Scientific and technological achievement conversion-based supply and demand intelligent recommendation matching system

Abstract

The invention discloses an intelligent supply and demand recommendation matching system based on scientific and technological achievement conversion, and relates to the technical field of resource services, the system comprises an information input module, an information management module, a supply and demand matching module, a pricing prediction module and a conversion recommendation module, user input information is subjected to distributed storage and classified labeling, and a knowledge graph is constructed; after preliminary supply and demand matching is carried out, semantic path reasoning is used for calculating the matching degree and sorting is carried out, a reinforcement learning model is adopted, the transaction price is dynamically predicted based on the transaction frequency, the industry permeability and the technical life cycle, based on a multi-factor evaluation algorithm, the matching degree, the transaction price satisfaction degree and the historical transaction rate are integrated, and a personalized recommendation list is generated; the problems that in the existing scientific and technological achievement conversion process, information of a recommendation matching system is scattered, opaque and single in evaluation dimension, and deep semantic association and user demand diversity are neglected are solved, and accurate and efficient matching services are provided for a scientific and technological achievement provider and a technical demander.

Description

Technical Field

[0001] The present invention relates to the field of transformation of scientific and technological achievements, and in particular to an intelligent recommendation and matching system based on supply and demand of scientific and technological achievements transformation. Background Art

[0002] With the rapid advancement of science and technology, various scientific research results continue to emerge. The transformation of scientific and technological achievements is the process of converting scientific research results into practical applications. This not only promotes economic and technological development and progress, but also promotes innovation and progress in all areas of society. However, the effectiveness and efficiency of scientific and technological achievement transformation are subject to multiple factors, including market demand, technology supply, and industrialization processes. The current scientific and technological achievement transformation process is plagued by problems such as information asymmetry between supply and demand, low matching efficiency, and low transformation success rates. As a result, most scientific and technological achievements often remain in the laboratory stage and fail to be effectively transformed into products and services needed by society. Therefore, how to efficiently and accurately match scientific and technological achievements with social needs has become a key task in accelerating achievement transformation and innovative development.

[0003] During the transformation of scientific and technological achievements, information between technology providers and technology demanders is fragmented and opaque, making it difficult for both parties to quickly find suitable partners. The matching process is time-consuming and inefficient. Traditional matching methods rely on keywords or simple rules, failing to capture the deep semantic connections between scientific achievements and technology needs, resulting in poorly relevant matching results. Existing recommendation matching systems typically assess conversion potential based on a single dimension, such as price or matching degree, while ignoring other important factors, such as historical transaction rates and user satisfaction. They also overlook the diversity of user needs, such as price sensitivity and historical conversion preferences. This leads to one-sided evaluations and poor recommendation effectiveness.

[0004] Therefore, an intelligent recommendation and matching system based on the supply and demand of scientific and technological achievements transformation is needed to solve the above problems. Summary of the Invention

[0005] In response to the shortcomings of the existing technology, the present invention discloses an intelligent recommendation and matching system based on the supply and demand of scientific and technological achievements transformation. Through AI-driven precise matching, dynamic pricing and full-process services, it provides accurate and efficient matching services for scientific and technological achievement providers and technology demanders.

[0006] The present invention adopts the following technical solutions: An intelligent recommendation and matching system for supply and demand based on the transformation of scientific and technological achievements, including: Information entry module: users register, select identities and log in through the information authentication interface. After logging in, users enter the corresponding information entry interface according to their identities to enter scientific and technological achievements information and technical requirements information. User identities include scientific and technological achievements providers and technical requirements parties. The information management module stores the input scientific and technological achievements and technical requirements information in a distributed manner in the scientific and technological achievements information database and the technical requirements information database, and labels each scientific and technological achievement and technical requirement with multi-dimensional tags. Through the tags of scientific and technological achievements, technical requirements, information about scientific and technological achievement providers and information about technical requirements, semantic association mining is carried out to build a knowledge graph of scientific and technological achievements and technical requirements. The supply and demand matching module uses preset labeling rules to preliminarily match scientific and technological achievement labels with technical requirement labels, generating many-to-many preliminary matching pairs of scientific and technological achievements and technical requirements. For the generated preliminary matching pairs, the semantic path reasoning method in the knowledge graph is used to calculate the semantic matching degree, and the preliminary matching pairs are prioritized according to the calculated semantic matching degree. The pricing prediction module uses a reinforcement learning model to predict the transaction price of scientific and technological achievements. The reinforcement learning model dynamically sets prices based on the transaction frequency, industry penetration rate, and technology life cycle stage of scientific and technological achievements with the same label. The conversion recommendation module uses a multi-factor conversion evaluation algorithm to generate a personalized conversion recommendation list. The multi-factor conversion evaluation algorithm performs a comprehensive evaluation based on the semantic matching degree between scientific and technological achievement information and technical demand information, the satisfaction with the predicted transaction price, and the historical transaction rate of the same label technology.

[0007] Furthermore, the user opens the information authentication interface on the electronic terminal, selects the user identity as the technology achievement provider or the technology demander, and enters the user name, contact information and login password information. The information entry module verifies the validity of the input information. After the authentication is passed, the user logs in using the account and password set during registration, SMS verification code or third-party login method. After the user logs in, the corresponding information entry interface is entered according to the user identity. The user enters the information content according to the format requirements. The technology achievement information entered by the technology achievement provider includes at least the name of the technology achievement, technical field, technology achievement description, technical background, application direction and technology achievement life cycle stage. The technology demand information entered by the technology demander includes at least the name of the technology demand, technical field, technology demand description, technical background, application direction and technology demand problem. When the user performs an operation, real-time verification is performed through role identification and permission list.

[0008] Furthermore, the scientific and technological achievement information database performs multi-dimensional classification based on the entered scientific and technological achievement information, and adds multi-dimensional labels to each scientific and technological achievement. The technical demand information database performs multi-dimensional classification based on the entered technical demand information, and adds multi-dimensional labels to each technical demand. The scientific and technological achievement information database and the technical demand information database use natural language models to extract keywords and perform semantic representation on the entered scientific and technological achievements and technical demand texts, and extract entities, relationships and attributes from the entered scientific and technological achievement information and technical demand information as well as the registered scientific and technological achievement provider information and technical demand party information to construct a knowledge graph.

[0009] Furthermore, the natural language model uses a pre-trained language model to encode the entered scientific and technological achievements information and the entered technical requirements information, converts the unstructured description into a vector representation, and extracts keywords through the attention mechanism. Then, the vector representation generated by the transformation of the technical field, technical background and application direction, as well as the keywords extracted from the scientific and technological achievements description and the technical requirements description, are calculated through cosine similarity to complete multi-dimensional classification.

[0010] Furthermore, the knowledge graph uses a graph structure to store the extracted entities, relationships and attributes. The nodes of the knowledge graph are scientific and technological achievements, technical requirements, providers of scientific and technological achievements and technical demanders. The edges of the knowledge graph are the relationship between scientific and technological achievements and technical requirements, the relationship between scientific and technological achievement information and providers of scientific and technological achievements, the relationship between technical demand information and technical demanders, the relationship between scientific and technological achievement information and technical demanders, and the relationship between technical demand information and providers of scientific and technological achievements. The knowledge graph crawls incremental information in real time through a dynamic update mechanism.

[0011] Furthermore, in the supply and demand matching module, a graph algorithm is used to find the path between scientific and technological achievement information and technical demand information in the knowledge graph, and a graph neural network GNN is used to map the semantic path into a low-dimensional vector to capture the implicit semantics of entities and relationships in the path. In the low-dimensional vector space, the similarity between node vectors is calculated by cosine similarity to evaluate the semantic relevance of the path, and the preliminary matching pairs are prioritized in descending order of similarity scores.

[0012] Furthermore, the working method of the reinforcement learning model includes the following steps: S1. Use the tag information, transaction frequency, industry penetration rate, and technology life cycle stage of scientific and technological achievements as state variables to describe the market performance and demand of current scientific and technological achievements in different dimensions; S2. Select different pricing actions based on the current status. Pricing actions include raising the price, lowering the price, and keeping the price unchanged. S3. Use a reward function to evaluate the effect of each pricing action and adjust the degree of reward based on the transaction volume of scientific and technological achievements; S4. Use deep Q network for iterative learning and update the pricing strategy based on the reward signal to obtain the optimal pricing strategy to maximize the transaction price of scientific and technological achievements or maximize profits.

[0013] Furthermore, the multi-factor conversion evaluation algorithm uses the semantic matching degree between scientific and technological achievement information and technical demand information, the satisfaction degree of predicted transaction price and the historical transaction rate of the same-label technology as evaluation factors. The satisfaction degree of predicted transaction price is obtained by calculating the similarity between the expected price of the scientific and technological achievement provider or the technology demander and the predicted transaction price. The historical transaction rate of the same-label technology is obtained by calculating the number of successful conversions of the same-label technology and the number of recommendations of the same-label technology. The evaluation factors are mapped to the interval [0,1] to eliminate dimensional differences, and the hierarchical analysis method is used to determine the weight of each factor. Each factor is weighted to obtain a comprehensive conversion evaluation score for each preliminary matching pair.

[0014] The beneficial effects of the present invention are: 1. This invention uses distributed storage technology to store scientific and technological achievement information and technical demand information separately in a dedicated database, and labels each scientific and technological achievement and technical demand with multi-dimensional tags. This not only improves the efficiency of information management, but also provides a rich data foundation for subsequent supply and demand matching. By mining the semantic associations between scientific and technological achievement labels, technical demand labels, information about scientific and technological achievement providers, and information about technical demanders, a knowledge graph of scientific and technological achievements and technical demands is constructed. The knowledge graph can more comprehensively reflect the complex relationship between scientific and technological achievements and technical demands, providing strong support for subsequent accurate matching.

[0015] 2. This invention uses preset tag rules for preliminary matching, generating many-to-many preliminary matching pairs of scientific and technological achievements and technical requirements. It then uses semantic path reasoning within the knowledge graph to calculate semantic matching. Combining tag matching with semantic analysis can more accurately identify potential connections between scientific and technological achievements and technical requirements. Preliminary matching pairs are prioritized based on the calculated semantic matching results, ensuring that users prioritize the scientific and technological achievements or technical requirements that best meet their needs, thereby improving matching efficiency and user satisfaction.

[0016] 3. This invention uses a reinforcement learning model to predict the transaction price of scientific and technological achievements. This model dynamically sets prices based on the transaction frequency, industry penetration, and technology lifecycle stage of similarly labeled scientific and technological achievements. This pricing approach more accurately reflects the market value of scientific and technological achievements, while also taking into account market dynamics and technological development trends, providing a reasonable price reference for the commercialization of scientific and technological achievements.

[0017] 4. This invention utilizes a multi-factor conversion evaluation algorithm to generate personalized conversion recommendation lists. This algorithm comprehensively considers the semantic match between scientific and technological achievement information and technical requirements, predicted transaction price satisfaction, and the historical transaction rate of similarly tagged technologies. This comprehensive evaluation method enables a more comprehensive assessment of the conversion potential of scientific and technological achievements, providing users with more accurate conversion recommendations. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a schematic diagram of the overall structure of the present invention; Figure 2 This is the overall flow chart of the reinforcement learning model in the present invention. DETAILED DESCRIPTION

[0019] The following is a combination of the embodiments of the present invention Figure 1 To the attached Figure 2 The technical solutions in the embodiments of the present invention are clearly and completely described. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0020] The embodiment of the present invention discloses an intelligent recommendation matching system for supply and demand based on the transformation of scientific and technological achievements. This system provides more efficient docking between supply and demand parties through intelligent matching of scientific and technological achievements and technical requirements, thereby promoting the transformation and application of scientific and technological achievements. Figure 1 , including: information entry module, information management module, supply and demand matching module, pricing prediction module and conversion recommendation module. Each module combines big data, artificial intelligence and reinforcement learning technology to ensure the system's intelligence, efficiency and accuracy.

[0021] In the information entry module, users register, select their identities, and log in through the information authentication interface. After logging in, users enter the corresponding information entry interface based on their identities to enter information about scientific and technological achievements and technical requirements. User identities include scientific and technological achievement providers and technical demanders. Users with different identities will enter different entry interfaces. Users enter relevant information based on their identities: The party requiring technical support shall fill in the name of the technical requirement, technical field, description of the technical requirement, technical background, application direction, and technical difficulty.

[0022] In the information management module, the entered scientific and technological achievement information and technical demand information are distributedly stored in the scientific and technological achievement information database and the technical demand information database. This method can ensure efficient access and management of data. Each scientific and technological achievement and technical demand is annotated with multi-dimensional labels, such as field, technology type, application stage, innovation, etc. The generation of labels is based on natural language processing (NLP) technology to ensure the accuracy and comprehensiveness of the labels. Based on label information, provider information, and demander information, semantic association mining is carried out to construct a knowledge graph of scientific and technological achievements and technical demands. Through graph analysis, the potential relationship and matching opportunities between scientific and technological achievements and demands are revealed. Natural language processing and deep learning technologies are used to achieve semantic understanding and graph construction of scientific and technological achievements and technical demands, enabling the system to process complex domain terms and multi-dimensional information and achieve accurate supply and demand matching.

[0023] In the supply-demand matching module, the system uses pre-set labeling rules to perform a preliminary match between scientific and technological achievement tags and technical requirement tags, generating many-to-many preliminary matching pairs. These matching pairs represent potential docking points between the supply and demand sides. Using semantic path reasoning within the knowledge graph, the semantic matching degree of these preliminary matching pairs is calculated. Semantic path reasoning enables in-depth analysis of the semantic similarities between scientific and technological achievements and technical requirements, enabling precise matching. Based on the calculated matching degree, preliminary matching pairs are prioritized, with highly matched pairs being recommended first.

[0024] The pricing prediction module uses a reinforcement learning model to predict the transaction price of scientific and technological achievements. Compared with traditional pricing models, it can adaptively adjust according to market changes, improving pricing flexibility and accuracy. The model dynamically determines pricing based on multiple factors, including: label matching (label similarity with the target technology); industry penetration (frequency and recognition of the scientific and technological achievement in the industry); technology life cycle stage (the stage of technological development of the scientific and technological achievement, for example, basic research, application development, market promotion, etc.); and dynamic adjustment: the reinforcement learning model dynamically adjusts pricing based on market changes, improving price prediction accuracy.

[0025] The conversion recommendation module uses a multi-factor conversion evaluation algorithm that combines multiple factors to generate a personalized conversion recommendation list. The multi-factor evaluation algorithm of the conversion recommendation module takes into account more comprehensive factors, enabling the system to conduct a comprehensive evaluation based on the characteristics of scientific and technological achievements and the specific needs of the demanders, and optimize the conversion effect. The following factors are mainly considered: semantic matching: the degree of semantic matching between scientific and technological achievements and technical requirements; transaction price prediction: the matching degree between the predicted transaction price and the budget of the technology demander; historical transaction rate: the historical transaction rate of the same label technology in the market.

[0026] Based on the comprehensive evaluation results, we provide highly targeted matching of scientific and technological achievements and technical needs that may lead to transactions, helping both supply and demand sides to quickly find the best cooperation opportunities.

[0027] After the information entry module enters data through user interaction, the data is passed to the information management module for storage and labeling. The information management module passes the data and labels to the supply and demand matching module for preliminary matching. The supply and demand matching module passes the preliminary matching results to the pricing prediction module for transaction price prediction, combining them with the inference results of the knowledge graph. The conversion recommendation module evaluates and generates personalized recommendations based on multiple factors, such as matching degree and transaction price prediction.

[0028] The system flow includes the following steps: User registration and login: After the user passes the identity authentication, he / she will enter the corresponding scientific and technological achievements or technical requirements entry interface.

[0029] Information entry and label generation: After the user enters the information, the system automatically generates multi-dimensional labels for scientific and technological achievements and technical requirements.

[0030] Supply and demand matching: The system generates preliminary matching pairs based on labeling rules and knowledge graph analysis, and performs semantic matching calculations and sorting.

[0031] Pricing prediction and conversion recommendation: Predict transaction prices through reinforcement learning models, and generate personalized conversion recommendation lists through multi-factor conversion evaluation.

[0032] In summary, the system, based on the matching of multi-dimensional tags with the knowledge graph, can efficiently and accurately match technology providers with technology demanders. The reinforcement learning model makes the pricing process dynamic, allowing it to adjust to real-time market changes and avoid the limitations of fixed pricing. The multi-factor conversion evaluation algorithm provides personalized recommendations for different demanders, improving conversion rates. Accurate matching and an efficient recommendation mechanism can significantly improve the efficiency of technology transfer and provide strong support for connecting technological innovation with industry needs.

[0033] When a user registers, selects an identity, and logs in through the information authentication interface, the user opens the system's information authentication interface on an electronic terminal (such as a computer, mobile phone, etc.). The user selects the user identity as "scientific and technological achievement provider" or "technology demander" according to his or her own situation, and enters the registration information: the user enters the user name, contact information (such as mobile phone number, email address, etc.) and login password information.

[0034] The information entry module verifies the validity of the information entered by the user, including but not limited to: whether the user name complies with naming rules (such as length and character type); whether the contact information is valid (such as whether the mobile phone number format is correct and the email address exists); and whether the login password meets security requirements (such as length and complexity). If the information entered by the user passes the validity verification, the authentication is successful.

[0035] Users can log in using one of the following methods: using the account and password set during registration; logging in via SMS verification code (the system sends a verification code to the user's registered mobile phone number, and the user enters the verification code to log in); logging in using a third-party login method (such as WeChat, Alipay, etc.).

[0036] After the user logs in, the system will enter the corresponding information entry interface based on the user's identity (scientific and technological achievement provider or technology demander).

[0037] The information entered by the technology achievement provider includes but is not limited to: Name of scientific and technological achievements, which accurately describes the name of the scientific and technological achievements; Technical field: select or enter the technical field to which the scientific and technological achievements belong; Description of scientific and technological achievements, including a detailed description of the content, characteristics, advantages, etc. of the scientific and technological achievements; Technical background, introducing the technical background and R&D process of scientific and technological achievements; Application direction, describing the application fields and potential markets of scientific and technological achievements; The stage of the life cycle of scientific and technological achievements: select the stage of the life cycle of the scientific and technological achievements (such as the research and development stage, application and promotion stage, etc.).

[0038] The information entered by the technical demander includes but is not limited to: Technical requirement name, which accurately describes the technical requirement; Technical field: select or enter the technical field to which the technical requirement belongs; Technical requirements description, which describes in detail the content, requirements, and expectations of the technical requirements; Technical background, which introduces the technical background and causes of technical requirements; Application direction, describing the application areas and expected effects of the technical requirements; Technical requirements difficulties, explain the main difficulties or challenges encountered in technical requirements.

[0039] When a user performs an operation, the system performs real-time verification using role identification and permission lists, ensuring that the user can only access and operate information and functions within their scope of authority. The system assigns each user a unique role identification to identify their identity and permissions. Based on the user's role identification, the system loads the corresponding permission list, clearly defining the information and functions the user can access and operate. Whenever a user performs any operation, the system performs real-time verification against the permission list to ensure that the user's operation is within their scope of authority.

[0040] Through the above processes and mechanisms, the intelligent recommendation and matching system for supply and demand of scientific and technological achievements transformation can ensure the accuracy and security of user information, while providing convenient and efficient information entry and matching recommendation services.

[0041] In the information management module, the scientific and technological achievement information database classifies scientific and technological achievement information in multiple dimensions based on their characteristics, such as technical field, application direction, and lifecycle stage. This classification helps users quickly locate scientific and technological achievements of interest. Each scientific and technological achievement is tagged with multiple dimensions, including technical keywords, application scenarios, innovation points, and maturity. This labeling enriches and accurately describes scientific and technological achievements, facilitating subsequent matching and recommendation.

[0042] Similar to the S&T achievement database, the Technology Demand Information Database also categorizes technology needs across multiple dimensions, such as technical field, type of need, and degree of urgency. Each technology need is tagged with multiple dimensions, such as technical difficulty, desired solution, and budget range. These tags help technology demanders more clearly express their needs and facilitate S&T achievement providers to quickly understand and respond to them.

[0043] The Scientific and Technological Achievement Information Database and the Technical Requirements Information Database use natural language models to extract keywords from the input text of scientific and technological achievements and technical requirements. These keywords summarize the main content of the text and provide an important basis for subsequent classification, labeling, and matching. Natural language models are also used to semantically represent text, converting it into a vector form that computers can understand. This representation helps capture semantic relationships between texts and improve matching accuracy.

[0044] During the knowledge graph construction process, entities are extracted from the input scientific and technological achievement and technical requirement information, as well as the registered information about the provider and demander of the scientific and technological achievement. These entities may include the name of the scientific and technological achievement, the name of the technical requirement, the technical field, the name of the provider, the name of the demander, etc. Relationships between entities are extracted, such as the relationship between scientific and technological achievements and application directions, the relationship between technical requirements and desired solutions, and the relationship between providers and scientific and technological achievements. These relationships form the skeleton of the knowledge graph. Entity attributes are extracted, such as the maturity of the scientific and technological achievement and the budget range of the technical requirement. These attributes provide rich information to the knowledge graph, making it more complete and accurate.

[0045] Based on the extracted entities, relationships, and attributes, a knowledge graph of scientific and technological achievements and technical requirements is constructed. The knowledge graph graphically displays scientific and technological achievements, technical requirements, and the relationships between them, providing users with an intuitive and comprehensive view of information.

[0046] Through the above-mentioned steps of multi-dimensional classification, labeling, natural language processing, and knowledge graph construction, the scientific and technological achievement information database and the technical demand information database can provide users with more accurate and efficient information matching and recommendation services, and promote the transformation and application of scientific and technological achievements.

[0047] The natural language model uses pre-trained language models (such as BERT, the GPT series, and RoBERTa) to encode the input scientific and technological achievements and technical requirements. These models are pre-trained on large-scale text data and can capture the semantic and grammatical information of the text. The pre-trained language models convert unstructured descriptions of scientific and technological achievements and technical requirements into vector representations. These vectors capture the semantic features of the text, enabling computers to understand and process the text.

[0048] Based on the pre-trained language model, the text is further processed using an attention mechanism to extract keywords. This mechanism automatically focuses on important parts of the text and ignores irrelevant information, thereby extracting the most representative keywords. Through this mechanism, the model can extract key technical terms, application scenarios, and innovative points from descriptions of scientific and technological achievements and technical requirements. These keywords provide an important basis for subsequent multi-dimensional classification and matching.

[0049] Structured or semi-structured information, such as technical fields, technical backgrounds, and application directions, can also be converted into vector representations using pre-trained language models or other encoding methods. These vectors represent the semantic characteristics of this information. Cosine similarity is calculated between keyword vectors extracted from descriptions of scientific and technological achievements and technical requirements and the vector representations generated by converting them into technical fields, technical backgrounds, and application directions. Cosine similarity is a metric that measures the similarity between two vectors, with values ​​ranging from -1 to 1. A larger value indicates greater similarity between the two vectors. By calculating cosine similarity, the degree of match between scientific and technological achievements or technical requirements in terms of technical fields, technical backgrounds, and application directions can be assessed.

[0050] Based on the cosine similarity calculation results, thresholds can be set or other classification algorithms can be used to assign scientific and technological achievements and technical requirements to corresponding categories. These categories can include different technical fields, application directions, or technical backgrounds. By comprehensively considering multiple dimensions such as technical fields, technical backgrounds, application directions, and keyword extraction results, a multi-dimensional classification of scientific and technological achievement information and technical requirement information is completed. This classification method helps users more comprehensively understand the characteristics of scientific and technological achievements and technical requirements, improving the accuracy and efficiency of matching.

[0051] Through the above process, the natural language model can effectively transform unstructured information about scientific and technological achievements and technical requirements into structured vector representations and keywords, and perform multi-dimensional classification through similarity calculations. This provides strong support for subsequent modules such as supply and demand matching, pricing prediction, and conversion recommendations.

[0052] Knowledge graphs use a graph structure to store extracted entities, relationships, and attributes. This structure intuitively displays complex relationships between entities, facilitating semantic reasoning and querying. Nodes in the graph represent entities (such as scientific and technological achievements, technical requirements, and the providers and demanders of scientific and technological achievements), while edges represent the relationships between entities. Each node and edge can also be accompanied by attribute information to provide richer descriptions.

[0053] The nodes of the knowledge graph include: Scientific and technological achievements: represents the scientific and technological achievement entity in the system, which may include attributes such as the name, description, technical field, and life cycle stage of the scientific and technological achievement.

[0054] Technical requirements: Represents the technical requirement entity in the system, which can include attributes such as the name, description, technical field, and expected solution of the technical requirement.

[0055] Provider of scientific and technological achievements: refers to the entity that provides scientific and technological achievements, such as scientific research institutions, universities, enterprises, etc., and can include attributes such as the provider's name, contact information, research direction, etc.

[0056] Technology demander: refers to the entity that proposes technology requirements, such as an enterprise or government department, and can include attributes such as the demander's name, contact information, and business field.

[0057] The edges of the knowledge graph include: The relationship between scientific and technological achievements and technological requirements: indicates the matching relationship between scientific and technological achievements and technological requirements, such as "scientific and technological achievement A meets technological requirement B".

[0058] The relationship between scientific and technological achievement information and the provider of scientific and technological achievements: indicates the ownership relationship between scientific and technological achievements and their providers, such as "scientific and technological achievement A was developed by provider C".

[0059] The relationship between technical requirement information and the technical demander: indicates the association between the technical requirement and the party proposing it, such as "technical requirement B is proposed by demander D".

[0060] The relationship between scientific and technological achievement information and technology demanders: indirectly indicates the potential interest or demand of technology demanders for scientific and technological achievements. Although this relationship is not direct, it can be inferred by analyzing the matching degree between technology demand and scientific and technological achievements.

[0061] The relationship between technology demand information and technology achievement providers: indirectly indicates that technology achievement providers may be interested in meeting specific technology needs. This relationship can be explored by analyzing the technical expertise of technology achievement providers and the needs of technology demanders.

[0062] To maintain the timeliness and accuracy of the knowledge graph, it employs a dynamic update mechanism to crawl new scientific and technological achievements, technical requirements, and changes in related entities and relationships in real time. This dynamic update mechanism can include regularly scanning data sources, monitoring data change notifications, and utilizing APIs to obtain the latest data. Once new information or changes are detected, the system immediately updates the corresponding nodes and edges in the knowledge graph. During the update process, the system must ensure data consistency and integrity to avoid data conflicts or loss. This can be achieved through technical means such as transaction processing and version control.

[0063] Through the above-mentioned construction method, node and edge definition, and dynamic update mechanism, the knowledge graph can reflect the latest status of scientific and technological achievements and technological needs and the relationship between them in real time and accurately, providing strong support for the intelligent recommendation and matching system for the transformation of scientific and technological achievements.

[0064] The supply-demand matching module uses graph algorithms, graph neural networks (GNNs), and cosine similarity calculations to perform semantic path analysis and prioritization. Within the knowledge graph, breadth-first search (BFS), depth-first search (DFS), or more complex path-finding algorithms (such as random walks and shortest path algorithms) are used to identify potential paths between scientific and technological achievement information and technical requirements. Path representation: These paths consist of a series of entities and relationships, such as "'Scientific and Technological Achievement A' to 'Satisfies' to 'Technical Requirement B'" or "'Scientific and Technological Achievement A' to 'Belongs to' to 'Technical Field X' to 'Related to' to 'Technical Requirement B'." Path analysis reveals indirect connections between scientific and technological achievements and technical requirements. These connections may be based on various factors, such as technical fields, application areas, and the background of the provider or demander.

[0065] Graph neural networks (such as GCN and GAT) are used to map the found semantic paths into low-dimensional vector representations. GNNs can capture the complex relationships between nodes and edges in a graph and convert them into a vector form that is easy for computers to process. Through the hierarchical propagation and aggregation mechanism of GNNs, the model can learn the implicit semantics of entities and relationships in the path. This semantic information is crucial for evaluating the relevance of the path. Each node (scientific and technological achievements, technical requirements, etc.) and edge (relationship) is represented as a low-dimensional vector that encodes its position in the graph and contextual information.

[0066] In a low-dimensional vector space, cosine similarity is used to calculate the similarity between the node vectors of scientific and technological achievements and the node vectors of technical requirements. Cosine similarity measures the directional similarity between two vectors; values ​​closer to 1 indicate greater similarity. The cosine similarity score can be used to assess the semantic relevance between scientific and technological achievements and technical requirements. A higher score indicates a closer semantic match between the two. Preliminary matching pairs are sorted in descending order based on their cosine similarity scores, with matching pairs with higher scores at the top. This sorting method ensures that the most relevant pairs of scientific and technological achievements and technical requirements are recommended to users first. Based on the priority sorting results, a recommendation list is generated for further review and evaluation by scientific and technological achievement providers and technical demanders.

[0067] Through the above process, the supply-demand matching module can efficiently and accurately identify the potential connections between scientific and technological achievements and technical requirements, and achieve precise matching and priority sorting through semantic path analysis and vector similarity calculation. This greatly improves the efficiency and success rate of scientific and technological achievement transformation.

[0068] Reinforcement learning model in the pricing of scientific and technological achievements, as shown in the attached Figure 2 As shown, the following steps are included: S1: State variable definition First, it is necessary to define state variables to describe the market performance and demand of current scientific and technological achievements in different dimensions. These state variables include: label information of scientific and technological achievements: such as technical fields, application directions, innovation points, etc. These labels help to distinguish the characteristics and market positioning of scientific and technological achievements. Transaction frequency: the trading activity of the scientific and technological achievements in the market, reflecting its popularity and market demand. Industry penetration rate: the popularity of the scientific and technological achievements in the target industry, reflecting its market acceptance and competitiveness. Technology life cycle stage: the life cycle stage of scientific and technological achievements (such as R&D stage, promotion stage, maturity stage, etc.). The market demand and pricing strategies at different stages may be different. These state variables together constitute a comprehensive description of the current market status of scientific and technological achievements by the reinforcement learning model.

[0069] S2: Pricing Action Selection Given the current state, the reinforcement learning model needs to select different pricing actions to respond to market changes. These include: Raising prices: When market demand for scientific and technological achievements is strong or the technology is advanced, prices can be appropriately raised to increase profits. Lowering prices: When market competition is fierce or rapid market capture is needed, prices can be lowered to attract more buyers. Maintaining prices: When market conditions are stable or further observation of market reactions is needed, the current price can be maintained. These pricing actions constitute the model's possible behavior space under specific conditions.

[0070] S3: Reward Function Evaluation In order to evaluate the effectiveness of each pricing action, a reward function needs to be designed to quantify the gains or losses brought about by the pricing action. The reward function can be defined based on the following factors: The transaction status of scientific and technological achievements: If the pricing action leads to the successful transaction of scientific and technological achievements, a positive reward is given; if the transaction fails, a negative reward or zero reward is given. Transaction price or profit: The higher the transaction price or the greater the profit, the reward value can be increased accordingly to encourage the model to choose a more profitable pricing strategy. Other market feedback: such as buyer satisfaction, market share changes, etc., can also be considered as factors in the reward function. Through the reward function, the model can learn what pricing actions to take in different states to obtain the maximum cumulative reward.

[0071] S4: Deep Q-Network Iterative Learning In order to obtain the optimal pricing strategy, a deep Q network (DQN) is used for iterative learning. The specific steps are as follows: Initialize DQN: Build a deep neural network as a Q-value function approximator, with state variables as input and the Q-value (i.e., expected cumulative reward) of each pricing action as output.

[0072] Experience Replay: During training, each state, action, reward, and next state are stored in an experience replay buffer. Training is performed by randomly sampling experience to improve data efficiency and stability.

[0073] Q value update: Update the Q value using the Bellman equation, that is: Where α is the learning rate, γ is the discount factor, r is the immediate reward, s' is the next state, and a' is the possible action in the next state.

[0074] Strategy optimization: By continuously iteratively updating the Q network parameters, the model gradually learns the strategy of selecting the optimal pricing action under a given state to maximize the transaction price of scientific and technological achievements or maximize profits.

[0075] Strategy execution: After training is completed, the learned optimal strategy will be applied to actual pricing decisions, and the pricing of scientific and technological achievements will be dynamically adjusted according to real-time market conditions.

[0076] Through the above steps, the reinforcement learning model can adaptively learn the optimal pricing strategy and achieve efficient pricing and transformation of scientific and technological achievements in a complex and changing market environment.

[0077] Furthermore, the multi-factor conversion evaluation algorithm takes the semantic matching degree between scientific and technological achievement information and technical demand information, the satisfaction with the predicted transaction price and the historical transaction rate of the same-label technology as evaluation factors. The satisfaction with the predicted transaction price is obtained by calculating the similarity between the expected price of the scientific and technological achievement provider or the technology demander and the predicted transaction price. The historical transaction rate of the same-label technology is obtained by calculating the number of successful conversions of the same-label technology and the number of recommendations of the same-label technology. The evaluation factors are mapped to the interval [0,1] to eliminate dimensional differences, and the hierarchical analysis method is used to determine the weight of each factor. Each factor is weighted to obtain a comprehensive conversion evaluation score for each preliminary matching pair.

[0078] The multi-factor transformation evaluation algorithm uses three main evaluation factors to comprehensively evaluate the matching quality and transformation potential of scientific and technological achievement information and technology demand information: Semantic Match: The degree of semantic match between scientific and technological achievement information and technical requirements information. This is typically calculated using natural language processing techniques (such as cosine similarity and Jaccard similarity) to calculate the similarity between the two description texts. This reflects the relevance of the content between scientific and technological achievements and technical requirements and is the basis for transformation evaluation.

[0079] Satisfaction with the predicted transaction price: This measure measures the acceptance of the predicted transaction price by the technology provider or technology demander. This is obtained by calculating the similarity between the expected price and the predicted transaction price. This can be measured using absolute difference, relative error, or a similarity function (such as cosine similarity). This reflects the impact of price factors on the conversion success rate and ensures that pricing strategies meet market expectations.

[0080] Historical Transaction Rate for Technologies with the Same Label: This is the ratio of the number of successful conversions to the number of recommendations for scientific and technological achievements with the same label (e.g., technical field, application direction, etc.). Calculation: Historical Transaction Rate = Number of Recommendations for Technologies with the Same Label / Number of Successful Conversions for Technologies with the Same Label. This reflects the conversion success rate of identical or similar technologies in the market, providing historical reference for current matching pairs.

[0081] Since the dimensions and value ranges of various evaluation factors may be different, they need to be mapped to the interval [0,1] to eliminate dimensional differences. Methods such as minimum-maximum normalization and Z-score standardization can be used.

[0082] The analytic hierarchy process (AHP) is used to assign weights to each evaluation factor: Through expert evaluation or questionnaires, compare the relative importance of each evaluation factor and construct a judgment matrix. For example, compare the relative importance of semantic match, satisfaction with the predicted transaction price, and historical transaction rate. Perform eigenvalue decomposition on the judgment matrix to obtain the weights of each evaluation factor. Ensure that the sum of the weights is 1 and that they meet the consistency test (CR < 0.1).

[0083] Multiply the normalized evaluation factor by its corresponding weight to obtain a comprehensive conversion evaluation score for each preliminary match. Preliminary matches are sorted in descending order based on the comprehensive conversion evaluation score, with matches with higher scores being recommended first. This provides a basis for decision-making for technology providers and technology demanders.

[0084] Through the above process, the multi-factor transformation evaluation algorithm can comprehensively and objectively evaluate the matching quality and transformation potential of scientific and technological achievements and technological needs, and improve the success rate and efficiency of the transformation of scientific and technological achievements.

[0085] Thus far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is clearly not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art may make equivalent changes or substitutions to the relevant technical features, and the technical solutions after such changes or substitutions will fall within the scope of protection of the present invention.

Claims

1. An intelligent recommendation matching system for supply and demand based on the transformation of scientific and technological achievements, characterized by: include: Information entry module: users register, select identities and log in through the information authentication interface. After logging in, users enter the corresponding information entry interface according to their identities to enter scientific and technological achievements information and technical requirements information. User identities include scientific and technological achievements providers and technical requirements parties. The information management module stores the input scientific and technological achievements and technical requirements information in a distributed manner in the scientific and technological achievements information database and the technical requirements information database, and labels each scientific and technological achievement and technical requirement with multi-dimensional tags. Through the tags of scientific and technological achievements, technical requirements, information about scientific and technological achievement providers and information about technical requirements, semantic association mining is carried out to build a knowledge graph of scientific and technological achievements and technical requirements. The supply and demand matching module uses preset labeling rules to preliminarily match scientific and technological achievement labels with technical requirement labels, generating many-to-many preliminary matching pairs of scientific and technological achievements and technical requirements. For the generated preliminary matching pairs, the semantic path reasoning method in the knowledge graph is used to calculate the semantic matching degree, and the preliminary matching pairs are prioritized according to the calculated semantic matching degree. The pricing prediction module uses a reinforcement learning model to predict the transaction price of scientific and technological achievements. The reinforcement learning model dynamically sets prices based on the transaction frequency, industry penetration rate, and technology life cycle stage of scientific and technological achievements with the same label. The conversion recommendation module uses a multi-factor conversion evaluation algorithm to generate a personalized conversion recommendation list. The multi-factor conversion evaluation algorithm performs a comprehensive evaluation based on the semantic matching degree between scientific and technological achievement information and technical demand information, the satisfaction with the predicted transaction price, and the historical transaction rate of the same label technology.

2. The intelligent recommendation matching system based on supply and demand of scientific and technological achievements transformation according to claim 1 is characterized in that: The user opens the information authentication interface on the electronic terminal, selects the user identity as the provider of scientific and technological achievements or the party in need of technology, and enters the user name, contact information and login password information. The information entry module verifies the validity of the input information. After the authentication is passed, the user logs in using the account and password set during registration, SMS verification code or third-party login method. After the user logs in, the user enters the corresponding information entry interface according to the user identity. The user enters the information content according to the format requirements. The scientific and technological achievement information entered by the provider of scientific and technological achievements includes at least the name of the scientific and technological achievement, technical field, description of the scientific and technological achievement, technical background, application direction and stage of the life cycle of the scientific and technological achievement. The technical requirement information entered by the technical demander includes at least the name of the technical requirement, technical field, description of the technical requirement, technical background, application direction and technical requirement problems. When the user performs an operation, real-time verification is performed through the role identification and permission list.

3. The intelligent recommendation matching system based on supply and demand of scientific and technological achievements transformation according to claim 1 is characterized in that: The scientific and technological achievement information database performs multi-dimensional classification based on the entered scientific and technological achievement information, and adds multi-dimensional labels to each scientific and technological achievement. The technical demand information database performs multi-dimensional classification based on the entered technical demand information, and adds multi-dimensional labels to each technical demand. The scientific and technological achievement information database and the technical demand information database use natural language models to extract keywords and semantically represent the entered scientific and technological achievements and technical demand texts, and extract entities, relationships and attributes from the entered scientific and technological achievement information and technical demand information as well as the registered scientific and technological achievement provider information and technical demand party information to construct a knowledge graph.

4. The intelligent recommendation matching system based on supply and demand of scientific and technological achievements transformation according to claim 3 is characterized in that: The natural language model uses a pre-trained language model to encode the entered scientific and technological achievement information and the entered technical requirement information, converts the unstructured description into a vector representation, and extracts keywords through the attention mechanism. Then, the vector representation generated by the transformation of technical fields, technical backgrounds and application directions, as well as the keywords extracted from the scientific and technological achievement description and the technical requirement description are calculated through cosine similarity to complete multi-dimensional classification.

5. The intelligent recommendation matching system based on supply and demand of scientific and technological achievements transformation according to claim 3 is characterized in that: The knowledge graph uses a graph structure to store extracted entities, relationships and attributes. The nodes of the knowledge graph are scientific and technological achievements, technical requirements, providers of scientific and technological achievements and technical demanders. The edges of the knowledge graph are the relationship between scientific and technological achievements and technical requirements, the relationship between scientific and technological achievement information and providers of scientific and technological achievements, the relationship between technical demand information and technical demanders, the relationship between scientific and technological achievement information and technical demanders, and the relationship between technical demand information and providers of scientific and technological achievements. The knowledge graph crawls incremental information in real time through a dynamic update mechanism.

6. The intelligent recommendation matching system based on supply and demand of scientific and technological achievements transformation according to claim 1 is characterized in that: In the supply and demand matching module, a graph algorithm is used to find the path between scientific and technological achievement information and technical demand information in the knowledge graph, and a graph neural network (GNN) is used to map the semantic path into a low-dimensional vector to capture the implicit semantics of entities and relationships in the path. In the low-dimensional vector space, the similarity between node vectors is calculated by cosine similarity to evaluate the semantic relevance of the path, and the preliminary matching pairs are prioritized in descending order of similarity scores.

7. The intelligent recommendation matching system based on supply and demand of scientific and technological achievements transformation according to claim 1 is characterized in that: The working method of the reinforcement learning model includes the following steps: S1. Use the tag information, transaction frequency, industry penetration rate, and technology life cycle stage of scientific and technological achievements as state variables to describe the market performance and demand of current scientific and technological achievements in different dimensions; S2. Select different pricing actions based on the current status. Pricing actions include raising the price, lowering the price, and keeping the price unchanged. S3. Use a reward function to evaluate the effect of each pricing action and adjust the degree of reward based on the transaction volume of scientific and technological achievements; S4. Use deep Q network for iterative learning and update the pricing strategy based on the reward signal to obtain the optimal pricing strategy to maximize the transaction price of scientific and technological achievements or maximize profits.

8. The intelligent recommendation matching system based on supply and demand of scientific and technological achievements transformation according to claim 1 is characterized in that: The multi-factor conversion evaluation algorithm uses the semantic matching degree between scientific and technological achievement information and technical demand information, the satisfaction degree of predicted transaction price and the historical transaction rate of the same-label technology as evaluation factors. The satisfaction degree of predicted transaction price is obtained by calculating the similarity between the expected price of the scientific and technological achievement provider or the technology demander and the predicted transaction price. The historical transaction rate of the same-label technology is obtained by calculating the number of successful conversions of the same-label technology and the number of recommendations of the same-label technology. The evaluation factors are mapped to the interval [0,1] to eliminate dimensional differences, and the hierarchical analysis method is used to determine the weight of each factor. Each factor is weighted to obtain a comprehensive conversion evaluation score for each preliminary matching pair.

Images