature fusion. (Learning rate = 0.2) to suppress overfitting; the output layer adopts a dual-output head structure, corresponding to two prediction targets respectively: the capability bottleneck output head uses linear output (regression result), and the conversion success rate output head uses the sigmoid activation function to output the probability of success (between 0 and 1). 3. Model training: using the attribute features of the node's historical transfer data as input features, and historical capability bottleneck and historical conversion success rate as supervision labels, a weighted joint loss function is adopted, the formula is total loss = 0.6 × bottleneck loss + 0.4 × success rate loss, the weights are set according to business priority, and the bottleneck prediction has a higher weight to reduce risk, using the Adam optimizer (learning rate = 0.001, , The batch size is set to 32, and the maximum number of iterations is 100. Training is stopped when the total loss on the validation set does not decrease for 10 consecutive iterations to avoid overfitting. The validation set metrics are monitored in real time during training. For classification tasks, AUC and F1 score are considered, and for regression tasks, R² and MAE are considered. The best performing model is selected and saved as the final transfer prediction model to obtain the trained transfer prediction model.

[0086] Finally, the entity feature information of multiple transfer nodes is input into multiple transfer prediction models to predict capacity bottlenecks and conversion success rates. The prediction results are then overlaid and fused based on a set of candidate technology transfer paths to obtain the path capacity bottleneck and conversion success rate for each candidate technology transfer path, outputting the transfer prediction result. For example, the entity feature information (such as efficiency, resource reserves, load, etc.) of multiple transfer nodes in a multi-level graph is input into the transfer prediction model corresponding to their respective clusters, predicting the capacity bottleneck and conversion success rate of each transfer node. For instance, a transfer node might have a 2-day delay and a 90% conversion success rate. For example, the prediction results of capacity bottlenecks and conversion success rates are overlaid and fused based on a set of candidate technology transfer paths. For capacity bottleneck prediction, the maximum capacity bottleneck value of all transfer nodes in the path can be taken as the path capacity bottleneck; for conversion success rate prediction, the product of the conversion success rates of all connected transfer nodes can be taken as the path conversion success rate. The final output is the transfer prediction result, which quantifies the potential risk and predicted success rate of each path and serves as the core basis for subsequent dynamic path optimization.

[0087] In summary, compared to existing technologies, this application combines historical transfer data and node attribute information to traverse the multi-level graph to predict capability bottlenecks and conversion success rates, obtaining transfer prediction results. In this way, the complex node relationships in the multi-level graph are transformed into quantifiable path evaluation indicators. This not only utilizes historical data to uncover node behavior patterns but also covers potential possibilities through randomly generated paths. Ultimately, it provides accurate capability bottleneck and conversion success rate assessments for each candidate path, laying a data-driven decision-making foundation for subsequent optimization.

[0088] S50: Based on the transfer prediction results and combined with the dynamic information of the upstream and downstream of the industrial chain, a multi-objective dynamic matching algorithm is used to optimize the path, obtain the target technology transfer path, and perform technology transfer industrial chain management based on the target technology transfer path.

[0089] Information in the upstream and downstream of the industrial chain is constantly changing, such as fluctuations in supply-side resources, adjustments in demand-side priorities, and changes in intermediary service capabilities. Traditional static paths, once determined, remain fixed and cannot respond to real-time changes in node status and environmental conditions. This leads to a gradual disconnect between the path and actual needs, which can easily result in mismatches in node capabilities, decreased transfer efficiency, or even path interruption, reducing the success rate of technology transfer and causing resource investment losses.

[0090] To address the aforementioned issues, this application utilizes a multi-objective dynamic matching algorithm to optimize the path based on the aforementioned transfer prediction results and combined with dynamic information from upstream and downstream of the industry chain, thereby obtaining the target technology transfer path and implementing technology transfer industry chain management based on the target technology transfer path.

[0091] Specifically, step S50 in the method includes:

[0092] Select the top N best candidate technology transfer paths from the transfer prediction results, and update the corresponding upstream and downstream dynamic information of the industry chain in real time.

[0093] Based on the dynamic information of the upstream and downstream of the industrial chain, the capability bottlenecks and conversion success rates of multiple transfer nodes corresponding to the N candidate technology transfer paths are predicted to obtain the first dynamic prediction result.

[0094] Based on the dynamic information of the upstream and downstream of the industrial chain, the capability bottlenecks and conversion success rates of the approximate transfer nodes of multiple transfer nodes corresponding to the N candidate technology transfer paths are predicted to obtain the second dynamic prediction result.

[0095] The optimal transfer node is marked by combining the first dynamic prediction result and the second dynamic prediction result, and the N candidate technology transfer paths are matched and updated according to the optimal transfer node marking result;

[0096] The system iterates and predicts the capability bottlenecks and conversion success rates of the updated N candidate technology transfer paths, and selects the optimal path based on the obtained path capability bottlenecks and conversion success rates, outputting the target technology transfer path.

[0097] In this embodiment, the top N candidate technology transfer paths with the best performance from the transfer prediction results are first selected, and the dynamic information of the upstream and downstream of the industry chain is updated in real time accordingly. For example, the top N candidate technology transfer paths with the best overall performance are selected from the transfer prediction results. For instance, the paths with the top 20% conversion success rate and the bottom 20% of capability bottlenecks are selected to ensure that the starting point of optimization has a high quality. At the same time, dynamic change data of the upstream and downstream of the industry chain are collected, such as resource fluctuations of technology suppliers, priority adjustments of demanders, resource allocation and changes in cooperation relationships of intermediary nodes. This dynamic information directly affects the real-time capabilities of nodes and is a key input for dynamic optimization.

[0098] Secondly, based on dynamic information from the upstream and downstream of the industry chain, the capability bottlenecks and conversion success rates of multiple transfer nodes corresponding to N alternative technology transfer paths are predicted to obtain the first dynamic prediction result. For example,

[0099] Based on updated dynamic information from upstream and downstream of the industry chain, the capability bottlenecks and conversion success rates of multiple transfer nodes corresponding to the top N alternative technology transfer paths are reassessed. For example, if the dynamic information shows that the efficiency of a certain node decreases by 10% due to staff turnover, the capability bottlenecks and conversion success rates are re-predicted through the transfer prediction model. The capability bottleneck is extended from the original 1 day to 2 days, and the conversion success rate is reduced from the original 85% to 70%, ensuring that the evaluation of the original path is synchronized with the real-time status.

[0100] Secondly, based on dynamic information from the upstream and downstream of the industry chain, the capacity bottlenecks and conversion success rates of approximate transfer nodes corresponding to multiple transfer nodes in the N candidate technology transfer paths are predicted to obtain a second dynamic prediction result. For example, for multiple transfer nodes in the candidate technology transfer paths, functionally similar approximate nodes are found, such as other nodes in the same node cluster. For instance, if the original path uses transfer node A1, and transfer node A2 in the same node cluster shows a low load in the dynamic information, the capacity bottlenecks and conversion success rates of transfer node A2 are predicted to obtain a second dynamic prediction result. For example, the capacity bottleneck is 0.5 days, and the conversion success rate is 90%. By introducing approximate replaceable transfer nodes, the limitations of the original path are broken, providing more possibilities for optimization.

[0101] Furthermore, the optimal transfer node is marked by combining the first dynamic prediction result and the second dynamic prediction result, and the N candidate technology transfer paths are matched and updated according to the optimal transfer node marking result. For example, for each transfer node in each candidate technology transfer path, its first dynamic prediction result is compared with the second dynamic prediction result of the corresponding approximate node, and the node with better performance is marked. For example, if the capacity bottleneck of the original transfer node A1 is 2 days and the conversion success rate is 70%, and the capacity bottleneck of the similar transfer node A2 is 0.5 days and the conversion success rate is 90%, then transfer node A2 is marked as the optimal node in this stage, and the corresponding node in the original path is replaced with the marked optimal node to generate the updated candidate path.

[0102] Finally, the updated N candidate technology transfer paths are iteratively predicted for capability bottlenecks and conversion success rates. Based on the obtained path capability bottlenecks and conversion success rates, the optimal path is selected, and the output is the target technology transfer path. For example, for the updated N candidate technology transfer paths, the above process of dynamic information collection → transfer node prediction → transfer node replacement is repeated for multiple iterations, such as 3-5 iterations. Each iteration is based on the latest industry chain dynamic information, recalculating the path capability bottlenecks and conversion success rates to ensure continuous adaptation to changes. After the iteration, the final performance of each path is comprehensively evaluated, and the path that simultaneously satisfies the lowest path capability bottleneck and the highest path conversion success rate is selected as the target technology transfer path.

[0103] In summary, compared to existing technologies, this application, based on the aforementioned transfer prediction results and combined with dynamic information from upstream and downstream of the industry chain, employs a multi-objective dynamic matching algorithm for path optimization to obtain the target technology transfer path, and then performs technology transfer industry chain management based on the target technology transfer path. This allows the technology transfer path to adapt to changes in the industry chain while achieving a multi-objective balance between minimizing capability bottlenecks and maximizing conversion success rates. It avoids the failure issues caused by environmental changes during the execution of static paths, ultimately achieving a dual improvement in technology transfer efficiency and stability.

[0104] In summary, the embodiments of this application have at least the following technical effects:

[0105] Compared to existing technologies, this application first constructs feature vectors for technology suppliers, demanders, and intermediary nodes, obtaining a set of node feature vectors. Thus, by constructing feature vectors and obtaining a set of node feature vectors, a unified data foundation is provided for subsequent path calculation based on directed acyclic graphs and optimization of dynamic matching algorithms, ensuring quantitative analysis and accurate decision-making throughout the entire technology transfer process.

[0106] Secondly, this application uses the node feature vector set and a directed acyclic graph method to obtain the initial transfer path. Thus, by transforming node features into a graphical network and calculating the optimal path, the core flow nodes and hierarchical relationships from supply to demand are clarified, realizing the transformation of technology transfer path from qualitative judgment to quantitative calculation, and providing a structured foundation for subsequent more refined management.

[0107] Furthermore, this application collects real-time information based on the initial transfer path and constructs a multi-level graph containing multiple transfer nodes and their collaborative relationships based on the real-time information collection results. In this way, the initial transfer path is refined into a structured network containing specific execution units and collaborative relationships, providing more realistic underlying data support for subsequent accurate prediction and optimization.

[0108] Furthermore, this application combines historical transfer data with node attribute information to traverse the multi-level graph to predict capability bottlenecks and conversion success rates, obtaining transfer prediction results. In this way, the complex node relationships in the multi-level graph are transformed into quantifiable path evaluation indicators. This not only utilizes historical data to uncover node behavior patterns but also covers potential possibilities through randomly generated paths. Ultimately, it provides accurate capability bottleneck and conversion success rate assessments for each candidate path, laying a data-driven decision-making foundation for subsequent optimization.

[0109] Finally, based on the aforementioned transfer prediction results and combined with dynamic information from upstream and downstream of the industry chain, this application employs a multi-objective dynamic matching algorithm to optimize the path, obtain the target technology transfer path, and execute technology transfer industry chain management according to the target technology transfer path. In this way, the technology transfer path adapts to changes in the industry chain while achieving a multi-objective balance between minimizing capability bottlenecks and maximizing conversion success rates. This avoids the failure issues caused by environmental changes during the execution of static paths, ultimately achieving a dual improvement in technology transfer efficiency and stability.

[0110] Through the aforementioned technical solutions, this application addresses the problem of inefficient matching caused by the ambiguity of main features by quantifying the multidimensional attributes of suppliers, demanders, and intermediary nodes using feature vector quantification technology. It also refines the collaborative relationships between nodes through multi-level graphs, overcoming the limitations of unclear node connections in traditional coarse frameworks. Furthermore, by combining a transfer prediction model with a multi-objective dynamic matching algorithm, it rapidly adjusts paths based on dynamic changes in the upstream and downstream of the industry chain, effectively avoiding the risk of path disconnection under static management. This improves the matching accuracy, process efficiency, and conversion success rate of technology transfer industry chain management, accelerating the transformation of technological achievements into industrial applications.

[0111] Example 2, as Figure 2 As shown, based on the same inventive concept as the technology transfer supply chain management method using a dynamic matching algorithm provided in Embodiment 1, this embodiment of the invention also provides a technology transfer supply chain management system using a dynamic matching algorithm, comprising:

[0112] Vector construction module 11 is used to construct feature vectors of technology suppliers, demanders and intermediary nodes, and obtain node feature vector sets;

[0113] Initial path acquisition module 12 is used to acquire an initial transition path based on the node feature vector set and in combination with the directed acyclic graph method;

[0114] The graph construction module 13 is used to collect real-time information based on the initial transfer path and construct a multi-level graph containing multiple transfer nodes and cooperative relationships based on the real-time information collection results.

[0115] The transfer prediction module 14 is used to combine historical transfer data and node attribute information to traverse the multi-level map to predict capacity bottlenecks and conversion success rates, and obtain transfer prediction results.

[0116] The optimized output module 15 is used to optimize the path based on the transfer prediction results and the dynamic information of the upstream and downstream of the industrial chain, using a multi-objective dynamic matching algorithm to obtain the target technology transfer path, and to perform technology transfer industrial chain management based on the target technology transfer path.

[0117] Specifically, the vector construction module 11 is used for:

[0118] The vector dimensions of the feature vector include technical attributes, subject attributes, and service attributes;

[0119] The technical attributes include at least the technology type, technology maturity, technical indicators, technological innovation points, and technological relevance;

[0120] The subject attributes include at least the subject size, subject financial status, subject credit rating, and subject cooperation history;

[0121] The service attributes include at least the service scope, service success rate, service efficiency, and service resources.

[0122] Furthermore, the phrase "constructing feature vectors for technology suppliers, demanders, and intermediary nodes, and obtaining a set of node feature vectors" includes:

[0123] Based on the supply capacity data of technology suppliers, we analyze and extract the technical attributes, subject attributes, and service attributes of technology supply nodes to form a feature vector of technology supply nodes.

[0124] Based on the demand data from the demand side, analyze and extract the technical attributes, subject attributes, and service attributes of the demand nodes to form a feature vector of the demand nodes;

[0125] The intermediary nodes are clustered to obtain a set of intermediary node categories. Based on big data, statistical analysis is performed on the set of intermediary node categories to extract typical capability parameters of each intermediary node category and generate an intermediary node feature vector.

[0126] The feature vectors of the technology supply nodes, the feature vectors of the demand nodes, and the feature vectors of the intermediary nodes are summarized to form the node feature vector set.

[0127] The initial path acquisition module 12 is specifically used for:

[0128] Based on the node feature vector set, a directed acyclic graph is constructed with the technology supply node as the starting point, multiple intermediary nodes as intermediate nodes, and the demand node as the ending point. The edges of the directed acyclic graph are unidirectional edges with the direction pointing from the technology supply node to the demand node.

[0129] In the directed acyclic graph, based on the principle of minimizing edge weight, the shortest path algorithm is used to calculate the path from the technology supply node to the demand node through at least one of the intermediary nodes to obtain the initial transfer path;

[0130] The initial transfer path is used to define the hierarchical path of technology transfer.

[0131] Specifically, the map construction module 13 is used for:

[0132] Based on the initial transfer path, extract multiple transfer nodes corresponding to each intermediary node and generate a list;

[0133] Information is retrieved from the list, and then regularized and standardized to obtain the entity feature information of the transfer node;

[0134] Based on the entity feature information, the initial transfer path is expanded accordingly to generate a multi-level graph. Each level of the multi-level graph corresponds to the technology supply node, the demand node, and multiple intermediary nodes. Each intermediary node has multiple transfer nodes under it. Each transfer node represents its capabilities and collaborative relationships in graph form.

[0135] Specifically, the transfer prediction module 14 is used for:

[0136] Based on the multi-level graph, multiple levels are traversed from top to bottom to randomly select multiple transfer nodes to obtain alternative technology transfer paths, and a set of alternative technology transfer paths is obtained iteratively.

[0137] Clustering is performed based on the node attribute information of the multiple transfer nodes obtained, and the clustering results are output as a list of multiple node clusters;

[0138] Based on the node cluster list, the historical transfer data is homomorphically partitioned to form multiple node historical transfer data clusters;

[0139] Using the historical transfer data clusters of the aforementioned nodes as sample data, multiple transfer prediction models are constructed and supervised training is performed.

[0140] The entity feature information of multiple transfer nodes is input into multiple transfer prediction models to predict the capacity bottleneck and conversion success rate. The prediction results of capacity bottleneck and conversion success rate are superimposed and fused according to the set of alternative technology transfer paths to obtain the path capacity bottleneck and path conversion success rate of each alternative technology transfer path, and the output is the transfer prediction result.

[0141] The optimized output module 15 is specifically used for:

[0142] Select the top N best candidate technology transfer paths from the transfer prediction results, and update the corresponding upstream and downstream dynamic information of the industry chain in real time.

[0143] Based on the dynamic information of the upstream and downstream of the industrial chain, the capability bottlenecks and conversion success rates of multiple transfer nodes corresponding to the N candidate technology transfer paths are predicted to obtain the first dynamic prediction result.

[0144] Based on the dynamic information of the upstream and downstream of the industrial chain, the capability bottlenecks and conversion success rates of the approximate transfer nodes of multiple transfer nodes corresponding to the N candidate technology transfer paths are predicted to obtain the second dynamic prediction result.

[0145] The optimal transfer node is marked by combining the first dynamic prediction result and the second dynamic prediction result, and the N candidate technology transfer paths are matched and updated according to the optimal transfer node marking result;

[0146] The system iterates and predicts the capability bottlenecks and conversion success rates of the updated N candidate technology transfer paths, and selects the optimal path based on the obtained path capability bottlenecks and conversion success rates, outputting the target technology transfer path.

[0147] In summary, the embodiments of this application have at least the following technical effects:

[0148] Compared to existing technologies, this application first constructs feature vectors for technology suppliers, demanders, and intermediary nodes through a vector construction module, obtaining a set of node feature vectors. This provides a unified data foundation for subsequent path calculation based on directed acyclic graphs (DAGs) and optimization of dynamic matching algorithms, ensuring quantitative analysis and accurate decision-making throughout the technology transfer process. Secondly, through an initial path acquisition module, based on the node feature vector set and combined with DAG methods, the initial transfer path is obtained, clarifying the core flow nodes and hierarchical relationships from supply to demand. This transforms the technology transfer path from qualitative judgment to quantitative calculation, providing a structured foundation for more refined management. Thirdly, through a graph construction module, real-time information is collected based on the initial transfer path, and a multi-level graph containing multiple transfer nodes and collaborative relationships is constructed based on the real-time information collection results. This refines the initial transfer path into a structured network containing specific execution units and collaborative relationships, providing more realistic underlying data support for subsequent accurate prediction and optimization. Furthermore, through the transfer prediction module, combining historical transfer data and node attribute information, multi-level graphs are traversed to predict capacity bottlenecks and conversion success rates, obtaining transfer prediction results. This provides accurate capacity bottleneck and conversion success rate assessments for each candidate path, laying a data-driven decision-making foundation for subsequent optimization. Finally, through the optimization output module, based on the transfer prediction results and combined with dynamic information from upstream and downstream of the industry chain, a multi-objective dynamic matching algorithm is used to optimize the path, obtaining the target technology transfer path. Technology transfer industry chain management is then executed based on the target technology transfer path, enabling the technology transfer path to adapt to changes in the industry chain while achieving a multi-objective balance between minimizing capacity bottlenecks and maximizing conversion success rates. This avoids the failure issues caused by environmental changes during the execution of static paths, ultimately achieving a dual improvement in technology transfer efficiency and stability. In this way, the matching accuracy, process efficiency, and conversion success rate of technology transfer industry chain management are improved, accelerating the transformation of technological achievements into industrial applications.

[0149] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0150] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied 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.

[0151] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0152] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0153] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0154] Although preferred embodiments of the invention have been described, those skilled in the art, once they have learned the basic inventive concept, can make other changes and modifications to these embodiments.

[0155] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of this invention and its equivalents, this invention also intends to include these modifications and variations.

Claims

1. A technology transfer industry chain management method using a dynamic matching algorithm, characterized in that, The method comprises the following steps: Constructing feature vectors of technology supply nodes, demand nodes and intermediary nodes, and obtaining a set of node feature vectors; Constructing a directed acyclic graph with the technology supply nodes as the starting point, the intermediary nodes as the intermediate nodes and the demand nodes as the terminal point according to the set of node feature vectors, wherein the edges of the directed acyclic graph are unidirectional and point from the technology supply nodes to the demand nodes; Calculating a path from the technology supply nodes to the demand nodes through at least one intermediary node in the directed acyclic graph by using a shortest path algorithm with the minimum edge weight as the principle, and obtaining an initial transfer path, wherein the initial transfer path is used to define a hierarchical path of technology transfer; Collecting real-time information according to the initial transfer path, and constructing a multi-level graph containing multiple transfer nodes and collaborative relationships according to the real-time information collection results; Predicting the capacity bottleneck and conversion success rate by traversing the multi-level graph in combination with historical transfer data and node attribute information, and obtaining a transfer prediction result; Optimizing the path by using a multi-objective dynamic matching algorithm in combination with upstream and downstream dynamic information of the industrial chain according to the transfer prediction result, obtaining a target technology transfer path, and performing technology transfer industrial chain management according to the target technology transfer path; The method of collecting real-time information according to the initial transfer path and constructing a multi-level graph containing multiple transfer nodes and collaborative relationships according to the real-time information collection results comprises the following steps: Extracting multiple transfer nodes corresponding to each intermediary node based on the initial transfer path, and generating a list; Calling information according to the list, and normalizing and standardizing to obtain entity feature information of the transfer nodes; According to the entity feature information, the initial transfer path is correspondingly expanded to generate a multi-level graph, each level of the multi-level graph corresponds to the technology supply nodes, the demand nodes and multiple intermediary nodes, and each intermediary node is in charge of multiple transfer nodes, and each transfer node represents the transfer node capacity and collaborative relationship in the form of a graph; The method of optimizing the path by using a multi-objective dynamic matching algorithm in combination with upstream and downstream dynamic information of the industrial chain according to the transfer prediction result to obtain a target technology transfer path comprises the following steps: Selecting the top N optimal candidate technology transfer paths in the transfer prediction result, and correspondingly updating the upstream and downstream dynamic information of the industrial chain in real time; According to the upstream and downstream dynamic information of the industrial chain, predicting the capacity bottleneck and conversion success rate of multiple transfer nodes corresponding to the N candidate technology transfer paths to obtain a first dynamic prediction result; According to the upstream and downstream dynamic information of the industrial chain, predicting the capacity bottleneck and conversion success rate of the approximate transfer nodes of multiple transfer nodes corresponding to the N candidate technology transfer paths to obtain a second dynamic prediction result; Combining the first dynamic prediction result and the second dynamic prediction result to mark the optimal transfer nodes, and updating the N candidate technology transfer paths according to the optimal transfer node marking result. The updated N alternative technology transfer paths are iterated to obtain the path capability bottleneck and the conversion success rate, and the optimal path is selected according to the obtained path capability bottleneck and the path conversion success rate, and the output is the target technology transfer path.

2. The technology transfer industry chain management method employing a dynamic matching algorithm according to claim 1, characterized in that, The vector dimension of the feature vector includes technical attributes, subject attributes and service attributes; The technical attributes at least include technical type, technical maturity, technical index, technical innovation point and technical correlation; The subject attributes at least include subject scale, subject financial condition, subject credit degree and subject cooperation history; The service attributes at least include service range, service success rate, service efficiency and service resource.

3. The technology transfer industry chain management method employing a dynamic matching algorithm according to claim 1, characterized in that, The feature vectors of the technology supply side, the demand side and the intermediary node are constructed to obtain a node feature vector set, including: Based on the supply capacity data of the technology supply side, the technical attributes, the subject attributes and the service attributes of the technology supply node are analyzed and extracted to form a technology supply node feature vector; Based on the demand data of the demand side, the technical attributes, the subject attributes and the service attributes of the demand node are analyzed and extracted to form a demand node feature vector; The intermediary nodes are clustered and divided to obtain a set of intermediary node categories, and the typical capability parameters of each intermediary node category are extracted based on big data statistical analysis of the set of intermediary node categories to generate an intermediary node feature vector; The technology supply node feature vector, the demand node feature vector and the intermediary node feature vector are summarized to form the node feature vector set.

4. The technology transfer industry chain management method employing a dynamic matching algorithm according to claim 3, characterized in that, Based on the multi-level graph, the multi-level graph is iterated from top to bottom to perform random selection of multiple transfer nodes, obtain an alternative technology transfer path, and iteratively obtain a set of alternative technology transfer paths; According to the node attribute information of the obtained multiple transfer nodes, the nodes are clustered and divided to output a cluster list of multiple nodes as a cluster division result; The historical transfer data is homomorphically divided according to the node cluster list to form multiple node historical transfer data clusters; Multiple transfer prediction models are constructed respectively using multiple node historical transfer data clusters as sample data and are supervised trained; The entity feature information of multiple transfer nodes is input into multiple transfer prediction models respectively to predict the capability bottleneck and the conversion success rate, and the capability bottleneck and the conversion success rate prediction results are superimposed and fused according to the set of alternative technology transfer paths to obtain the path capability bottleneck and the path conversion success rate of each alternative technology transfer path, and the output is the transfer prediction result. The method for performing any one of claims 1-4, comprising:

5. A technology transfer industry chain management system employing a dynamic matching algorithm, characterized by, A vector construction module for constructing feature vectors of technology supply side, demand side and intermediary nodes to obtain a node feature vector set; An initial path acquisition module for obtaining an initial transfer path based on the node feature vector set in combination with a directed acyclic graph method; ​ A graph construction module is configured to collect real-time information according to the initial transfer path, and construct a multi-level graph including multiple transfer nodes and cooperative relationships according to the real-time information collection result; A transfer prediction module is configured to combine historical transfer data and node attribute information, traverse the multi-level graph to predict the capacity bottleneck and the conversion success rate, and obtain a transfer prediction result; An optimization output module is configured to combine upstream and downstream dynamic information of the industrial chain according to the transfer prediction result, use a multi-objective dynamic matching algorithm to optimize the path, obtain a target technology transfer path, and perform technology transfer industrial chain management according to the target technology transfer path.

Images