Scientific and technological achievement intelligent matching and transaction recommendation method and system based on knowledge graph

Abstract

The invention provides a scientific and technological achievement intelligent matching and transaction recommendation method and system based on a knowledge graph, and relates to the technical field of science and technology, and the method comprises the steps: extracting a technical gene sequence based on a scientific and technological achievement data set, generating a technical state vector, constructing a technical association intensity matrix and a technical iteration path graph, and calculating a technical potential energy distribution field; and forming an initial knowledge graph. And monitoring the fluctuation of the knowledge graph state feature set through the intelligent sensor, and dynamically optimizing the knowledge graph according to the fluctuation condition. Converting technical requirements of a user into a query state, performing state similarity calculation with the knowledge graph, identifying an optimal technical unit combination, predicting a state evolution trajectory, evaluating a regional adaptation degree, and generating a technical implementation scheme. By dynamically optimizing the knowledge graph and accurately matching the user demands, the efficiency and accuracy of scientific and technological achievement matching and transaction recommendation can be effectively improved, and scientific and technological achievement conversion is promoted.

Description

Technical Field

[0001] The present invention relates to technology, and in particular to an intelligent matching and trading recommendation method and system for scientific and technological achievements based on a knowledge graph. Background Art

[0002] The transformation of scientific and technological achievements is a key link in scientific and technological innovation activities and an important way to realize the integrated development of science and technology and the economy. Traditional methods for matching and trading scientific and technological achievements mainly rely on manual search, offline docking, etc., with low efficiency and high costs, and it is difficult to meet the growing demand for the transformation of scientific and technological achievements. In recent years, with the rapid development of technologies such as big data and artificial intelligence, intelligent matching and trading recommendation methods for scientific and technological achievements based on knowledge graphs have gradually emerged, providing new ideas and methods for the transformation of scientific and technological achievements.

[0003] Insufficient mining of technical relevance: Most existing methods perform scientific and technological achievement matching based on keyword matching or simple semantic similarity calculation, making it difficult to deeply mine the deep relevance between scientific and technological achievements, resulting in low matching accuracy and high missed matching rates.

[0004] Insufficient prediction of technological evolution trends: Existing methods lack the ability to predict the development trends of technologies, making it difficult to effectively evaluate the future development potential and market value of scientific and technological achievements, resulting in blindness in transaction decisions.

[0005] Insufficient consideration of regional adaptability of technical implementation plans: Existing methods rarely consider the regional differences in technical implementation, such as factors like policy environment, industrial foundation, and talent resources, resulting in low landing and feasibility of technical implementation plans. Summary of the Invention

[0006] Embodiments of the present invention provide an intelligent matching and trading recommendation method and system for scientific and technological achievements based on a knowledge graph, which can solve the problems in the prior art.

[0007] In the first aspect of the embodiments of the present invention,

[0008] There is provided an intelligent matching and trading recommendation method for scientific and technological achievements based on a knowledge graph, including:

[0009] Obtain a scientific and technological achievement dataset, and extract technical gene sequences from the scientific and technological achievement dataset through a deep feature extractor; input the technical gene sequences into a high-dimensional coding network to generate technical state vectors; construct a technical association strength matrix based on the technical state vectors, and the technical association strength matrix records the coupling degree between technical units; calculate a technical evolution coefficient according to the coupling degree to generate a technical iteration path diagram; use the technical iteration path diagram to quantitatively calculate the innovation potential values of each technical unit to form a technical potential distribution field; integrate the technical association strength matrix, the technical iteration path diagram, and the technical potential distribution field into an initial knowledge graph, and output the state feature set of the initial knowledge graph;

[0010] Deploy an intelligent sensor to monitor the fluctuations of the state feature set in real time; when the monitoring result of the state fluctuation set exceeds a preset dynamic threshold, trigger a graph optimization signal; according to the graph optimization signal, call a coupling calculation unit to recalculate the technical association strength matrix; based on the updated technical association strength matrix, use a dynamic optimization algorithm to optimize the technical iteration path diagram; use the optimized technical iteration path diagram to recalculate the technical potential distribution field; integrate the updated technical association strength matrix, the technical iteration path diagram, and the technical potential distribution field into a new knowledge graph state, and output the updated state feature set;

[0011] Convert the user's technical requirements into a query state, calculate the state similarity with the updated state feature set to obtain the state similarity; identify the optimal technical unit combination based on the state similarity; use the technical iteration path diagram to predict the state evolution trajectory of the optimal technical unit combination; combine the technical potential distribution field to evaluate the regional adaptability of the technical implementation, generate a technical implementation plan, and record the user's feedback information on the technical implementation plan; convert the feedback information into a state correction factor to adjust the dynamic threshold to achieve the dynamic optimization of the new knowledge graph state.

[0012] Calculate a technical evolution coefficient according to the coupling degree to generate a technical iteration path diagram; use the technical iteration path diagram to quantitatively calculate the innovation potential values of each technical unit to form a technical potential distribution field; integrate the technical association strength matrix, the technical iteration path diagram, and the technical potential distribution field into an initial knowledge graph, and the state feature set output by the initial knowledge graph includes:

[0013] Input the coupling degree into the technology evolution prediction model, which calculates the evolution coefficient of the technology unit based on the dominance ratio, association strength ratio, and technology maturity difference between technology units; combine the evolution coefficient with the technology development constraint conditions, which include the time sequence limit, technology generation span limit, and technology branch convergence limit of technology development; generate a technology iteration path map based on the evolution coefficient and the technology development constraint conditions, and the edge weight of the technology iteration path map is determined by the ratio of the evolution coefficients of the source technology unit and the target technology unit;

[0014] Calculate the innovation potential value of the technology unit using the technology iteration path map, map the innovation potential value to the geographical space coordinate system, and construct a technology potential distribution field based on the spatial aggregation effect, regional relevance, and regional technology foundation;

[0015] Perform tensor fusion on the technology association strength matrix, the technology iteration path map, and the technology potential distribution field to construct an initial knowledge graph with multi-dimensional associations; the initial knowledge graph includes a topological structure layer, a technology evolution layer, and a potential distribution layer of technology units;

[0016] Extract features from the initial knowledge graph, extract the association features of the topological structure layer based on the graph neural network, extract the time sequence features of the technology evolution layer based on the recurrent neural network, and extract the spatial features of the potential distribution layer based on the convolutional neural network; fuse the association features, the time sequence features, and the spatial features to form a state feature set of the knowledge graph.

[0017] Deploy an intelligent sensor to monitor the fluctuations of the state feature set in real time; when the monitoring result of the state fluctuation set exceeds a preset dynamic threshold, trigger a graph optimization signal; according to the graph optimization signal, call the coupling calculation unit to recalculate the technology association strength matrix; based on the updated technology association strength matrix, optimize the technology iteration path map using a dynamic optimization algorithm, including:

[0018] Deploy an intelligent sensor to monitor the fluctuations of the state feature set. The intelligent sensor collects technology node feature data, technology association feature data, and graph global feature data, calculates the feature deviation degree based on the technology node feature data, calculates the coupling change rate based on the technology association feature data, and calculates the situation evolution rate based on the graph global feature data;

[0019] Input the feature deviation degree, the coupling change rate, and the situation evolution rate into a feature fusion model, and the feature fusion model uses a weighted combination method to generate a state feature fluctuation index, which characterizes the real-time situation of the knowledge graph;

[0020] Build a dynamic threshold calculation model. The dynamic threshold calculation model establishes a benchmark threshold based on the historical data distribution of the state feature fluctuation index, and dynamically adjusts the benchmark threshold according to the technology development cycle function and the environmental impact factor to obtain an adaptive dynamic threshold;

[0021] Compare the state feature fluctuation index with the adaptive dynamic threshold in real time. When the state feature fluctuation index exceeds the adaptive dynamic threshold, calculate the influence degree of the fluctuation occurrence area, and generate a map optimization signal including an optimization area, an optimization depth, and an optimization direction based on the influence degree;

[0022] The map optimization signal triggers the coupling calculation unit to perform an update operation. The coupling calculation unit obtains the latest technology data, recalculates the technology association strength of the corresponding area based on the latest technology data and the optimization area, and generates an updated technology association strength matrix;

[0023] Input the updated technology association strength matrix into the path optimization model. The path optimization model identifies the path nodes that need to be adjusted based on the optimization depth, constructs path evolution constraint conditions according to the optimization direction, and uses the dynamic programming algorithm to obtain the optimal technology iteration path map according to the path nodes and the path evolution constraint conditions.

[0024] The map optimization signal triggers the coupling calculation unit to perform an update operation. The coupling calculation unit obtains the latest technology data, recalculates the technology association strength of the corresponding area based on the latest technology data and the optimization area, and generating an updated technology association strength matrix includes:

[0025] The map optimization signal triggers the coupling calculation unit to perform an update operation. The coupling calculation unit uses the locality-sensitive hashing algorithm to divide the influence area range into multiple sub-regions, calculates the local density and relative distance of the technology nodes in the multiple sub-regions, identifies the technology nodes, and constructs an influence propagation network centered on the technology nodes;

[0026] The coupling calculation unit obtains the latest technology data from the multi-source technology database, performs semantic analysis on the latest technology data using a deep learning model, extracts a technology feature vector including technology function attributes, application scenario features, and innovation point descriptions, and establishes a time-series feature extraction module to capture the dynamic trend information of technology development; the coupling calculation unit calculates the technology association strength within the influence area range based on the technology feature vector and the dynamic trend information;

[0027] The coupling calculation unit uses a multi-level coupling calculation method to recalculate the technology association strength within the influence area range, and the coupling calculation unit generates an updated technology association strength matrix based on the calculation results of the multi-level coupling calculation method.

[0028] Using the optimized technical iteration path diagram, recalculate the technical potential energy distribution field; integrate the updated technical association strength matrix, the technical iteration path diagram, and the technical potential energy distribution field into a new knowledge graph state, and output the updated state feature set, including:

[0029] Extract the path node degree centrality, path betweenness centrality, and path cluster cohesion in the optimized technical iteration path diagram, and construct a technology diffusion matrix, which records the propagation strength and coverage of technical knowledge in the path network;

[0030] Based on the path node degree centrality, the path betweenness centrality, the path cluster cohesion, and the technology diffusion matrix, recalculate the technical potential energy distribution field, which includes basic potential energy, position potential energy, and development potential energy. The basic potential energy is calculated based on the technical innovation level, the position potential energy is calculated based on the network structure status, and the development potential energy is calculated based on the future evolution trend;

[0031] Perform data standardization processing on the updated technical association strength matrix, convert the technical iteration path diagram into a standard network representation, convert the technical potential energy distribution field into a unified digital format, and construct a multi-dimensional data integration framework;

[0032] Establish a vertical mapping relationship and a horizontal connection relationship in the multi-dimensional data integration framework. The vertical mapping relationship determines the corresponding method of data at different levels, and the horizontal connection relationship determines the interaction method of data at the same level. Extract the common features of multi-dimensional data based on the tensor decomposition method; input the common features into the feature alignment module. The feature alignment module unifies the expression methods of different dimensions, uses an adaptive weight allocation algorithm to determine the importance of features in each dimension, and integrates the aligned features into a new knowledge graph state;

[0033] Extract static features and dynamic features based on the new knowledge graph state. The static features include structural features, distribution features, and attribute features, and the dynamic features include evolution features, flow features, and interaction features; input the static features and the dynamic features into a feature fusion device. The feature fusion device uses a multi-layer perceptron for feature combination and realizes feature dimensionality reduction through the principal component analysis method to generate an updated state feature set.

[0034] Convert the user's technical requirements into a query state, calculate the state similarity with the updated state feature set to obtain the state similarity; identify the optimal technical unit combination based on the state similarity; predict the state evolution trajectory of the optimal technical unit combination using the technical iteration path diagram, including:

[0035] Receive the technical requirements of users, extract the technical field features, functional requirement features, and performance index features in the technical requirements of users by using a deep semantic analysis model, construct a feature representation matrix, and the feature representation matrix quantitatively describes the multi-dimensional attributes of the requirements;

[0036] Based on the feature representation matrix, construct a query state vector through a feature mapping network. The feature mapping network includes a static feature encoding layer and a dynamic feature encoding layer. The static feature encoding layer processes technical attributes and functional attributes, and the dynamic feature encoding layer processes temporal features and evolutionary features;

[0037] Calculate the similarity between the query state vector and the updated state feature set, use the attention mechanism to assign adaptive weights to different feature dimensions, calculate the distance between state vectors through a deep metric learning model, and generate a state similarity matrix;

[0038] Use the state similarity matrix to construct a technical unit affinity graph. The nodes of the technical unit affinity graph represent technical units, and the edge weights represent state similarities. Use a community discovery algorithm to identify high-cohesion technical unit combinations, and screen the optimal technical unit combinations based on a combination scoring function;

[0039] Conduct a historical evolution analysis on the optimal technical unit combination, extract an evolution feature sequence based on the technology iteration path graph, and the evolution feature sequence includes changes in technology maturity, innovation activity, and association strength;

[0040] Input the evolution feature sequence into a temporal prediction model. The temporal prediction model adopts a long short-term memory network structure, captures long-term dependencies through a gating mechanism, combines the attention mechanism to identify temporal patterns, and generates a technology evolution prediction sequence;

[0041] Construct a multi-dimensional evolution path graph based on the technology evolution prediction sequence. The nodes of the multi-dimensional evolution path graph represent technology states, and the edges represent state transition probabilities. Optimize the path selection strategy through a Monte Carlo tree search algorithm, and output the optimal state evolution trajectory.

[0042] Combine the technical potential distribution field to evaluate the regional adaptability of technology implementation, generate a technology implementation plan, and record the feedback information of the user on the technology implementation plan; convert the feedback information into a state correction factor for adjusting the dynamic threshold to achieve the dynamic optimization of the new knowledge graph state, including:

[0043] Evaluate the regional adaptability of technology implementation based on the technical potential distribution field, and generate a technology implementation plan according to the regional adaptability. The technology implementation plan includes a technology path planning and resource allocation plan formulated based on the regional adaptability;

[0044] A multi-modal feedback collection system is adopted to record the implementation effect feedback of users on the technical implementation solution. The multi-modal feedback collection system obtains a multi-dimensional feature feedback data set, and the multi-dimensional feature feedback data set includes text feedback data, voice feedback data, and behavior feedback data; the multi-dimensional feature feedback data set is input into a deep semantic understanding network, and the deep semantic understanding network extracts semantic features and emotional features to generate a feedback feature vector;

[0045] A state correction factor is constructed based on the feedback feature vector, and the state correction factor includes a technical implementation effect score and an optimization direction indication; the state correction factor is input into a dynamic threshold adjustment module, and the dynamic threshold adjustment module updates the triggering threshold of the evolution rule of the knowledge graph state according to the technical implementation effect score and the optimization direction indication;

[0046] The dynamic optimization of the knowledge graph state is realized by using a graph structure optimization network. The graph structure optimization network determines the rule triggering condition according to the dynamic threshold adjustment module and applies the rule that meets the triggering condition to the structure update of the knowledge graph; a multi-objective optimization framework is constructed, and the multi-objective optimization framework uses the state correction factor as an optimization constraint to guide the evolution direction of the knowledge graph state;

[0047] A new knowledge graph state is generated based on the optimization result of the multi-objective optimization framework, and the optimization effect of the new knowledge graph state is fed back to the multi-modal feedback collection system to form a continuously optimized closed-loop mechanism.

[0048] In the second aspect of the embodiments of the present invention,

[0049] An intelligent matching and trading recommendation system for scientific and technological achievements based on a knowledge graph is provided, including:

[0050] A first unit is used to obtain a scientific and technological achievement data set, extract a technical gene sequence from the scientific and technological achievement data set through a deep feature extractor; input the technical gene sequence into a high-dimensional coding network to generate a technical state vector; construct a technical association strength matrix based on the technical state vector, and the technical association strength matrix records the coupling degree between technical units; calculate a technical evolution coefficient according to the coupling degree to generate a technical iteration path diagram; quantitatively calculate the innovation potential value of each technical unit by using the technical iteration path diagram to form a technical potential distribution field; integrate the technical association strength matrix, the technical iteration path diagram, and the technical potential distribution field into an initial knowledge graph, and output the state feature set of the initial knowledge graph;

[0051] A second unit for deploying intelligent sensors to monitor the fluctuations of the state feature set in real time; when the monitoring result of the state fluctuation set exceeds a preset dynamic threshold, triggering a map optimization signal; according to the map optimization signal, calling a coupling calculation unit to recalculate the technical association strength matrix; based on the updated technical association strength matrix, using a dynamic optimization algorithm to optimize the technical iteration path map; using the optimized technical iteration path map to recalculate the technical potential distribution field; integrating the updated technical association strength matrix, the technical iteration path map, and the technical potential distribution field into a new knowledge map state, and outputting an updated state feature set;

[0052] A third unit for converting the user's technical requirements into a query state, calculating the state similarity with the updated state feature set to obtain the state similarity; identifying the optimal technical unit combination based on the state similarity; predicting the state evolution trajectory of the optimal technical unit combination using the technical iteration path map; evaluating the regional adaptability of technology implementation in combination with the technical potential distribution field, generating a technology implementation plan, and recording the user's feedback information on the technology implementation plan; converting the feedback information into a state correction factor for adjusting the dynamic threshold to achieve dynamic optimization of the new knowledge map state.

[0053] In the third aspect of the embodiments of the present invention,

[0054] Provided is an electronic device, including:

[0055] A processor;

[0056] A memory for storing instructions executable by the processor;

[0057] Wherein, the processor is configured to call the instructions stored in the memory to execute the method described above.

[0058] In the fourth aspect of the embodiments of the present invention,

[0059] Provided is a computer-readable storage medium, on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the method described above is implemented.

[0060] The beneficial effects of this application are as follows:

[0061] 1. Improve the matching and transaction efficiency of scientific and technological achievements: By means of technologies such as technical gene sequences, technical state vectors, and technical association strength matrices, a knowledge map of scientific and technological achievements is constructed, which can more accurately match the user's technical requirements with scientific and technological achievements, reduce information asymmetry, and improve transaction efficiency.

[0062] 2. Achieve precise recommendation of technological achievements: Based on state similarity calculation, prediction of technological evolution trajectories, and evaluation of regional adaptability, the optimal combination of technological units and technological implementation plans can be recommended to users, avoiding blind selection of technologies and improving the success rate of technological achievement transformation.

[0063] 3. Continuously optimize the knowledge graph and recommendation effect: Utilize user feedback information to dynamically adjust the dynamic threshold of the knowledge graph, achieve continuous optimization of the knowledge graph, and thus continuously improve the accuracy and effectiveness of scientific and technological achievement matching and transaction recommendation. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 It is a schematic flowchart of the method for intelligent matching and transaction recommendation of scientific and technological achievements based on a knowledge graph according to an embodiment of the present invention;

[0065] Figure 2 It is a schematic structural diagram of the system for intelligent matching and transaction recommendation of scientific and technological achievements based on a knowledge graph according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0066] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the scope of protection of the present invention.

[0067] The technical solutions of the present invention will be described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments.

[0068] Figure 1 It is a schematic flowchart of the method for intelligent matching and transaction recommendation of scientific and technological achievements based on a knowledge graph according to an embodiment of the present invention, as Figure 1 shown, and the method includes:

[0069] S11. Obtain a scientific and technological achievement dataset, and extract technical gene sequences from the scientific and technological achievement dataset through a deep feature extractor; input the technical gene sequences into a high-dimensional coding network to generate technical state vectors; construct a technical association strength matrix based on the technical state vectors, where the technical association strength matrix records the coupling degree between technical units; calculate a technical evolution coefficient according to the coupling degree, generate a technical iteration path diagram; use the technical iteration path diagram to quantitatively calculate the innovation potential values of each technical unit to form a technical potential distribution field; integrate the technical association strength matrix, the technical iteration path diagram, and the technical potential distribution field into an initial knowledge graph, and output the state feature set of the initial knowledge graph;

[0070] S12. Deploy an intelligent sensor to monitor the fluctuations of the state feature set in real time; when the monitoring result of the state fluctuation set exceeds a preset dynamic threshold, trigger a graph optimization signal; according to the graph optimization signal, call a coupling calculation unit to recalculate the technical association strength matrix; based on the updated technical association strength matrix, use a dynamic optimization algorithm to optimize the technical iteration path diagram; use the optimized technical iteration path diagram to recalculate the technical potential distribution field; integrate the updated technical association strength matrix, the technical iteration path diagram, and the technical potential distribution field into a new knowledge graph state, and output the updated state feature set;

[0071] S13. Convert the user's technical requirements into a query state, calculate the state similarity with the updated state feature set to obtain the state similarity; identify the optimal technical unit combination based on the state similarity; use the technical iteration path diagram to predict the state evolution trajectory of the optimal technical unit combination; combine the technical potential distribution field to evaluate the regional adaptability of the technical implementation, generate a technical implementation plan, and record the user's feedback information on the technical implementation plan; convert the feedback information into a state correction factor to adjust the dynamic threshold to achieve the dynamic optimization of the new knowledge graph state.

[0072] In an alternative embodiment, calculating a technical evolution coefficient according to the coupling degree, generating a technical iteration path diagram; using the technical iteration path diagram to quantitatively calculate the innovation potential values of each technical unit to form a technical potential distribution field; integrating the technical association strength matrix, the technical iteration path diagram, and the technical potential distribution field into an initial knowledge graph, and outputting the state feature set of the initial knowledge graph includes:

[0073] Input the coupling degree into the technology evolution prediction model. The technology evolution prediction model calculates the evolution coefficient of the technology unit based on the dominance ratio, correlation strength ratio, and technology maturity difference between technology units; combine the evolution coefficient with the technology development constraint conditions, where the technology development constraint conditions include the time sequence limit of technology development, the technology generation span limit, and the technology branch convergence limit; generate a technology iteration path diagram based on the evolution coefficient and the technology development constraint conditions, and the edge weight of the technology iteration path diagram is determined by the ratio of the evolution coefficients of the source technology unit and the target technology unit;

[0074] Use the technology iteration path diagram to calculate the innovation potential value of the technology unit, map the innovation potential value to the geographical space coordinate system, and construct a technology potential distribution field based on the spatial aggregation effect, regional relevance, and regional technology foundation;

[0075] Perform tensor fusion on the technology correlation strength matrix, the technology iteration path diagram, and the technology potential distribution field to construct an initial knowledge graph with multi-dimensional associations; the initial knowledge graph includes a topological structure layer, a technology evolution layer, and a potential distribution layer of technology units;

[0076] Extract features from the initial knowledge graph, extract the association features of the topological structure layer based on the graph neural network, extract the time sequence features of the technology evolution layer based on the recurrent neural network, and extract the spatial features of the potential distribution layer based on the convolutional neural network; fuse the association features, the time sequence features, and the spatial features to form a state feature set of the knowledge graph.

[0077] First, obtain the coupling degree data between technology units. For example, by analyzing patent text data, scientific and technological literature data, or technology cooperation network data, the frequency of co-occurrence or the strength of cooperation relationships between different technology units can be obtained, and this is used as a measure of the coupling degree between technology units. Suppose there are technology units A, B, C, and D, and through data analysis, the coupling degrees between them are: A - B: 0.8, A - C: 0.5, A - D: 0.2, B - C: 0.6, B - D: 0.3, C - D: 0.7. These data will form a technology correlation strength matrix.

[0078] Next, calculate the technology evolution coefficient based on the coupling degree and generate a technology iteration path diagram. The calculation of the technology evolution coefficient needs to consider the dominance ratio, correlation intensity ratio, and technology maturity difference between technology units. For example, if technology unit A has a higher dominance, stronger correlation intensity, and higher technology maturity compared to technology unit B, the evolution coefficient from A to B will be relatively low, and vice versa. According to the calculated evolution coefficient and technology development constraints (such as temporal constraints of technology development, technology generation span constraints, and technology branch convergence constraints), a technology iteration path diagram can be generated. The nodes in the diagram represent technology units, the edges represent the evolution paths between technology units, and the weight of the edge is determined by the ratio of the evolution coefficients of the source technology unit and the target technology unit. For example, if the evolution coefficient from A to B is 0.2 and the evolution coefficient from B to C is 0.5, the weight of the edge from A to B is 0.2 / 0.5 = 0.4.

[0079] Then, use the technology iteration path diagram to calculate the innovation potential value of technology units and construct a technology potential distribution field. The innovation potential value of a technology unit can be calculated by analyzing factors such as its position, connection relationship, and evolution coefficient in the technology iteration path diagram. For example, a technology unit located at the center of the technology iteration path diagram, connected to multiple other technology units, and having a high evolution coefficient usually has a higher innovation potential value. Mapping the calculated innovation potential value to a geographical space coordinate system and combining factors such as spatial aggregation effect, regional correlation, and regional technology foundation, a technology potential distribution field can be constructed. For example, if a certain region aggregates multiple technology units with high innovation potential values, the overall technology potential of this region will also be relatively high.

[0080] Next, fuse the technology correlation intensity matrix, the technology iteration path diagram, and the technology potential distribution field to construct an initial knowledge graph with multi-dimensional associations. This knowledge graph includes the topological structure layer, technology evolution layer, and potential distribution layer of technology units, corresponding to the technology correlation intensity matrix, the technology iteration path diagram, and the technology potential distribution field respectively. For example, the topological structure layer information of technology unit A in the knowledge graph includes its connection relationship and connection strength with other technology units, the technology evolution layer information includes its position and evolution coefficient in the technology iteration path diagram, and the potential distribution layer information includes its position and innovation potential value in the geographical space coordinate system.

[0081] Finally, feature extraction is performed on the initial knowledge graph to form a state feature set of the knowledge graph. The graph neural network is used to extract the correlation features of the topological structure layer, the recurrent neural network is used to extract the temporal features of the technology evolution layer, and the convolutional neural network is used to extract the spatial features of the potential energy distribution layer. These features are fused to form a state feature set of the knowledge graph, which is used for subsequent technology prediction and analysis. For example, the extracted state feature set can be input into a machine learning model to predict the future development trend of technologies or identify technology fields with high innovation potential.

[0082] The solution of this application can:

[0083] Improve the accuracy of technology prediction: By constructing a knowledge graph with multi-dimensional associations, the relationships and evolution laws between technology units can be understood more comprehensively, thereby improving the accuracy of technology prediction. Identify technology fields with high innovation potential: By analyzing the technology potential distribution field, technology units and regions with high innovation potential can be identified, providing decision-making support for technology R & D and investment. Promote technology innovation and development: By constructing a technology iteration path graph, the direction and path of technology evolution can be better understood, thereby promoting technology innovation and development.

[0084] In an optional implementation manner, an intelligent sensor is deployed to monitor the fluctuations of the state feature set in real time; when the monitoring result of the state fluctuation set exceeds a preset dynamic threshold, a graph optimization signal is triggered; according to the graph optimization signal, a coupled computing unit is called to recalculate the technology association strength matrix; based on the updated technology association strength matrix, optimizing the technology iteration path graph using a dynamic optimization algorithm includes:

[0085] Deploy an intelligent sensor to monitor the fluctuations of the state feature set. The intelligent sensor collects technology node feature data, technology association feature data, and graph global feature data, calculates the feature deviation degree based on the technology node feature data, calculates the coupling change rate based on the technology association feature data, and calculates the situation evolution rate based on the graph global feature data;

[0086] Input the feature deviation degree, the coupling change rate, and the situation evolution rate into a feature fusion model. The feature fusion model uses a weighted combination method to generate a state feature fluctuation index, and the state feature fluctuation index characterizes the real-time situation of the knowledge graph;

[0087] Construct a dynamic threshold calculation model. The dynamic threshold calculation model establishes a benchmark threshold based on the historical data distribution of the state feature fluctuation index, and dynamically adjusts the benchmark threshold according to the technology development cycle function and the environmental impact factor to obtain an adaptive dynamic threshold;

[0088] Compare the state feature fluctuation index with the adaptive dynamic threshold in real time. When the state feature fluctuation index exceeds the adaptive dynamic threshold, calculate the influence degree of the fluctuation occurrence area, and generate a map optimization signal including an optimization area, an optimization depth, and an optimization direction based on the influence degree;

[0089] The map optimization signal triggers the coupling calculation unit to perform an update operation. The coupling calculation unit obtains the latest technical data, recalculates the technical association strength of the corresponding area based on the latest technical data and the optimization area, and generates an updated technical association strength matrix;

[0090] Input the updated technical association strength matrix into the path optimization model. The path optimization model identifies the path nodes that need to be adjusted based on the optimization depth, constructs path evolution constraint conditions according to the optimization direction, and uses the dynamic programming algorithm to obtain the optimal technical iteration path map based on the path nodes and the path evolution constraint conditions.

[0091] First, deploy intelligent sensors to monitor the fluctuations of the technical situation feature set in real time. The intelligent sensors collect data from multiple data sources, including but not limited to patent databases, scientific and technological literature databases, industry reports, news and information, etc. The collected data covers three aspects: technical node feature data, such as the maturity, market share, development trend, etc. of the technology; technical association feature data, such as the cooperation relationship, competition relationship, dependence relationship, etc. between technologies; and map global feature data, such as the overall development level of the technology, the distribution of hot spots, the emergence of emerging technologies, etc.

[0092] Then, the intelligent sensors perform feature extraction and calculation based on the collected data. For example, calculate the feature deviation based on the technical node feature data to measure the deviation degree of the development state of each technical node from its expected state. Suppose the expected maturity of a certain technical node is 0.8, and the actual maturity is 0.6, then its feature deviation can be calculated as 0.2. Calculate the coupling change rate based on the technical association feature data to measure the change speed of the association relationship between technologies. For example, if the cooperation relationship between two technologies significantly strengthens within a certain period of time, then its coupling change rate will increase accordingly. Calculate the situation evolution rate based on the map global feature data to measure the change speed of the overall technical situation. For example, if the technology development speed in a certain field suddenly accelerates, then the situation evolution rate of this field will increase accordingly.

[0093] Next, input the calculated feature deviation degree, coupling change rate, and situation evolution rate into the feature fusion model. This model uses a weighted combination method to fuse the three indicators into a single state feature fluctuation indicator, which is used to characterize the real-time situation of the knowledge graph. For example, according to the experience of domain experts or historical data analysis, different weights can be assigned to the three indicators, and they are weighted and summed to obtain the state feature fluctuation indicator. Suppose the weights of the feature deviation degree, coupling change rate, and situation evolution rate are 0.4, 0.3, and 0.3 respectively, and their values are 0.2, 0.5, and 0.4 respectively. Then the state feature fluctuation indicator is 0.2 * 0.4 + 0.5 * 0.3 + 0.4 * 0.3 = 0.35.

[0094] To determine whether the situation change has reached the level that requires path optimization, a dynamic threshold calculation model is constructed. This model first establishes a benchmark threshold based on the historical data distribution of the state feature fluctuation indicator. For example, the average value and standard deviation of the state feature fluctuation indicator over a past period of time can be statistically analyzed, and the average value plus twice the standard deviation is used as the benchmark threshold. Then, the benchmark threshold is dynamically adjusted according to the technology development cycle function and environmental impact factor to obtain an adaptive dynamic threshold. For example, during a period of rapid technology development, the threshold can be appropriately lowered to more sensitively capture situation changes; while during a period of stable technology development, the threshold can be appropriately raised to avoid frequent path adjustments. Suppose the benchmark threshold is 0.4, and currently it is in a period of rapid technology development with an environmental impact factor of 0.9. Then the adaptive dynamic threshold can be calculated as 0.4 * 0.9 = 0.36.

[0095] Compare the state feature fluctuation indicator with the adaptive dynamic threshold in real time. When the state feature fluctuation indicator exceeds the adaptive dynamic threshold, calculate the influence degree of the fluctuation occurrence area. For example, the number of technology nodes involved in the fluctuation area, their importance, and the association relationships between them can be analyzed to evaluate the influence degree. Based on the influence degree, a graph optimization signal including the optimization area, optimization depth, and optimization direction is generated. For example, if the influence degree is large, the optimization depth may need to be deeper, and the optimization direction may need to be adjusted more significantly. Suppose the state feature fluctuation indicator is 0.37, which exceeds the adaptive dynamic threshold of 0.36, and the fluctuation is mainly concentrated in the field of artificial intelligence with a high influence degree evaluated. Then the generated graph optimization signal may be: the optimization area is the field of artificial intelligence, the optimization depth is high, and the optimization direction is to strengthen the research in the direction of deep learning technology.

[0096] The map optimization signal triggers the coupling calculation unit to perform an update operation. The coupling calculation unit obtains the latest technical data, recalculates the technical association strength of the corresponding area based on the latest technical data and the optimization area, and generates an updated technical association strength matrix. For example, if the technical association strength in the field of artificial intelligence needs to be updated, the coupling calculation unit will collect the latest data on the relationships between various technologies in the field of artificial intelligence and recalculate the association strength between them.

[0097] Finally, the updated technical association strength matrix is input into the path optimization model. The path optimization model identifies the path nodes that need to be adjusted based on the optimization depth, and constructs path evolution constraint conditions according to the optimization direction. For example, if the optimization depth is relatively high, more path nodes may need to be adjusted; if the optimization direction is to strengthen deep learning technology, the technical nodes related to deep learning need to be given priority. According to the path nodes and the path evolution constraint conditions, the dynamic programming algorithm is used to obtain the optimal technology iteration path graph. For example, the dynamic programming algorithm can be used to find an optimal path from the current technical state to the target technical state under the condition of meeting the constraint conditions.

[0098] The solution of this application can:

[0099] Improve the efficiency of technology R & D: By monitoring the changes in the technology situation in real time and dynamically adjusting the technology iteration path, it is possible to avoid wasting resources in the wrong direction, thereby improving the efficiency of technology R & D. Enhance the success rate of technology R & D: By adaptively adjusting the path according to the changes in the technology situation, it is possible to better adapt to the technology development trend and market demand, thereby enhancing the success rate of technology R & D. Increase the flexibility of technology path planning: The combination of dynamic thresholds and path optimization algorithms makes technology path planning more flexible and better able to cope with the complex and changeable technology development environment.

[0100] In an alternative embodiment, the map optimization signal triggers the coupling calculation unit to perform an update operation. The coupling calculation unit obtains the latest technical data, and recalculates the technical association strength of the corresponding area based on the latest technical data and the optimization area. Generating the updated technical association strength matrix includes:

[0101] The map optimization signal triggers the coupling calculation unit to perform an update operation. The coupling calculation unit uses the locality-sensitive hashing algorithm to divide the influence area range into multiple sub-areas, calculates the local density and relative distance of the technical nodes in the multiple sub-areas, identifies the technical nodes, and constructs an influence propagation network centered on the technical nodes;

[0102] The coupling calculation unit obtains the latest technical data from the multi-source technical database, performs semantic analysis on the latest technical data using a deep learning model, extracts technical feature vectors including technical function attributes, application scenario features, and innovation point descriptions, and establishes a time-series feature extraction module to capture dynamic trend information of technological development; the coupling calculation unit calculates the technical association strength within the influence area based on the technical feature vectors and the dynamic trend information;

[0103] The coupling calculation unit recalculates the technical association strength within the influence area using a multi-level coupling calculation method, and the coupling calculation unit generates an updated technical association strength matrix based on the calculation results of the multi-level coupling calculation method.

[0104] First, the system receives a graph optimization signal, triggering the coupling calculation unit to perform an update operation. For example, the system can generate a graph optimization signal regularly (e.g., daily or weekly), or trigger the update operation according to a user request or a specific event.

[0105] When the coupling calculation unit receives the graph optimization signal, it first determines the influence area range that needs to be updated. This influence area can be the entire technology graph or a specific part of the graph, such as a specific technology field or technology cluster. For example, if it is detected that the technology in the field of artificial intelligence is developing rapidly, the field of artificial intelligence can be set as the influence area range.

[0106] Next, the coupling calculation unit uses the locality-sensitive hashing algorithm to divide the influence area range into multiple sub-areas. For example, the influence area can be divided into different sub-areas according to the application scenario or function attribute of the technology. Assuming the influence area is the field of artificial intelligence, it can be divided into sub-areas such as computer vision, natural language processing, and machine learning.

[0107] Then, the coupling calculation unit calculates the local density and relative distance of the technical nodes in each sub-area. The local density refers to the number of other technical nodes related to a certain technical node around it. The relative distance refers to the semantic similarity or functional correlation between two technical nodes. By calculating the local density and relative distance, representative technical nodes in each sub-area can be identified. For example, in the computer vision sub-area, "convolutional neural network" may be a representative technical node.

[0108] Centered on these representative technical nodes, the coupling calculation unit constructs an influence propagation network. The influence propagation network describes the mutual influence and association relationships between different technical nodes. For example, "convolutional neural network" may be associated with other technical nodes such as "image recognition" and "object detection".

[0109] Meanwhile, the coupling calculation unit obtains the latest technology data from multi-source technology databases (such as patent databases, scientific and technological literature databases, news information websites, etc.). These data can include technical papers, patent specifications, news reports, etc.

[0110] After obtaining the latest technology data, the coupling calculation unit uses deep learning models (such as BERT, GPT, etc.) to perform semantic analysis on this data and extract technical feature vectors containing technical function attributes, application scenario features, and innovation point descriptions. For example, for the "automatic driving" technology, the extracted technical feature vectors may include features such as "vehicle control", "environmental perception", "path planning", etc. Suppose the latest technology data contains a paper on "deep learning-based automatic driving technology", then technical features such as "deep learning", "vehicle control", "environmental perception", etc. can be extracted.

[0111] In addition, the coupling calculation unit also establishes a time-series feature extraction module to capture the dynamic trend information of technological development. For example, indicators such as the appearance frequency and citation times of a certain technology within a period of time can be counted to reflect the development trend of this technology. Suppose the number of mentions of the "deep learning" technology has increased significantly in the past year, then it can be regarded as a technological development trend.

[0112] Based on the extracted technical feature vectors and dynamic trend information, the coupling calculation unit uses a multi-level coupling calculation method to recalculate the technological association strength within the influence area. The multi-level coupling calculation method can consider factors at different levels, such as semantic similarity, functional relevance, development trend, etc. between technical nodes.

[0113] Finally, the coupling calculation unit generates an updated technological association strength matrix based on the calculation results of the multi-level coupling calculation method. This matrix reflects the association strength between different technical nodes. For example, if the association strength between "deep learning" and "automatic driving" is very high, it indicates that there is a close connection between these two technologies.

[0114] The solution of this application can:

[0115] High degree of automation: The present invention can automatically obtain the latest technology data from multi-source technology databases and use deep learning models for semantic analysis, thereby realizing the automatic update of technological association strength, greatly reducing the cost and workload of manual intervention. High accuracy: The present invention uses deep learning models to perform semantic analysis on technical data, which can more accurately capture the association relationship between technologies, thereby improving the accuracy of calculating technological association strength. Strong dynamics: The present invention can capture the dynamic trend information of technological development and incorporate it into the calculation of technological association strength, so that the technology map can timely reflect the latest dynamics of technological development.

[0116] In an alternative embodiment, the optimized technology iteration path diagram is used to recalculate the technology potential distribution field; the updated technology association strength matrix, the technology iteration path diagram, and the technology potential distribution field are integrated into a new knowledge graph state, and the updated state feature set output includes:

[0117] Extract the path node degree centrality, path betweenness centrality, and path cluster cohesion in the optimized technology iteration path diagram to construct a technology diffusion matrix, which records the propagation intensity and coverage of technical knowledge in the path network;

[0118] Recalculate the technology potential distribution field based on the path node degree centrality, the path betweenness centrality, the path cluster cohesion, and the technology diffusion matrix. The technology potential distribution field includes basic potential, position potential, and development potential. The basic potential is calculated based on the technological innovation level, the position potential is calculated based on the network structure status, and the development potential is calculated based on future evolution trends;

[0119] Perform data standardization processing on the updated technology association strength matrix, convert the technology iteration path diagram into a standard network representation, convert the technology potential distribution field into a unified digital format, and construct a multi-dimensional data integration framework;

[0120] Establish vertical mapping relationships and horizontal connection relationships in the multi-dimensional data integration framework. The vertical mapping relationships determine the corresponding methods of data at different levels, and the horizontal connection relationships determine the interaction methods of data at the same level. Extract the common features of the multi-dimensional data based on the tensor decomposition method; input the common features into the feature alignment module. The feature alignment module unifies the expression methods of different dimensions, uses an adaptive weight assignment algorithm to determine the importance of features in each dimension, and integrates the aligned features into a new knowledge graph state;

[0121] Extract static features and dynamic features based on the new knowledge graph state. The static features include structural features, distribution features, and attribute features, and the dynamic features include evolution features, flow features, and interaction features; input the static features and the dynamic features into a feature fusion device. The feature fusion device uses a multi-layer perceptron for feature combination and realizes feature dimensionality reduction through the principal component analysis method to generate an updated state feature set.

[0122] First, obtain the technology iteration path diagram. The technology iteration path diagram describes the evolutionary relationships and development sequences among different technologies. For example, a technology iteration path diagram can be constructed based on patent data, literature data, industry reports, etc., where nodes represent a technology and edges represent the evolutionary relationships between technologies. A simplified case: Suppose there are three technologies A, B, and C, where A evolves into B and B evolves into C, then the path is A->B->C.

[0123] Then, calculate the technology correlation strength matrix. This matrix reflects the degree of correlation among different technologies. For example, the technology correlation strength can be calculated based on technology co-occurrence, citation relationships, semantic similarity, etc. Case: Suppose the correlation strength between A and B is 0.8, the correlation strength between B and C is 0.7, and the correlation strength between A and C is 0.5, then the matrix can be represented as [[1,0.8,0.5],[0.8,1,0.7],[0.5,0.7,1]].

[0124] Next, calculate the technology potential distribution field. The technology potential distribution field includes basic potential, position potential, and development potential. The basic potential is calculated based on the technology innovation level, such as the number of patents, citation times, etc. The position potential is calculated based on the network structure status, such as the degree centrality, betweenness centrality of nodes, etc. The development potential is calculated based on future evolutionary trends, such as the technology development speed, market demand, etc. Case: The basic potential of technology A is 10, the position potential is 5, and the development potential is 8; the basic potential of technology B is 8, the position potential is 7, and the development potential is 9; the basic potential of technology C is 6, the position potential is 9, and the development potential is 7.

[0125] According to the optimized technology iteration path diagram, recalculate the technology potential distribution field. For example, considering the technology development trend and market demand changes, adjust the development potential of technology C to 10.

[0126] Extract the path node degree centrality, path betweenness centrality, and path cluster cohesion in the technology iteration path diagram. The node degree centrality represents the number of edges connected to the node, the betweenness centrality represents the number of times the node appears on the shortest path in the network, and the cluster cohesion represents the tightness of the group to which the node belongs. Case: In the A->B->C path, the betweenness centrality of B is the highest.

[0127] Construct the technology diffusion matrix. The technology diffusion matrix records the propagation strength and coverage of technical knowledge in the path network. For example, the technology diffusion matrix can be calculated based on the path length, node centrality, etc. Case: Suppose the influence of A on B is 0.8 and the influence of B on C is 0.7, then the diffusion matrix can be represented as [[1,0.8,0.56],[0,1,0.7],[0,0,1]] (0.56 = 0.8 * 0.7).

[0128] Recalculate the technical potential distribution field based on the above metrics. For example, multiply the positional potential and the development potential by the degree centrality of the corresponding nodes as the new positional potential and development potential.

[0129] Perform data standardization on the updated technical association strength matrix. For example, scale all values to between 0 and 1. Convert the technical iteration path graph into a standard network representation, such as an adjacency matrix. Convert the technical potential distribution field into a unified digital format.

[0130] Construct a multi-dimensional data integration framework. Establish vertical mapping relationships and horizontal connection relationships in the framework. For example, use the technical association strength matrix, the technical iteration path graph, and the technical potential distribution field as different dimensions of the framework respectively. The vertical mapping relationship determines the corresponding method of data at different levels. For example, associate data of different dimensions of the same technology. The horizontal connection relationship determines the interaction method of data at the same level. For example, the association strength between different technologies.

[0131] Extract the common features of multi-dimensional data based on the tensor decomposition method. Input the common features into the feature alignment module to unify the expression forms of different dimensions. Use an adaptive weight assignment algorithm to determine the importance of features in each dimension. Integrate the aligned features into a new knowledge graph state.

[0132] Extract static features and dynamic features based on the new knowledge graph state. Static features include structural features (such as network density), distribution features (such as potential distribution), and attribute features (such as technology categories). Dynamic features include evolutionary features (such as the technology development speed), flow features (such as the technology diffusion speed), and interaction features (such as technology cooperation relationships).

[0133] Input the static features and dynamic features into a feature fusion device. For example, use a multi-layer perceptron for feature combination. Implement feature dimensionality reduction through the principal component analysis method to generate an updated state feature set.

[0134] The solution of this application can:

[0135] Improve the accuracy of technology prediction: By integrating multi-dimensional data and constructing a more comprehensive knowledge graph state, the present invention can capture the laws of technology development more accurately, thereby improving the accuracy of technology prediction. Enhance the interpretability of technology prediction: By analyzing the technical iteration path and the potential distribution field, the present invention can reveal the internal driving factors and evolution mechanisms of technology development, enhancing the interpretability of technology prediction. Support technology planning and decision-making: The technology prediction results provided by the present invention can provide a scientific basis for technology planning, investment decision-making, etc., helping enterprises and governments better grasp the direction of technology development.

[0136] In an alternative embodiment, the technical requirements of the user are converted into a query state, and the state similarity is calculated with the updated state feature set to obtain the state similarity; based on the state similarity, the optimal technical unit combination is identified; using the technical iteration path graph to predict the state evolution trajectory of the optimal technical unit combination includes:

[0137] Receive the user's technical requirements, use a deep semantic analysis model to extract the technical field features, functional requirement features, and performance index features in the user's technical requirements, and construct a feature representation matrix, which quantitatively describes the multi-dimensional attributes of the requirements;

[0138] Based on the feature representation matrix, construct a query state vector through a feature mapping network. The feature mapping network includes a static feature encoding layer and a dynamic feature encoding layer. The static feature encoding layer processes technical attributes and functional attributes, and the dynamic feature encoding layer processes temporal features and evolution features;

[0139] Calculate the similarity between the query state vector and the updated state feature set, use the attention mechanism to assign adaptive weights to different feature dimensions, and calculate the distance between state vectors through a deep metric learning model to generate a state similarity matrix;

[0140] Use the state similarity matrix to construct a technical unit affinity graph. The nodes of the technical unit affinity graph represent technical units, and the edge weights represent state similarity. Use a community discovery algorithm to identify high-cohesion technical unit combinations, and screen the optimal technical unit combination based on a combination scoring function;

[0141] Conduct a historical evolution analysis on the optimal technical unit combination, and extract an evolution feature sequence based on the technical iteration path graph. The evolution feature sequence includes changes in technical maturity, innovation activity, and association strength;

[0142] Input the evolution feature sequence into a temporal prediction model. The temporal prediction model uses a long short-term memory network structure, captures long-term dependencies through a gating mechanism, and combines the attention mechanism to identify temporal patterns to generate a technical evolution prediction sequence;

[0143] Construct a multi-dimensional evolution path graph based on the technical evolution prediction sequence. The nodes of the multi-dimensional evolution path graph represent technical states, and the edges represent state transition probabilities. Optimize the path selection strategy through the Monte Carlo tree search algorithm, and output the optimal state evolution trajectory.

[0144] First, the system receives the technical requirement text submitted by the user. For example, the requirement submitted by the user is "Develop a voice recognition system for smart home, with a recognition accuracy of over 95% and a response speed of less than 1 second."

[0145] Next, the system uses a deep semantic analysis model to extract the features in the user's technical requirements. Taking "develop a voice recognition system for smart home, with the requirement that the recognition accuracy rate reaches over 95% and the response speed is less than 1 second" as an example, the technical field features extracted by the system are "voice recognition" and "smart home", the functional requirement features are "voice recognition" and "control smart home devices", and the performance index features are "recognition accuracy rate > 95%" and "response speed < 1s". Then, the system constructs these features into a feature representation matrix. Assuming that this matrix contains four dimensions: technical field, functional requirement, recognition accuracy rate, and response speed. Then the feature representation matrix of this requirement can be expressed as: [voice recognition, smart home, voice recognition, control smart home devices, 0.95, 1], where 0.95 and 1 represent the numerical values of the recognition accuracy rate and the response speed respectively.

[0146] Then, the system inputs the feature representation matrix into the feature mapping network to generate a query state vector. The feature mapping network consists of a static feature encoding layer and a dynamic feature encoding layer. The static feature encoding layer processes technical attributes and functional attributes, for example, encoding "voice recognition" and "smart home" into specific vector representations. The dynamic feature encoding layer processes temporal features and evolutionary features. Since the current requirement does not have clear temporal and evolutionary features, the dynamic feature encoding layer can supplement according to the current technological development trend, such as the latest progress of voice recognition technology and the future development direction of the smart home market. Assuming that the static feature encoding layer encodes "voice recognition" as [0.1, 0.2, 0.7], "smart home" as [0.8, 0.1, 0.1], "voice recognition" as [0.1, 0.2, 0.7], and "control smart home devices" as [0.5, 0.3, 0.2]. The dynamic feature encoding layer supplements two dimensions according to the current technological development trend, such as the market growth rate and technological maturity, encoded as [0.9, 0.8]. The finally generated query state vector is [0.1, 0.2, 0.7, 0.8, 0.1, 0.1, 0.1, 0.2, 0.7, 0.5, 0.3, 0.2, 0.9, 0.8].

[0147] The system calculates the similarity between the generated query state vector and the updated state feature set. Assuming that the state feature set contains the state vectors of multiple technical units, the system uses an attention mechanism to assign adaptive weights to different feature dimensions, and calculates the distance between the query state vector and the state vector of each technical unit through a deep metric learning model, and finally generates a state similarity matrix. For example, assuming that there are two technical units A and B in the state feature set, and their state vectors are [0.1, 0.3, 0.6,...] and [0.2, 0.2, 0.6,...] respectively, the calculated state similarity matrix is [[1, 0.8], [0.8, 1]], where the diagonal elements are 1, indicating that the similarity of itself to itself is 1, and other elements represent the similarity between different technical units.

[0148] The system constructs a technical unit affinity graph using the state similarity matrix. The nodes in the graph represent technical units, and the edge weights represent state similarities. For example, according to the above state similarity matrix, an affinity graph containing two nodes A and B can be constructed, and the edge weight between nodes A and B is 0.8. Then, the system uses a community discovery algorithm to identify highly cohesive combinations of technical units. In this example, since there are only two technical units, they form a combination. The system filters the optimal combination of technical units based on a combination scoring function. Assuming that the scoring function considers factors such as the performance, cost, and reliability of technical units, the combination of A and B is finally selected as the optimal combination of technical units.

[0149] The system conducts a historical evolution analysis on the optimal combination of technical units, extracts an evolution feature sequence based on the technology iteration path graph, such as changes in technology maturity, innovation activity, and association strength. Assuming that the historical evolution data of technical units A and B shows that their technology maturity is continuously improving, the innovation activity remains stable, and the association strength is gradually increasing.

[0150] The system inputs the evolution feature sequence into a time series prediction model, such as a long short-term memory network, to predict the technology evolution trend. Assuming that the prediction results show that the technology maturity of technical units A and B will continue to improve, the innovation activity will increase, and the association strength will further increase.

[0151] Finally, the system constructs a multi-dimensional evolution path graph based on the technology evolution prediction sequence, and optimizes the path selection strategy through the Monte Carlo tree search algorithm to output the optimal state evolution trajectory. For example, the prediction results may show that technical units A and B will fuse into a new technical unit C, or they will develop into more advanced technical units A' and B' respectively.

[0152] The solution of this application can:

[0153] Improve the accuracy of technology prediction: Through deep semantic analysis and deep metric learning, understand user needs more accurately, and find matching technology units, thereby improving the accuracy of technology prediction. Optimize the technology combination plan: Through state similarity calculation and community discovery algorithms, identify highly cohesive technology unit combinations, and screen the optimal combinations based on the combination scoring function, thereby optimizing the technology combination plan. Predict the technology evolution trajectory: Through historical evolution analysis and time series prediction models, predict the future development trends of technology units, and construct a multi-dimensional evolution path map to provide reference for technology planning and decision-making.

[0154] In an alternative embodiment, the regional adaptability of technology implementation is evaluated in combination with the technology potential distribution field to generate a technology implementation plan, and the feedback information of the user on the technology implementation plan is recorded; the feedback information is converted into a state correction factor for adjusting the dynamic threshold, and the dynamic optimization of the new knowledge graph state is realized, including:

[0155] Evaluate the regional adaptability of technology implementation based on the technology potential distribution field, and generate a technology implementation plan according to the regional adaptability. The technology implementation plan includes a technology path planning and resource allocation plan formulated based on the regional adaptability;

[0156] Adopt a multi-modal feedback acquisition system to record the implementation effect feedback of the user on the technology implementation plan. The multi-modal feedback acquisition system obtains a multi-dimensional feature feedback data set, and the multi-dimensional feature feedback data set includes text feedback data, voice feedback data and behavior feedback data; input the multi-dimensional feature feedback data set into a deep semantic understanding network, and the deep semantic understanding network extracts semantic features and emotional features to generate a feedback feature vector;

[0157] Construct a state correction factor based on the feedback feature vector. The state correction factor includes a technology implementation effect score and an optimization direction indication; input the state correction factor into a dynamic threshold adjustment module, and the dynamic threshold adjustment module updates the evolution rule trigger threshold of the knowledge graph state according to the technology implementation effect score and the optimization direction indication;

[0158] Adopt a graph structure optimization network to realize the dynamic optimization of the knowledge graph state. The graph structure optimization network determines the rule trigger conditions according to the dynamic threshold adjustment module, and applies the rules that meet the trigger conditions to the structure update of the knowledge graph; construct a multi-objective optimization framework, and the multi-objective optimization framework uses the state correction factor as an optimization constraint to guide the evolution direction of the knowledge graph state;

[0159] Generate a new knowledge graph state based on the optimization results of the multi-objective optimization framework, and feedback the optimization effect of the new knowledge graph state to the multi-modal feedback acquisition system to form a continuously optimized closed-loop mechanism.

[0160] First, evaluate the regional adaptability of technology implementation. For example, assume that a knowledge graph about "smart home" is to be constructed. It is necessary to evaluate the applicability of different technologies (such as natural language processing, knowledge representation learning, etc.) in different regions (such as smart home device control, home environment monitoring, etc.). The adaptability scores of each technology in each region can be obtained through methods such as expert scoring and data analysis, forming a technology potential distribution field. For example, the adaptability score of natural language processing technology in the smart home device control region is 0.9, and the adaptability score in the home environment monitoring region is 0.7.

[0161] Generate a technology implementation plan according to the regional adaptability. The technology implementation plan includes technology path planning and resource allocation plan. For example, according to the above adaptability scores, it is determined to give priority to using natural language processing technology in the smart home device control region and allocate corresponding computing resources and human resources.

[0162] Next, use a multi-modal feedback collection system to record user feedback. This system can collect users' text feedback (such as comments when users use the smart home APP), voice feedback (such as instructions when users control devices through voice assistants), and behavior feedback (such as users' usage habits of smart home devices). For example, if the user feedback is "I hope I can control the light color by voice", then this feedback will be recorded.

[0163] Input the multi-dimensional feature feedback data set into the deep semantic understanding network. This network can extract semantic features and emotional features and generate feedback feature vectors. For example, the semantic features of the above user feedback can be expressed as "voice control", "light", "color", and the emotional feature can be expressed as "expectation".

[0164] Construct a state correction factor based on the feedback feature vector. The state correction factor includes a technology implementation effect score and an optimization direction indication. For example, according to the above feedback feature vector, it can be judged that users have a strong demand for the function of voice control of light color. Therefore, the technology implementation effect score is relatively high, and the optimization direction indication is to strengthen the voice control function.

[0165] Input the state correction factor into the dynamic threshold adjustment module. This module updates the triggering threshold of the evolution rule of the knowledge graph state according to the technology implementation effect score and the optimization direction indication. For example, if the technology implementation effect score is relatively high, then lower the triggering threshold of the corresponding rule to make it easier to be triggered, thereby accelerating the evolution speed of the knowledge graph.

[0166] Implement dynamic optimization of the knowledge graph state using a graph structure optimization network. This network determines the rule trigger conditions according to the dynamic threshold adjustment module and applies the rules that meet the trigger conditions to the structure update of the knowledge graph. For example, if the trigger condition of the "add new entity" rule is met, "light color" will be added to the knowledge graph.

[0167] Construct a multi-objective optimization framework, use the state correction factor as an optimization constraint to guide the evolution direction of the knowledge graph state. For example, use the optimization direction indication feedback by users as one of the optimization objectives to ensure that the evolution direction of the knowledge graph is consistent with user needs.

[0168] Finally, generate a new knowledge graph state based on the optimization results of the multi-objective optimization framework, and feedback the optimization effect of the new knowledge graph state to the multi-modal feedback acquisition system to form a continuous optimization closed-loop mechanism. For example, apply the updated knowledge graph to the smart home system and collect users' usage feedback on the new functions for further optimizing the knowledge graph.

[0169] The solution of this application can:

[0170] Improve the quality of the knowledge graph: By driving the dynamic evolution of the knowledge graph through user feedback, the content and structure of the knowledge graph can be continuously improved to make it more accurate, more complete, and closer to user needs. Enhance the user experience: By continuously optimizing the knowledge graph, the performance and user experience of application systems such as smart homes can be improved. For example, more accurate voice control, more intelligent scene recommendations, etc. Realize personalized services: By analyzing user feedback, the personalized needs of users can be understood, and thus more personalized services can be provided. For example, customize smart home scenarios according to users' living habits.

[0171] Figure 2 This is a schematic structural diagram of the intelligent matching and trading recommendation system for scientific and technological achievements based on the knowledge graph according to the embodiments of the present invention, as Figure 2 shown, the system includes:

[0172] The first unit is used to obtain a scientific and technological achievement data set, extract technical gene sequences from the scientific and technological achievement data set through a deep feature extractor; input the technical gene sequences into a high-dimensional coding network to generate technical state vectors; construct a technical association strength matrix based on the technical state vectors, and the technical association strength matrix records the coupling degree between technical units; calculate a technical evolution coefficient according to the coupling degree, generate a technical iteration path graph; use the technical iteration path graph to quantitatively calculate the innovation potential values of each technical unit to form a technical potential distribution field; integrate the technical association strength matrix, the technical iteration path graph, and the technical potential distribution field into an initial knowledge graph, and output the state feature set of the initial knowledge graph;

[0173] A second unit for deploying intelligent sensors to monitor the fluctuations of the state feature set in real time; when the monitoring result of the state fluctuation set exceeds a preset dynamic threshold, triggering a graph optimization signal; according to the graph optimization signal, calling a coupling calculation unit to recalculate the technical association strength matrix; based on the updated technical association strength matrix, using a dynamic optimization algorithm to optimize the technical iteration path graph; using the optimized technical iteration path graph to recalculate the technical potential distribution field; integrating the updated technical association strength matrix, the technical iteration path graph, and the technical potential distribution field into a new knowledge graph state, and outputting an updated state feature set;

[0174] A third unit for converting the user's technical requirements into a query state, calculating the state similarity with the updated state feature set to obtain the state similarity; identifying an optimal technical unit combination based on the state similarity; predicting the state evolution trajectory of the optimal technical unit combination using the technical iteration path graph; evaluating the regional adaptability of technology implementation in combination with the technical potential distribution field to generate a technology implementation plan, and recording the user's feedback information on the technology implementation plan; converting the feedback information into a state correction factor for adjusting the dynamic threshold to achieve dynamic optimization of the new knowledge graph state.

[0175] In the third aspect of the embodiments of the present invention,

[0176] Provided is an electronic device, including:

[0177] A processor;

[0178] A memory for storing instructions executable by the processor;

[0179] Wherein, the processor is configured to call the instructions stored in the memory to execute the method described above.

[0180] In the fourth aspect of the embodiments of the present invention,

[0181] Provided is a computer-readable storage medium, on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the method described above is implemented.

[0182] The present invention can be a method, device, system, and / or computer program product. The computer program product may include a computer-readable storage medium, on which computer-readable program instructions for executing various aspects of the present invention are uploaded.

[0183] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A method for intelligent matching and transaction recommendation of scientific and technological achievements based on knowledge graph, characterized in that: include: Acquire a scientific and technological achievement data set, and extract a technical gene sequence from the scientific and technological achievement data set by using a deep feature extractor; Inputting the technology gene sequence into a high-dimensional coding network to generate a technology state vector; constructing a technology association strength matrix based on the technology state vector, wherein the technology association strength matrix records the coupling degree between technology units; Calculating the technology evolution coefficient according to the coupling degree to generate a technology iteration path map; using the technology iteration path map to quantitatively calculate the innovation potential energy value of each technology unit to form a technology potential energy distribution field; Integrate the technology association intensity matrix, the technology iteration path map, and the technology potential energy distribution field into an initial knowledge graph, and output a state feature set of the initial knowledge graph; Deploy intelligent sensors to monitor the fluctuation of the state feature set in real time; when the monitoring result of the state fluctuation set exceeds a preset dynamic threshold, trigger a graph optimization signal; According to the graph optimization signal, calling the coupling calculation unit to recalculate the technology association intensity matrix; Based on the updated technology association intensity matrix, the technology iteration path map is optimized by using a dynamic optimization algorithm; and the technology potential energy distribution field is recalculated by using the optimized technology iteration path map; Integrate the updated technology association intensity matrix, the technology iteration path map, and the technology potential energy distribution field into a new knowledge graph state, and output an updated state feature set; The user's technical requirements are converted into query states, and the state similarity is calculated with the updated state feature set to obtain the state similarity; the optimal technology unit combination is identified based on the state similarity; the state evolution trajectory of the optimal technology unit combination is predicted using the technology iteration path diagram; the regional adaptability of the technology implementation is evaluated in combination with the technology potential energy distribution field, a technology implementation plan is generated, and the user's feedback information on the technology implementation plan is recorded; the feedback information is converted into a state correction factor for adjusting the dynamic threshold to achieve dynamic optimization of the new knowledge graph state.

2. The method according to claim 1, characterized in that Calculating the technology evolution coefficient according to the coupling degree to generate a technology iteration path map; using the technology iteration path map to quantitatively calculate the innovation potential energy value of each technology unit to form a technology potential energy distribution field; Integrating the technology association intensity matrix, the technology iteration path map, and the technology potential energy distribution field into an initial knowledge graph, and outputting a state feature set of the initial knowledge graph includes: The coupling degree is input into a technology evolution prediction model, and the technology evolution prediction model calculates the evolution coefficient of the technology unit based on the dominance ratio, the correlation strength ratio and the technology maturity difference between the technology units; the evolution coefficient is combined with the technology development constraint conditions, and the technology development constraint conditions include the timing restriction of technology development, the technology generation span restriction and the technology branch convergence restriction; a technology iteration path map is generated based on the evolution coefficient and the technology development constraint conditions, and the edge weight of the technology iteration path map is determined by the evolution coefficient ratio of the source technology unit and the target technology unit; The innovation potential energy value of the technology unit is calculated by using the technology iteration path map, the innovation potential energy value is mapped to the geographic space coordinate system, and the technology potential energy distribution field is constructed based on the spatial aggregation effect, regional correlation and regional technology foundation; The technology association intensity matrix, the technology iteration path map and the technology potential energy distribution field are tensor-fused to construct a multi-dimensional associated initial knowledge graph; the initial knowledge graph includes a topological structure layer, a technology evolution layer and a potential energy distribution layer of a technology unit; Perform feature extraction on the initial knowledge graph, extract the association features of the topological structure layer based on the graph neural network, extract the temporal features of the technology evolution layer based on the recursive neural network, and extract the spatial features of the potential energy distribution layer based on the convolutional neural network; fuse the association features, the temporal features and the spatial features to form a state feature set of the knowledge graph.

3. The method according to claim 1, characterized in that Deploy intelligent sensors to monitor the fluctuation of the state feature set in real time; when the monitoring result of the state fluctuation set exceeds a preset dynamic threshold, trigger a graph optimization signal; According to the graph optimization signal, calling the coupling calculation unit to recalculate the technology association intensity matrix; Based on the updated technology association intensity matrix, optimizing the technology iteration path map using a dynamic optimization algorithm includes: Deploy intelligent sensors to monitor the fluctuation of the state feature set, wherein the intelligent sensors collect technical node feature data, technical association feature data and graph global feature data, calculate feature deviation based on the technical node feature data, calculate coupling change rate based on the technical association feature data, and calculate situation evolution rate based on the graph global feature data; The feature deviation, the coupling change rate and the situation evolution rate are input into a feature fusion model, the feature fusion model uses a weighted combination method to generate a state feature fluctuation index, and the state feature fluctuation index represents the real-time situation of the knowledge graph; Constructing a dynamic threshold calculation model, wherein the dynamic threshold calculation model establishes a benchmark threshold based on the historical data distribution of the state characteristic fluctuation index, and dynamically adjusts the benchmark threshold according to the technology development cycle function and the environmental impact factor to obtain an adaptive dynamic threshold; The state feature fluctuation index is compared with the adaptive dynamic threshold in real time, and when the state feature fluctuation index exceeds the adaptive dynamic threshold, the influence degree of the fluctuation occurrence area is calculated, and a graph optimization signal including an optimization area, an optimization depth and an optimization direction is generated based on the influence degree; The graph optimization signal triggers the coupling calculation unit to perform an update operation, the coupling calculation unit obtains the latest technical data, recalculates the technical association strength of the corresponding area based on the latest technical data and the optimized area, and generates an updated technical association strength matrix; The updated technology association intensity matrix is ​​input into the path optimization model. The path optimization model identifies the path nodes that need to be adjusted based on the optimization depth, constructs path evolution constraints according to the optimization direction, and uses a dynamic programming algorithm to obtain the optimal technology iteration path diagram based on the path nodes and the path evolution constraints.

4. The method according to claim 3, characterized in that The graph optimization signal triggers the coupling calculation unit to perform an update operation, the coupling calculation unit obtains the latest technical data, recalculates the technical association strength of the corresponding area based on the latest technical data and the optimized area, and generates an updated technical association strength matrix including: The graph optimization signal triggers the coupling calculation unit to perform an update operation. The coupling calculation unit uses a local sensitive hashing algorithm to divide the influence area into multiple sub-areas, calculates the local density and relative distance of the technology nodes in the multiple sub-areas, identifies the technology nodes, and constructs an influence propagation network with the technology nodes as the center; The coupling calculation unit obtains the latest technology data from a multi-source technology database, uses a deep learning model to perform semantic analysis on the latest technology data, extracts a technology feature vector containing technology function attributes, application scenario characteristics and innovation point descriptions, and establishes a time series feature extraction module to capture dynamic trend information of technology development; the coupling calculation unit calculates the technology association strength within the impact area based on the technology feature vector and the dynamic trend information; The coupling calculation unit recalculates the technical association strength of the impact area range by using a multi-level coupling calculation method, and generates an updated technical association strength matrix based on the calculation results of the multi-level coupling calculation method.

5. The method according to claim 1, characterized in that Recalculate the technology potential energy distribution field by using the optimized technology iteration path map; integrate the updated technology association intensity matrix, the technology iteration path map, and the technology potential energy distribution field into a new knowledge graph state, and output the updated state feature set including: Extracting the path node degree centrality, path betweenness centrality and path cluster cohesion in the optimized technology iteration path graph, and constructing a technology diffusion matrix, wherein the technology diffusion matrix records the diffusion intensity and coverage of technology knowledge in the path network; Recalculate the technology potential energy distribution field based on the path node degree centrality, the path betweenness centrality, the path cluster cohesion and the technology diffusion matrix. The technology potential energy distribution field includes basic potential energy, position potential energy and development potential energy. The basic potential energy is calculated based on the technology innovation level, the position potential energy is calculated based on the network structure status, and the development potential energy is calculated based on the future evolution trend. Performing data standardization on the updated technology association intensity matrix, converting the technology iteration path map into a standard network representation, converting the technology potential energy distribution field into a unified digital format, and constructing a multi-dimensional data integration framework; A vertical mapping relationship and a horizontal connection relationship are established in the multidimensional data integration framework, wherein the vertical mapping relationship determines the correspondence mode of data at different levels, and the horizontal connection relationship determines the interaction mode of data at the same level, and the common features of multidimensional data are extracted based on the tensor decomposition method; the common features are input into the feature alignment module, and the feature alignment module unifies the expression mode of different dimensions, adopts an adaptive weight allocation algorithm to determine the importance of features of each dimension, and integrates the aligned features into a new knowledge graph state; Based on the new knowledge graph state, static features and dynamic features are extracted, the static features include structural features, distribution features and attribute features, and the dynamic features include evolution features, flow features and interaction features; the static features and the dynamic features are input into a feature fusion device, and the feature fusion device uses a multi-layer perceptron to combine features, and implements feature dimensionality reduction through a principal component analysis method to generate an updated state feature set.

6. The method according to claim 1, characterized in that Convert the user's technical requirements into a query state, and perform state similarity calculation with the updated state feature set to obtain state similarity; and identify the optimal technical unit combination based on the state similarity; Predicting the state evolution trajectory of the optimal technology unit combination using the technology iteration path diagram includes: Receive user technical requirements, extract technical field features, functional requirement features and performance indicator features in the user technical requirements using a deep semantic analysis model, and construct a feature characterization matrix, which quantitatively describes the multi-dimensional attributes of the requirements; Based on the feature representation matrix, a query state vector is constructed through a feature mapping network, wherein the feature mapping network includes a static feature coding layer and a dynamic feature coding layer, wherein the static feature coding layer processes technical attributes and functional attributes, and the dynamic feature coding layer processes temporal features and evolutionary features; Calculate the similarity between the query state vector and the updated state feature set, use the attention mechanism to assign adaptive weights to different feature dimensions, calculate the distance between state vectors through a deep metric learning model, and generate a state similarity matrix; The state similarity matrix is ​​used to construct a technology unit affinity graph, wherein the nodes of the technology unit affinity graph represent technology units, and the edge weights represent state similarities. A community discovery algorithm is used to identify high-cohesion technology unit combinations, and an optimal technology unit combination is selected based on a combination scoring function. Performing a historical evolution analysis on the optimal technology unit combination, and extracting an evolution feature sequence based on the technology iteration path map, wherein the evolution feature sequence includes changes in technology maturity, changes in innovation activity, and changes in association intensity; Inputting the evolution feature sequence into a time series prediction model, the time series prediction model adopts a long short-term memory network structure, captures long-term dependencies through a gating mechanism, and combines an attention mechanism to identify time series patterns to generate a technology evolution prediction sequence; A multidimensional evolution path diagram is constructed based on the technology evolution prediction sequence. The nodes of the multidimensional evolution path diagram represent the technology status, and the edges represent the state transition probability. The path selection strategy is optimized through the Monte Carlo tree search algorithm to output the optimal evolution trajectory.

7. The method according to claim 1, characterized in that Evaluate the regional adaptability of technology implementation in combination with the technology potential energy distribution field, generate a technology implementation plan, and record the user's feedback information on the technology implementation plan; Converting the feedback information into a state correction factor for adjusting the dynamic threshold to achieve dynamic optimization of the new knowledge graph state includes: Evaluate the regional adaptability of technology implementation based on the technology potential distribution field, and generate a technology implementation plan based on the regional adaptability, wherein the technology implementation plan includes a technology path planning and a resource allocation plan based on the regional adaptability; A multimodal feedback collection system is used to record user feedback on the implementation effect of the technical implementation plan, and the multimodal feedback collection system obtains a multidimensional feature feedback data set, and the multidimensional feature feedback data set includes text feedback data, voice feedback data, and behavior feedback data; the multidimensional feature feedback data set is input into a deep semantic understanding network, and the deep semantic understanding network extracts semantic features and emotional features to generate a feedback feature vector; Constructing a state correction factor based on the feedback feature vector, wherein the state correction factor includes a technical implementation effect score and an optimization direction indication; inputting the state correction factor into a dynamic threshold adjustment module, wherein the dynamic threshold adjustment module updates the evolution rule trigger threshold of the knowledge graph state according to the technical implementation effect score and the optimization direction indication; A graph structure optimization network is used to realize dynamic optimization of the knowledge graph state. The graph structure optimization network determines the rule triggering conditions according to the dynamic threshold adjustment module, and applies the rules that meet the triggering conditions to the structural update of the knowledge graph; a multi-objective optimization framework is constructed. The multi-objective optimization framework uses the state correction factor as an optimization constraint to guide the evolution direction of the knowledge graph state; A new knowledge graph state is generated based on the optimization results of the multi-objective optimization framework, and the optimization effect of the new knowledge graph state is fed back to the multimodal feedback acquisition system to form a closed-loop mechanism for continuous optimization.

8. A scientific and technological achievement intelligent matching and transaction recommendation system based on knowledge graph, used to implement the method described in any one of claims 1 to 7, characterized in that: include: The first unit is used to obtain a scientific and technological achievement data set, and extract a technical gene sequence from the scientific and technological achievement data set through a deep feature extractor; Inputting the technology gene sequence into a high-dimensional coding network to generate a technology state vector; constructing a technology association strength matrix based on the technology state vector, wherein the technology association strength matrix records the coupling degree between technology units; Calculating the technology evolution coefficient according to the coupling degree to generate a technology iteration path map; using the technology iteration path map to quantitatively calculate the innovation potential energy value of each technology unit to form a technology potential energy distribution field; Integrate the technology association intensity matrix, the technology iteration path map, and the technology potential energy distribution field into an initial knowledge graph, and output a state feature set of the initial knowledge graph; The second unit is used to deploy intelligent sensors to monitor the fluctuation of the state feature set in real time; when the monitoring result of the state fluctuation set exceeds a preset dynamic threshold, a graph optimization signal is triggered; According to the graph optimization signal, calling the coupling calculation unit to recalculate the technology association intensity matrix; Based on the updated technology association intensity matrix, the technology iteration path map is optimized by using a dynamic optimization algorithm; and the technology potential energy distribution field is recalculated by using the optimized technology iteration path map; Integrate the updated technology association intensity matrix, the technology iteration path map, and the technology potential energy distribution field into a new knowledge graph state, and output an updated state feature set; The third unit is used to convert the user's technical requirements into a query state, calculate the state similarity with the updated state feature set, and obtain the state similarity; identify the optimal technology unit combination based on the state similarity; use the technology iteration path diagram to predict the state evolution trajectory of the optimal technology unit combination; combine the technology potential energy distribution field to evaluate the regional adaptability of the technology implementation, generate a technology implementation plan, and record the user's feedback information on the technology implementation plan; convert the feedback information into a state correction factor, which is used to adjust the dynamic threshold to achieve dynamic optimization of the new knowledge graph state.

9. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the method described in any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the computer program instructions are executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images