A port application technology evaluation method and system based on a knowledge graph

Abstract

The application discloses a port application technology evaluation method and system based on a knowledge graph, relates to the field of port technology evaluation, and comprises the following steps: matching a port technology knowledge graph with a port application technology to be evaluated to obtain corresponding target standard technology and a target evaluation index system; generating basic scores of each evaluation index in the target evaluation index system through knowledge graph reasoning, adjusting the basic scores based on the correlation between the evaluation indexes and benefits, and obtaining single scores of each evaluation index; determining dynamic weights of each evaluation index based on the correlation between the target standard technology and the evaluation indexes and the correlation between the target standard technology and an application scenario; and performing weighted calculation according to the single scores of each evaluation index and the dynamic weights to obtain a comprehensive score. The application solves the problems that in a traditional port evaluation method, evaluation factors lack correlation and coupling, and a static evaluation mode cannot adapt to dynamic change requirements of a port scene.

Description

Technical Field

[0001] This invention relates to the field of port technology assessment, and more particularly to a method and system for assessing port application technologies based on knowledge graphs. Background Technology

[0002] As a core hub of international trade, ports are accelerating their transformation towards intelligence and green development. Port application technology assessment is a key link to ensure scientific technology selection and efficient application.

[0003] Currently, the assessment of port application technologies still relies on manual experience scoring and independent indicator calculation. It has not built a port-specific knowledge system and related assessment rules, and lacks objective, unified and port-adaptable assessment standards. The indicators and benefits are isolated from each other and cannot achieve correlation reasoning, resulting in insufficient objectivity and industry adaptability in the assessment.

[0004] Meanwhile, existing assessment methods often adopt fixed weights and static assessment models, failing to dynamically adjust the assessment logic based on the differences in port operation scenarios. This makes them unsuitable for the diverse and dynamically changing port assessment needs, and difficult to support accurate assessment and scientific selection of technologies.

[0005] Therefore, there is an urgent need for a new customized port assessment method to solve the problems of lack of correlation and coupling among assessment factors in traditional port assessment methods and the inability of static assessment models to adapt to the dynamic changes in port scenarios. Summary of the Invention

[0006] The purpose of this invention is to provide a port application technology evaluation method and system based on knowledge graphs, which can solve the problems of lack of correlation and coupling among various evaluation factors and the inability of static evaluation models to adapt to the dynamic changes in port scenarios in traditional port evaluation methods.

[0007] To achieve the above objectives, the present invention provides the following technical solution: In a first aspect, the present invention provides a port application technology evaluation method based on knowledge graphs, comprising: Based on the port technology knowledge graph, the application technologies of the port to be evaluated are matched to obtain the target standard technologies and their target evaluation index system corresponding to the application technologies of the port to be evaluated; the target evaluation index system includes multiple evaluation indicators. The basic scores of each evaluation indicator in the target evaluation indicator system are generated by reasoning through knowledge graphs, and the basic scores are adjusted based on the relationship between the evaluation indicators and benefits to obtain the individual scores of each evaluation indicator. Based on the correlation between the target standard technology and the evaluation indicators, as well as the correlation between the target standard technology and the application scenario, the dynamic weight of each evaluation indicator is determined. The comprehensive score is obtained by weighting the individual scores of each evaluation indicator with their corresponding dynamic weights.

[0008] Optionally, basic scores for each evaluation indicator in the target evaluation indicator system can be generated through knowledge graph reasoning, including: Based on the subgraph matching reasoning algorithm, the feature subgraph of the port application technology to be evaluated is matched with the target standard technology subgraph pre-stored in the port technology knowledge graph to obtain the basic score of each evaluation indicator. The feature subgraph of the port application technology to be evaluated is constructed from the feature data of the port application technology to be evaluated according to the structure of the port technology knowledge graph. The target standard technology subgraph contains the evaluation indicator system and the standard features of each evaluation indicator.

[0009] Optionally, the base scores can be adjusted based on the correlation between the evaluation indicators and the benefits to obtain individual scores for each evaluation indicator, including: Based on the path reasoning algorithm, the implicit relationship between each evaluation indicator and the benefit is explored, and the basic score is corrected to obtain the corrected score. The fuzzy inference algorithm is used to process the uncertain feature data in the feature data of the port application technology to be evaluated, and the correction score is calibrated based on the processing results to output the individual score of each evaluation index.

[0010] Optionally, based on the correlation between standard technologies and evaluation indicators in the port technology knowledge graph, and the correlation between standard technologies and application scenarios, the dynamic weights of each evaluation indicator are determined, including: Based on the correlation between the port application technologies to be evaluated and the evaluation indicators, the entity correlation degree and path length between the port application technologies to be evaluated and the evaluation indicators are determined. The importance of a scenario is determined based on the correlation between the port application technologies and application scenarios to be evaluated. The dynamic weights of each evaluation indicator are calculated based on entity relevance, path length, and scenario importance.

[0011] Optionally, dynamic weights for each evaluation metric are calculated based on entity relevance, path length, and scenario importance, including: Substituting entity relevance, path length, and scene importance into the formula: ; The dynamic weights of each evaluation indicator are calculated; among them, Dynamic weights; For entity relevance; This is the sum of the entity correlations of all evaluation indicators; This is the path length; For scene importance; This is the sum of the importance of all scenarios; , and The weighting coefficients and .

[0012] Optionally, the method also includes: game calibration based on a large language model for multiple agents; the multiple agents include an efficiency-priority agent, a cost-control agent, and a safety-red-line agent; Multi-agent game calibration based on large language models includes: The intelligent agent system uses the individual scores and dynamic weights of the indicators that each intelligent agent is responsible for as the initial anchor values, and divides the feature subgraphs related to the evaluation dimensions of each intelligent agent from the port technology knowledge graph and assigns them to the corresponding intelligent agents; among them, the efficiency priority intelligent agent receives the individual scores and dynamic weights of the operation efficiency indicators, the cost control intelligent agent receives the individual scores and dynamic weights of the cost control indicators, and the safety red line intelligent agent receives the individual scores and dynamic weights of the safety compliance indicators. Each agent engages in multiple rounds of interactive debate based on its own set optimization goals, and conducts graph tracing and evidence retrieval based on its corresponding feature subgraph. If an agent disagrees with its own single-item score, it calls the port technology knowledge graph query interface based on the path nodes and relationships in the corresponding feature subgraph to extract relevant path nodes and evolutionary relationships as rebuttal evidence, and uses a dynamic scoring evolution mechanism to update its own quantitative score in real time. Based on the port multidimensional decision-making collaborative convergence algorithm, the global collaborative steady-state determination mechanism is used to monitor the scoring adjustment status of each agent in real time. When the global system fluctuation value of the agent system is less than the preset convergence threshold, the game is determined to have reached a collaborative steady state and the debate is terminated. The calibrated scores of the three evaluation indicators of efficiency, cost and safety after convergence are extracted.

[0013] Optionally, before matching the port application technologies to be evaluated based on the port technology knowledge graph, the method may also include: Obtain initial characteristic data of the port application technologies to be evaluated; the initial characteristic data includes technical parameter data, on-site operation data, cost input data, safety and compliance data, and scenario adaptation data; The initial feature data is preprocessed to obtain the feature data. The preprocessing includes adaptive denoising, data completion, outlier removal and normalization mapping operations performed sequentially.

[0014] Optionally, the adaptive noise reduction operation is achieved by using a filtering threshold that adapts to port machinery vibration, tidal cycle, and loading and unloading operation pulse interference.

[0015] Optionally, the data completion operation is implemented based on the time sequence association rules of the port operation process.

[0016] Compared with existing technologies, this invention provides a port application technology evaluation method based on knowledge graphs. By constructing a port technology knowledge graph, it achieves multi-element correlation and coupling of technology, evaluation indicators, application scenarios, and benefits. It completes the correlation evaluation of indicators and benefits through knowledge graph reasoning and determines dynamic weights through scenario adaptation. This completely gets rid of the traditional evaluation mode of manual scoring, independent indicator calculation, and static fixed weights. It fundamentally solves the problems of strong subjectivity, disconnect between indicators and benefits, lack of correlation reasoning, and poor scenario adaptability of traditional evaluation methods. While ensuring the objectivity and consistency of the evaluation, it realizes the dynamic adjustment of evaluation logic and weights according to port scenarios, which significantly improves the accuracy of evaluation and the adaptability and scientific nature of decision-making in the port industry.

[0017] Secondly, the present invention also provides a port application technology evaluation system based on knowledge graphs, comprising: The matching module is used to match the port application technologies to be evaluated based on the port technology knowledge graph, and obtain the target standard technologies and their target evaluation index systems corresponding to the port application technologies to be evaluated; the target evaluation index system includes multiple evaluation indicators. The scoring module is used to generate basic scores for each evaluation indicator in the target evaluation indicator system through knowledge graph reasoning, and to adjust the basic scores based on the correlation between the evaluation indicators and benefits to obtain individual scores for each evaluation indicator. The weight determination module is used to determine the dynamic weight of each evaluation indicator based on the relationship between the target standard technology and the evaluation indicators, as well as the relationship between the target standard technology and the application scenario. The weighting module is used to calculate the comprehensive score by weighting the individual scores of each evaluation indicator with their corresponding dynamic weights.

[0018] The beneficial effects of the knowledge graph-based port application technology evaluation system provided by this invention are similar to those of the aforementioned knowledge graph-based port application technology evaluation method, and will not be repeated here. Attached Figure Description

[0019] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this invention, illustrate exemplary embodiments of the invention and are used to explain the invention, but do not constitute an undue limitation of the invention. In the drawings: Figure 1 A flowchart illustrating a knowledge graph-based port application technology evaluation method provided as an embodiment of the present invention; Figure 2 A partial structural schematic diagram of a port technology knowledge graph provided as an embodiment of the present invention; Figure 3This is a schematic diagram of the structure of a port application technology evaluation system based on knowledge graph, provided as an embodiment of the present invention. Detailed Implementation

[0020] To facilitate a clear description of the technical solutions in the embodiments of the present invention, the terms "first" and "second" are used to distinguish identical or similar items with essentially the same function and effect. For example, the first threshold and the second threshold are merely used to distinguish different thresholds and do not limit their order. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and that the terms "first" and "second" are not necessarily different.

[0021] It should be noted that in this invention, the terms "exemplary" or "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplary" or "for example" in this invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0022] In this invention, "at least one" means one or more, and "more than one" means two or more. "And / or" describes the relationship between the associated objects, indicating that three relationships can exist.

[0023] like Figure 1 As shown, this embodiment of the invention provides a port application technology evaluation method based on knowledge graphs, which may include: Step S1: Based on the port technology knowledge graph, match the port application technologies to be evaluated to obtain the target standard technologies and their target evaluation index system corresponding to the port application technologies to be evaluated; the target evaluation index system includes multiple evaluation indicators. Understandably, the method also includes constructing a port technology knowledge graph before matching the port application technologies to be evaluated based on the port technology knowledge graph.

[0024] Constructing a port technology knowledge graph can specifically include: (1) Multi-source data collection: Collect multi-source data related to port application technology in all dimensions, including port technical solution documents, equipment operation and maintenance records, scheduling system business data, historical evaluation reports, industry standards and specifications, on-site monitoring data, etc., covering core dimensions such as technology, indicators, scenarios, and benefits.

[0025] (2) Entity and Relationship Extraction: Based on the predefined classification rules and association specifications of the port domain ontology, a port domain-fine-tuned BERT model is used to extract multiple types of entities, including technology entities, evaluation indicator entities, application scenario entities, and benefit entities. An attention-based relationship extraction model is used to mine the relationships between entities, including but not limited to technology-application scenario mapping relationships, technology-evaluation indicator adaptation relationships, evaluation indicator-benefit relationships, and technology-alternative technology evolution relationships. For example, the port domain ontology may include a port application technology classification system, an evaluation indicator definition system, a scenario adaptation rule system, and a technology evolution rule system. The classification system can be divided into 6 major categories and 23 subcategories according to technical functions. The evaluation indicator definition system clarifies the calculation methods, threshold ranges, and industry standards for each evaluation indicator.

[0026] (3) Entity and Relationship Integration: The extracted entities are ambiguously resolved, and conflicting relationships are verified and integrated using port industry knowledge rules to ensure the consistency of knowledge.

[0027] (4) Graph storage and incremental update: Neo4j graph database is used to store entity, relation and attribute information to build an initial port application technology exclusive knowledge graph; incremental update triggers are configured so that when new port application technologies, industry standards are updated or new evaluation data are added, the entity attributes and relation strength are automatically updated to ensure the timeliness and accuracy of the graph.

[0028] As can be seen from the above, the port technology knowledge graph constructed in this embodiment includes multiple entities and the relationships between them. These multiple entities include, but are not limited to, standard technology entities (hereinafter referred to as standard technologies), evaluation indicator entities (hereinafter referred to as evaluation indicators), application scenario entities (hereinafter referred to as application scenarios), and benefit entities (hereinafter referred to as benefits).

[0029] For example, refer to Figure 2 Standard technologies can include intelligent port machinery technologies, green and energy-saving technologies, and safety monitoring technologies. It is understood that standard technologies can also include other standard technology entities such as scheduling optimization technologies, cargo flow technologies, and emergency response technologies.

[0030] For example, refer to Figure 2 Evaluation metrics can include operational efficiency metrics, cost control metrics, and scenario adaptability metrics. It is understood that evaluation metrics can also include other evaluation metrics such as safety compliance metrics, technology maturity metrics, and ease of operation and maintenance metrics.

[0031] For example, refer to Figure 2Application scenarios can include container terminals, bulk cargo terminals, and port-adjacent logistics parks. It is understandable that application scenarios can also include other application-type entities such as oil and chemical terminals and roll-on / roll-off terminals.

[0032] For example, refer to Figure 2 Benefits can include reduced operating costs and reduced safety risks. Understandably, benefits can also include other tangible benefits such as increased operational efficiency, reduced energy consumption, and reduced labor input.

[0033] See Figure 2 The relationships include, but are not limited to: (1) the scenario mapping relationship between standard technologies and application scenarios: used to characterize port operation scenarios adapted to different standard technologies; (2) the indicator adaptation relationship between standard technologies and evaluation indicators: used to characterize the exclusive evaluation indicator system corresponding to different standard technologies; (3) the benefit relationship between evaluation indicators and benefits: used to characterize the port business benefit dimension corresponding to each evaluation indicator; (4) the weight relationship between application scenarios and evaluation indicators: used to characterize the dynamic weight adaptation rules of each evaluation indicator under different application scenarios; (5) the value mapping relationship between standard technologies and benefits: used to characterize the port business benefits that different standard technologies can achieve. The above relationships are only illustrative examples. This invention does not limit the relationships between entities in the knowledge graph and may include any reasonable relationships related to port technology evaluation.

[0034] Step S2: Generate basic scores for each evaluation indicator in the target evaluation indicator system through knowledge graph reasoning, and adjust the basic scores based on the correlation between the evaluation indicators and benefits to obtain individual scores for each evaluation indicator. Step S3: Based on the correlation between the target standard technology and the evaluation indicators, and the correlation between the target standard technology and the application scenario, determine the dynamic weight of each evaluation indicator; Step S4: Calculate the comprehensive score by weighting the individual scores of each evaluation indicator with their corresponding dynamic weights.

[0035] In this embodiment, in the initial stage of the assessment, technology matching is carried out based on a pre-constructed port technology knowledge graph. This enables the establishment of a port-specific, interconnected assessment system from the root, solving the problems of fragmented assessment factors, lack of unified standards, and excessive reliance on subjective human judgment in traditional assessments. This port technology knowledge graph integrates multiple entities, including standard technologies, assessment indicators, application scenarios, and benefits, and establishes relationships between these entities. It transforms port-specific technology types, assessment indicators, operational application scenarios, and business benefits into standardized assessment knowledge carriers, forming a port-specific and objectively unified assessment benchmark, completely eliminating reliance on human experience. Based on this knowledge graph, matching the port application technology to be assessed with entities within the graph accurately identifies the target standard technology corresponding to the port application technology being assessed. Then, through the pre-set relationships between standard technologies and assessment indicators in the graph, a target assessment indicator system adapted to the port application technology being assessed is automatically generated. This creates a strong correlation between technology and assessment indicators, breaking the independent accounting and fragmented state of indicators in traditional assessments and achieving initial interconnected coupling of assessment factors.

[0036] Next, in the indicator scoring stage, knowledge graph reasoning is used to complete score calculation and adjustment, enabling the correlation between indicators and benefits, further addressing the issues of subjective evaluation and the disconnect between indicators and benefits. The basic scores for each indicator in the target evaluation indicator system are generated using knowledge graph reasoning algorithms. The characteristics of standard technologies within the graph serve as objective evidence, and quantitative algorithms complete the score calculation without human intervention, replacing the traditional expert subjective scoring model and providing an objective technical basis for the scoring process. Simultaneously, based on the correlation between evaluation indicators and benefits in the knowledge graph, the basic scores are adaptively adjusted, linking the quantitative values ​​of the indicators to the actual business value of the port. This ensures that the scoring results align with the actual benefits brought by the technology application, achieving correlation reasoning between evaluation factors and completely resolving the one-sided results caused by traditional evaluations that only calculate indicator values ​​without considering business benefits.

[0037] In the weight determination stage, dynamic weights are determined based on the correlation between target standard technologies and evaluation indicators, as well as between target standard technologies and application scenarios. This breaks away from the traditional static evaluation model with fixed weights, solving the problem that evaluation cannot adapt to the dynamic changes in port scenarios. Traditional evaluations use unchanging weight settings, ignoring the differences in evaluation emphasis for different port scenarios such as container terminals, oil and chemical terminals, and bulk cargo terminals. This step, however, incorporates scenario adaptability into the weight calculation logic by using the correlation between target standard technologies and application scenarios in a knowledge graph. At the same time, it adjusts the weight ratio based on the closeness of the correlation between target standard technologies and evaluation indicators, allowing the evaluation weights to dynamically and adaptively adjust according to port operation scenarios and technology types. This completely replaces the traditional static weight model, making the evaluation logic adaptable to the multi-scenario and dynamically changing usage needs of ports.

[0038] In the comprehensive scoring stage, the individual scores of each evaluation indicator are weighted and calculated with their corresponding dynamic weights to obtain the comprehensive score. This completes the final implementation of the dynamic and correlated evaluation, forming a closed loop and thoroughly solving several shortcomings of traditional evaluation. This stage combines the individual scores obtained through correlation reasoning in the early stage with the dynamic weights adapted to the scenario. It not only continues the correlation and coupling logic of the evaluation factors throughout the entire process, but also implements the evaluation rules of dynamic adaptation to the scenario. The entire process is driven by the port's exclusive knowledge graph, without human intervention or fixed static rule constraints. Ultimately, it achieves objective, correlated, and dynamically adapted customized and accurate evaluation of the port, fully achieving the purpose of the invention.

[0039] Before matching the port application technologies to be evaluated based on the port technology knowledge graph, the method also includes: preprocessing the initial feature data to obtain feature data; the initial feature data includes at least equipment operation data, operation status data and environmental monitoring data; and the preprocessing includes adaptive denoising operation, data completion operation, outlier removal operation and standardization mapping operation performed in sequence.

[0040] Specifically, initial feature data and feature data can include, but are not limited to, technical parameter data, on-site operational data, cost input data, safety and compliance data, and scenario adaptation data. In port operation scenarios, the aforementioned feature data is susceptible to strong interference, obstruction, and intermittent transmission, resulting in non-random missing data, timing misalignments, and asynchronous multi-source heterogeneous data. Therefore, preprocessing must be based on the port's unique operational processes and equipment linkages, rather than using mathematical interpolation or filtering methods that are universally applicable to other scenarios.

[0041] For example, the adaptive noise reduction operation is implemented by using filtering thresholds that adapt to the vibration of port machinery, tidal cycles, and pulse interference from loading and unloading operations. Specifically, a 50Hz high-pass filter threshold is set for the vibration noise of port machinery operation, a 12-hour periodic smoothing filter threshold is set for tidal cycle interference, and a 0.5s pulse blanking filter threshold is set for pulse interference from loading and unloading operations.

[0042] In this embodiment, the adaptive denoising operation can accurately remove extreme interference caused by port-specific environmental noise (such as mechanical vibration, tidal cycle, loading and unloading operation pulses) in the port field operation data, effectively improving the purity and reliability of feature data, and laying a high-quality data foundation for subsequent technical matching evaluation.

[0043] For example, the data completion operation is implemented based on the time sequence association rules of port operation processes. Specifically, it is based on the standard operation time sequence chain of "quay crane hoisting → flatbed truck transfer → yard storage" in a container terminal, and completes the missing operation status data according to the association relationship between the preceding and following time nodes.

[0044] In this embodiment, the data completion operation can intelligently fill in missing data based on the temporal correlation characteristics of port operations, ensuring data integrity and business authenticity, avoiding data distortion problems caused by general filling methods, and making the processed data highly consistent with the actual operation rules of the port.

[0045] For example, outlier removal may specifically include: using an outlier detection algorithm based on port industry-specific thresholds to remove outlier data that does not conform to the reasonable range of the port industry.

[0046] For example, the standardization process specifically includes: using a standardization mapping algorithm to convert technical parameter data from different sources and with different dimensions into data with a unified dimension.

[0047] In an exemplary embodiment, step S1: Based on the port technology knowledge graph, the port application technology to be evaluated is matched to obtain the target standard technology and its target evaluation index system corresponding to the port application technology to be evaluated. The specific implementation process is as follows: the feature data of the port application technology to be evaluated is matched with the standard technology entities in the port technology knowledge graph, and the standard technology entity with the highest matching degree is taken as the target standard technology; based on the relationship between the target standard technology and the evaluation index, the matching path between the feature data of the target standard technology and the evaluation index is mined through the path matching algorithm, and the core evaluation index and auxiliary evaluation index are selected to form the target evaluation index system.

[0048] In one optional implementation, based on the correlation between the target standard technology and the evaluation indicators, a path matching algorithm is used to mine the matching paths between the feature data of the target standard technology and the evaluation indicators, and core evaluation indicators and auxiliary evaluation indicators are selected to form a target evaluation indicator system, including: Calculate the shortest path length in the graph between the target standard technology and each evaluation index; When the shortest path length is less than a preset threshold and there is a strong connection between the target standard technology and the core demand entity of the port, the evaluation index is anchored as the core evaluation index. When the shortest path length is greater than or equal to a preset threshold or when the target standard technology is only indirectly related to the evaluation indicator through secondary benefits, the evaluation indicator will be dynamically classified as an auxiliary evaluation indicator.

[0049] For example, core port requirements include efficiency, safety, and cost.

[0050] For example, the core evaluation indicators are decisive indicators, including safety compliance indicators, operational efficiency indicators, and cost control indicators.

[0051] For example, auxiliary evaluation indicators are supplementary indicators, including technology maturity indicators, ease of operation and maintenance indicators, and scenario adaptability indicators.

[0052] This embodiment uses a port technology knowledge graph to match the applied technologies and standard technologies of the port to be evaluated. Based on graph path calculation and association determination rules, it automatically divides core and auxiliary evaluation indicators, achieving a more refined indicator system construction on the basis of existing evaluation logic. It has the following advantages: This embodiment relies on the shortest path length and strong connectivity association determination of the graph to automatically stratify indicators, dividing evaluation indicators into decisive core indicators and supplementary auxiliary indicators, forming a differentiated indicator selection logic. This fills the gap in existing evaluations that lack stratified indicator definition and have coarse evaluation granularity. Simultaneously, it uses the core demand entities of the port as the core determination basis for indicator division, integrating port-specific operational demands such as efficiency, safety, and cost into the indicator selection rules. This differs from the indiscriminate selection of indicators in general evaluations, ensuring that the indicator system fundamentally aligns with the evaluation focus of the port industry. Furthermore, by using the direct and indirect association rules of target standard technology and evaluation indicators, as well as secondary benefits, the indicators are classified. The customized association logic of the port-specific knowledge graph is reused throughout the process, rather than the generalized association rules of the general domain. This enables the port-specific association rules to be implemented in the indicator system construction stage, allowing the application of knowledge graphs to extend from technical matching to the detailed stage of precise indicator selection, and further improving the complete logic of port customized evaluation.

[0053] In an exemplary embodiment, basic scores for each evaluation indicator in the target evaluation indicator system are generated through knowledge graph reasoning, including: Based on the subgraph matching reasoning algorithm, the feature subgraph of the port application technology to be evaluated is matched with the target standard technology subgraph pre-stored in the port technology knowledge graph to obtain the basic score of each evaluation indicator. The feature subgraph of the port application technology to be evaluated is constructed from the feature data of the port application technology to be evaluated according to the structure of the port technology knowledge graph. The target standard technology subgraph contains the evaluation indicator system and the standard features of each evaluation indicator.

[0054] Specifically, the aforementioned feature subgraphs and feature data have a clear logical relationship: the feature subgraph is a graphical representation of the feature data, the feature data is the attribute parameters, and the feature subgraph is a set of entities and relationships that carry such attribute parameters; the feature subgraph is constructed based on the feature data and according to the structural mapping of the port technology knowledge graph; the feature subgraph is used to perform subgraph matching reasoning, and the feature data is the basis for constructing the feature subgraph and the data source.

[0055] In a specific embodiment, using a complete example of a container terminal, if the port application technology to be evaluated is unmanned truck intelligent scheduling technology, its characteristic data is as follows: 1) Technical parameters: Scheduling response time < 2s, concurrent processing capacity 1000+ train trips.

[0056] 2) On-site operation data: average waiting time for trucks is 15s, and the path planning deviation is 0.5m.

[0057] 3) Safety and compliance data: ISO28000 supply chain security certification passed, 365 days of accident-free operation.

[0058] 4) Scene adaptation data: Adapted to the high-density operating environment of container terminals.

[0059] Next, the feature data will be constructed into a feature subgraph: Based on the structure of the port technology knowledge graph (technology-indicators-scenarios, etc.), these data are mapped into a graph as follows: (1) Node (entity) construction: Node A: Unmanned truck scheduling technology (Attributes: Response time < 2s, Concurrency 1000+).

[0060] Node B: Operational efficiency indicators (attributes: truck waiting time, path deviation).

[0061] Node C: Safety compliance metrics (attributes: ISO certification, incident-free period).

[0062] Node D: Container terminal scenario (Attributes: High density, multiple vehicle trips).

[0063] (2) Relationship (edge) construction: Edge A→B: Technical-indicator adaptation relationship (corresponding to the association between operational data and indicators).

[0064] Edge A→C: Technology-Compliance Relationship (corresponding to the association between security data and compliance indicators).

[0065] Edge A→D: Technology-Scene Mapping Relationship (corresponding to scene adaptation data).

[0066] This forms a complete "feature sub-graph of port application technologies to be evaluated".

[0067] Next, we will perform subgraph matching and reasoning operations: The feature subgraph of the port application technology to be evaluated already contains the two core evaluation indicator entities required for this evaluation: operational efficiency indicators and safety compliance indicators. This feature subgraph is then matched with the target standard technology subgraph pre-stored in the atlas (e.g., the complete graph structure of "standard unmanned truck scheduling technology" stored in the atlas). A general scoring range of 0-100 points is used as the base score, and the comparison is performed simultaneously across three dimensions to ultimately calculate the overall similarity matching score between the two. (1) Compare whether the "number and type of entities" are consistent. Sub-graph entities to be evaluated: unmanned truck scheduling technology, operational efficiency evaluation indicators, safety and compliance evaluation indicators, and container terminal scenarios.

[0068] Target standard sub-graph entities: standard unmanned truck scheduling technology, operation efficiency evaluation indicators, safety and compliance evaluation indicators, and container terminal scenarios.

[0069] Comparison results: The entities correspond completely, and the similarity in this part is at its maximum.

[0070] (2) Compare whether the attribute values ​​of each entity are close. Attributes of entities to be evaluated in the subgraph: response time of unmanned truck scheduling technology <2s, concurrent traffic 1000+ trips; truck waiting time of 15s and path planning deviation of 0.5m for operational efficiency indicators.

[0071] Target standard subgraph entity attributes: response time of standard unmanned truck scheduling technology <2s, concurrent traffic 1000+ trips; waiting time of operation efficiency indicators ≤20s, path planning deviation ≤1m.

[0072] Comparison results: All entity attribute values ​​meet the standards, and the matching degree can reach over 95%.

[0073] (3) Compare whether the "relationships between entities" are consistent. Subgraph relationships to be evaluated: Technology → Indicator Adaptation Relationship, Technology → Compliance Relationship, Technology → Scenario Mapping Relationship.

[0074] Target standard subgraph relationships: technology → indicator adaptation relationship, technology → compliance relationship, technology → application scenario mapping relationship.

[0075] Comparison results: The relationships between entities are completely identical, and the similarity in this part is at its maximum.

[0076] Based on the comprehensive comparison of the above three dimensions, the system can directly extract the matching degree of each evaluation indicator from the evaluation indicator entities already included in the corresponding feature subgraph in the matching results, and then generate the basic score of the corresponding evaluation indicator: a matching degree of 95% for the operation efficiency indicator attribute corresponds to a basic score of 95 points for operation efficiency, and a matching degree of 98% for the safety compliance indicator attribute corresponds to a basic score of 98 points for safety compliance. If the matching degree of the indicator is high, it means that the technical standard of the indicator dimension is met, and the basic score is high; if the matching degree of the indicator is low (for example, the response time is slower than 2 seconds, and the matching degree of operation efficiency is only 60%), it means that there is a defect in the indicator dimension, and the basic score of operation efficiency is only 60 points.

[0077] In an exemplary embodiment, the base score is adjusted based on the correlation between the evaluation indicators and the benefits to obtain individual scores for each evaluation indicator, including: Based on the path reasoning algorithm, the implicit relationship between each evaluation indicator and the benefit is explored, and the basic score is corrected to obtain the corrected score. The fuzzy inference algorithm is used to process uncertain feature data (such as fluctuation data caused by environmental disturbances and semi-structured operation data) in the feature data of port application technologies to be evaluated, and the correction score is calibrated based on the processing results to output the individual score of each evaluation indicator.

[0078] Specifically, in this embodiment, the port technology knowledge graph only directly records the explicit relationships between technology and indicators, and between technology and benefits. There is no direct relationship between the evaluation indicators and benefits, which are implicit relationships. Such implicit relationships need to be mined through path reasoning algorithms. For example, the path reasoning algorithm can determine the influence weight of each evaluation indicator on the corresponding benefit by traversing the association path of "evaluation indicator → technology → benefit" in the knowledge graph. Based on the influence weight, the basic score is corrected, thereby making up for the deficiency of the basic score that only considers the matching degree of technical parameters and does not combine the actual benefit value of the port, and thus obtaining the corrected score.

[0079] Taking the above-mentioned intelligent scheduling technology for unmanned container trucks as a specific example: through path reasoning mining, the operational efficiency index corresponds to the benefit of "improved operational efficiency," with an impact weight of 1.05; the safety compliance index corresponds to the benefit of "reduced safety risks," with an impact weight of 1.02; the operational efficiency correction score = 95 points × 1.05 = 99.75 points, and the safety compliance correction score = 98 points × 1.02 = 99.96 points; the slight fluctuation data caused by tidal interference in the port application technology to be evaluated is fuzzed to eliminate the impact of data uncertainty, and the correction scores are slightly calibrated; the final output single-item score is: operational efficiency single-item score 99 points, and safety compliance single-item score 100 points.

[0080] In an exemplary implementation, step S3: Based on the correlation between standard technologies and evaluation indicators in the port technology knowledge graph, and the correlation between standard technologies and application scenarios, determine the dynamic weight of each evaluation indicator, including: Based on the correlation between the port application technologies to be evaluated and the evaluation indicators, the entity correlation degree and path length between the port application technologies to be evaluated and the evaluation indicators are determined. The importance of a scenario is determined based on the correlation between the port application technologies and application scenarios to be evaluated. The dynamic weights of each evaluation indicator are calculated based on entity relevance, path length, and scenario importance.

[0081] In one optional implementation, the dynamic weights of each evaluation metric are calculated based on entity relevance, path length, and scene importance, including: Substituting entity relevance, path length, and scene importance into the formula: (1); The dynamic weights of each evaluation indicator are calculated; among them, Dynamic weights; For entity relevance; This is the sum of the entity correlations of all evaluation indicators; This is the path length; For scene importance; This is the sum of the importance of all scenarios; , and The weighting coefficients and .

[0082] In this embodiment, the dynamic weight quantification calculation method relies on multi-factor normalization fusion and an adaptively adjustable weighting formula to generate weights. This calculation method transforms three different dimensions of features—entity correlation, path length, and scene importance—into directly fused quantitative parameters through normalization processing, effectively eliminating the dimensional differences between different dimensions of data, allowing the three types of influencing factors to participate in weight calculation in a coordinated manner, and ensuring the stability and consistency of the weight calculation results. The quantification method of taking the reciprocal of the path length in formula (1) can intuitively map the degree of correlation between entities in the knowledge graph. The shorter the correlation path, the higher the corresponding weight contribution. This transforms the inherent logic of the graph topology into a calculable quantitative rule, realizing the accurate quantification of correlation relationships. , and It satisfies the constraint that the sum is 1 and is adaptively adjustable. It can flexibly adjust the contribution ratio of each factor according to the evaluation focus of different port scenarios such as container terminals, bulk cargo terminals, and oil and chemical terminals, so that the generation of dynamic weights has the ability to adapt to different scenarios. At the same time, it completes the automatic calculation of weights through a formulaic unified calculation rule, without the need for manual intervention to set weight thresholds. It transforms the original qualitative weight allocation into a standardized quantitative calculation, realizing the accurate, automated, and dynamic generation of evaluation weights, filling the gap in the existing technology that cannot generate port scenario-adaptive weights through multi-factor quantitative fusion.

[0083] In an exemplary embodiment, the method further includes: game calibration based on a large language model using multiple agents. Game calibration based on a large language model can specifically include: (1) The intelligent agent system uses the individual scores and dynamic weights of the indicators responsible for each intelligent agent as the initial anchor values, and divides the feature subgraphs related to the evaluation dimensions of each intelligent agent from the port technology knowledge graph and assigns them to the corresponding intelligent agents; among them, the efficiency priority intelligent agent receives the individual scores and dynamic weights of the operation efficiency indicators, the cost control intelligent agent receives the individual scores and dynamic weights of the cost control indicators, and the safety red line intelligent agent receives the individual scores and dynamic weights of the safety compliance indicators. (2) Each agent engages in multiple rounds of interactive debate based on its own set optimization goals, and conducts graph tracing and evidence retrieval based on its corresponding feature subgraph. If each agent has objections to its own single-item score, it calls the port technology knowledge graph query interface based on the path nodes and relationships in the corresponding feature subgraph to extract relevant path nodes and evolutionary relationships as rebuttal evidence, and uses the dynamic scoring evolution mechanism to update its own quantitative score in real time. (3) Based on the port multidimensional decision-making collaborative convergence algorithm, the global collaborative steady state judgment mechanism is used to monitor the scoring adjustment status of each agent in real time. When the global system fluctuation value of the agent system is less than the preset convergence threshold, the game is judged to have reached the collaborative steady state and the debate is terminated. The calibrated scores of the three evaluation indicators of efficiency, cost and safety after convergence are extracted.

[0084] After extracting the calibrated scores of the three indicators of efficiency, cost and security after convergence, they are combined with the original individual scores of other evaluation indicators that did not participate in the game, such as technology maturity, ease of operation and maintenance, and scenario adaptability. The scores are then re-weighted and integrated according to their respective dynamic weights to obtain the final calibrated comprehensive evaluation result, thereby calibrating the deviation caused by the original static benchmark rule calculation.

[0085] Regarding this embodiment of intelligent agents, it should be noted that when using multiple intelligent agents for game calibration, a large language model can be used as the underlying inference engine and encapsulated into an intelligent agent system with independent execution capabilities through software engineering methods. This intelligent agent system can include multiple intelligent agents (such as the three intelligent agents mentioned above), and the internal construction of each intelligent agent includes the following four main functional modules.

[0086] (1) Perspective instruction module, used to inject specific identity presets, optimize target boundaries and scoring bias rules into the underlying big model, so as to ensure that the agent always maintains a specific departmental position in the evaluation.

[0087] (2) Graph interaction toolchain, which encapsulates the graph query interface. When the agent needs to find evidence, the toolchain can automatically convert the natural language analysis intent generated by the agent into a knowledge graph query statement and extract the corresponding nodes and relation subgraphs in the graph database.

[0088] (3) Local state work area, which is used to temporarily store the input parameters of the port application technology to be evaluated, the initial global basic score (that is, the single score corresponding to the evaluation index), and the historical question records generated by other agents in multiple rounds of debate.

[0089] (4) Quantization output parser, which is responsible for intercepting the divergent text generated by the large model, forcing it to be formatted into standard structured data output, and extracting specific quantization scores and path evidence chains.

[0090] Building upon the modules described above, the agent system loads a large language model onto each agent and equips them with components such as a graph interaction toolchain. For the three agents used for evaluation, specific input and output definitions are established for each agent, as follows: The efficiency-first intelligent agent is designed to simulate the decision-making logic of port operations and production scheduling departments, aiming to maximize operational efficiency and throughput. Its input data includes basic feature vectors corresponding to efficiency evaluation indicators, individual scores, and feature subgraphs pre-segmented from the port's technical knowledge graph that are strongly correlated with operational time, flow rate, and throughput improvement. Its output data consists of efficiency dimension scores and the graph path evidence chain supporting those scores. The basic feature vectors are vectors constructed from the feature data. The benchmark quantitative score is the individual score of the corresponding evaluation indicator.

[0091] The cost control intelligent agent is designed to simulate the decision-making logic of the port's finance and asset management departments, and its optimization goal is to minimize the total cost over the entire life cycle. Its input data consists of the basic feature vectors corresponding to the cost assessment indicators, individual scores, and feature subgraphs related to hardware investment, energy consumption standards, and subsequent maintenance frequency in the port's technical knowledge graph. Its output data consists of cost dimension score values ​​and graph path evidence chains that refute or support the investment.

[0092] The Safety Red Line Intelligent Agent is designed to simulate the decision-making logic of port safety supervision departments, and its optimization goal is to possess the highest level of risk aversion attribute and veto power. Its input data consists of basic feature vectors corresponding to safety assessment indicators, individual scores, and feature subgraphs in the port technology knowledge graph related to the impact of extreme weather, chain reactions of equipment failures, and operator compliance. Its output data consists of safety dimension score values ​​and risk warning evidence chains. If a path directly leading to personnel casualties or major equipment shutdowns is found in the graph, the intelligent agent has the right to output an extremely low blocking score.

[0093] In practice, multiple intelligent agents work together. The intelligent agent system distributes the individual scores of the corresponding evaluation indicators and the entity names of the target standard technologies corresponding to the port application technologies to be evaluated to the three intelligent agents. Each intelligent agent, in combination with the assigned feature subgraph, independently calculates and outputs the initial first-round internal score and evidence chain.

[0094] Next, the cross-examination phase begins. The system broadcasts the scores and evidence chains output by the three agents in the first round. For example, when the efficiency-first agent outputs a score for improved efficiency, the safety-red-line agent receives this output and will automatically trigger its built-in graph interaction toolchain to initiate a special search on the port technology knowledge graph, attempting to find hidden correlation paths that may lead to a decrease in safety redundancy due to improved efficiency. If evidence of a reverse path is found, the safety agent will generate a rebuttal instruction and lower its own score output.

[0095] Next, in the cross-questioning and dynamic scoring phase, each agent must output its quantitative score from the previous round. And the quantitative score of the current round The generation and evaluation process of this score is as follows: First, the agent parses the newly retrieved graph path evidence through a large language model, evaluates and quantifies the support strength of the evidence, and simultaneously reads the cross-departmental compromise pressure values ​​exerted by other agents in the local state workspace. Subsequently, the agent's quantization output parser strictly follows the preset agent dynamic scoring evolution method, weighs the weight between the strength of new evidence and the compromise resistance, and calculates the specific quantification score for the current round. The updated evidence chain is then formatted and output. As the number of questioning rounds increases, the port multi-dimensional decision-making collaborative convergence algorithm running in the system's background continuously collects the latest quantitative scores from each agent's output. When the global system fluctuation value calculated in the background is less than the preset limit convergence threshold, the system referee module forcibly cuts off the graph interaction toolchain of the agents and terminates the debate; the system finally extracts the original individual scores of the three agents in the last round of output performance score, cost score, and safety score and other indicators, and performs weighted fusion to generate the final comprehensive evaluation score, and summarizes the evidence chain output by them into a multi-perspective selection evaluation report.

[0096] For example, in one specific implementation, the individual scores and dynamic weights of the six evaluation indicators are calculated as follows: Work efficiency index: Individual score 95 points, weight 0.35; Cost control indicator: 90 points for each item, weight 0.30; Safety compliance indicator: Individual score 98 points, weight 0.25; Technology Maturity Index: Individual score 92 points, weight 0.05; Ease of operation and maintenance index: 88 points for this item, weight 0.03; Scene adaptability index: Single item score 94 points, weight 0.02.

[0097] The intelligent agent game calibration process is as follows: (1) Agent allocation and initial anchoring: The agent system only allocates the individual scores, dynamic weights and corresponding feature subgraphs of the three indicators of work efficiency, cost control and safety compliance to the efficiency-priority agent, cost control agent and safety red line agent respectively; while the three indicators of technology maturity, operation and maintenance convenience and scenario adaptability do not participate in the game and keep their original scores and weights unchanged.

[0098] (2) Multiple rounds of debate and graph tracing: The efficiency-first agent believes that the job efficiency score of 95 is too low. Based on the assigned efficiency feature subgraph, it retrieves graph path evidence and corrects the score to 98. The cost control agent accepts the initial score and keeps it unchanged at 90. The safety red line agent finds that there is a slight risk of decreased safety redundancy in efficiency improvement and corrects the score to 95.

[0099] (3) Collaborative convergence and termination: The system calculates the global fluctuation value through the port multidimensional decision collaborative convergence algorithm. When the fluctuation value is less than the limit convergence threshold, the debate is terminated and three calibration scores are obtained: operational efficiency: 98 points, cost control: 90 points, safety compliance: 95 points.

[0100] (4) Final comprehensive score calculation: The scores of the three calibrated indicators are combined with the original scores of technology maturity, operation and maintenance convenience and scenario adaptability that did not participate in the game, and weighted and integrated according to their respective dynamic weights: final comprehensive score = 98×0.35+90×0.30+95×0.25+92×0.05+88×0.03+94×0.02=94.47 points.

[0101] It should be noted that during the aforementioned multiple rounds of debate, the scoring adjustments by each agent were not arbitrary text generation, but strictly controlled by the agent system's unique port multi-dimensional decision-making collaborative convergence algorithm. This algorithm ensures that the game process develops towards a convergent steady state by quantifying the agents' outputs. The algorithm consists of the following two core formulas: The first part is the dynamic scoring evolution formula for agents, used to quantitatively calculate the score changes of each agent after each round of debate. Its mathematical expression is: (2); variable and These represent the agent's current round (i.e., the [number]th round). (round) and the previous round (i.e., the 1st round) The quantitative score is calculated in rounds. The initial value of the score comes from the initial individual score calculated by the system, and the subsequent values ​​are updated in real time by each evaluation agent in multiple rounds of game.

[0102] variable This represents the influence coefficient of the evidence presented in the graph. This coefficient is extracted by the system based on pre-defined prior knowledge of the specific business scenario at the port. For example, in the scenario of a dangerous goods terminal, the evidence influence coefficient of the safety assessment agent will be set to the highest level by the system.

[0103] variable This represents the strength of path evidence newly retrieved by the agent from the port technology knowledge graph in the current round. This parameter is an autonomous output of the agent, driven entirely by a large language model. The agent parses the retrieved graph subgraph and, based on its inherent logical reasoning chain, quantifies it into a specific numerical value, representing the strength of the evidence's support or opposition to the current technology selection.

[0104] variable This represents the constraint coefficient between different agents. This parameter originates from the configuration settings of the port management architecture and is used to simulate real cross-departmental game resistance. The formula indicates that an agent's score in the next round depends not only on the strength of new evidence mined by its own large model but also on the absolute difference between the current scores of agents from other interest perspectives.

[0105] The second part is the global cooperative steady-state determination formula, which serves as the system's adjudication mechanism to determine whether multiple rounds of debate should be terminated. Its mathematical expression is: (3); In formula (3), the variable This represents the global system fluctuation value. This value is calculated in real-time by the system's background referee module, which collects the scores from each agent. (Variable) Represents the weight of the cross-agent disagreement penalty. (Variable) This represents the round in which agents collaborate to make a decision. This parameter is initialized by the system administrator based on the port's risk tolerance for introducing new technologies. The lower the tolerance, the greater the penalty weight, meaning the system requires all agents to reach a high degree of consensus on the score before the debate can end.

[0106] The system continuously calculates the global system fluctuation value after each round of interaction. When this fluctuation value is less than the system's preset convergence threshold, the system determines that each agent has reached a cooperative steady state where it can no longer unilaterally expand its own indicator benefits. At this point, the multi-round debate is forced to automatically converge and stop. The system extracts the quantitative score of the final compromise of each agent, and outputs a comprehensive evaluation score and detailed technical selection recommendations after weighted fusion.

[0107] In this embodiment, intelligent agent game theory is used to evaluate the three indicators of efficiency, cost, and safety, primarily for the following reasons: First, efficiency, cost, and safety are the three most core, rigid, and conflict-ridden decision dimensions in port operation and management, directly determining whether a technology can be implemented, whether it is compliant, and whether it is economically viable—they are veto-level critical indicators. Second, there are inherent differences in stance among the operations department, finance department, and safety supervision department. The operations department pursues maximum efficiency, the finance department pursues minimum cost, and the safety supervision department insists on the inviolability of safety red lines. These three parties have inherent conflicts of interest, requiring dynamic equilibrium through game theory. Traditional static weighting methods cannot realistically simulate this actual decision-making process. Furthermore, other indicators such as technology maturity, ease of operation and maintenance, and scenario adaptability are auxiliary and reference indicators, without significant departmental opposition or conflict of interest. Therefore, they do not require correction through multi-agent debate and can be directly calculated using static results.

[0108] This embodiment boasts the following unique technical advantages: By employing a two-tiered evaluation model combining static rule calculation with dynamic game-theoretic calibration, it retains the objectivity of knowledge graph reasoning while realistically simulating the cross-departmental collaborative decision-making process in ports, thus overcoming the shortcomings of traditional evaluation methods that are overly mechanical and detached from practical application scenarios. Simultaneously, by calibrating the three core rigid indicators of efficiency, cost, and safety through game-theoretic analysis, it significantly reduces computational complexity while ensuring that the evaluation results closely align with the three core concerns of ports: efficiency, investment, and safety, making them more relevant to actual engineering applications. Furthermore, by excluding auxiliary indicators such as technology maturity and ease of operation and maintenance from the game-theoretic process, it effectively avoids diluting the core decision-making weight with non-core factors, making the overall evaluation logic more focused, the evaluation conclusions more stable, and the evaluation more persuasive. Finally, each calibrated score is accompanied by a corresponding knowledge graph path evidence chain, forming a traceable and interpretable multi-perspective decision-making basis, solving the black-box problem of traditional evaluations that only output scores without providing reasonable explanations.

[0109] In an exemplary embodiment, the evaluation method may further include: (1) Based on the preset score threshold of the port industry, the comprehensive score is compared numerically, and the evaluation level of the port application technology to be evaluated is determined according to the comparison result. For example, the evaluation level may include excellent, good, qualified, unqualified. (2) Using the standard technology obtained by matching as a reference, and combining the technology substitution and evolution relationship pre-built in the port technology knowledge graph, the port application technology to be evaluated is compared horizontally with similar technologies and vertically with historical version evolution, forming a multi-perspective evaluation result that includes scoring differences, technology upgrade range, and ranking of similar technologies. (3) Extract the indicator nodes in the evaluation indicator system whose individual scores are lower than the preset benchmark value, and combine them with the association path between the evaluation indicators and risks in the port technology knowledge graph to identify risk warning information; (4) Based on the optimal characteristic parameters of the standard technology and the characteristic differences of the port application technology to be evaluated, combined with the correlation between the evaluation indicators and benefits, technical optimization suggestions are generated; (5) Integrate individual scores, comprehensive scores, evaluation levels, multi-perspective evaluation results, risk warning information and technical optimization suggestions to output a complete port application technology evaluation report.

[0110] In an exemplary embodiment, a complete port application technology assessment report can be presented intuitively using a combination of knowledge graph visualization and dashboard visualization. The knowledge graph visualization displays the network of relationships between the port application technologies being assessed and their indicators, scenarios, and benefits, clearly presenting the assessment logic. The dashboard visualization displays the score distribution of each indicator, comparison curves with similar technologies, and longitudinal evolution trend graphs. The assessment report allows users to view the assessment details and reasoning process of specific indicators through drill-down and filtering operations.

[0111] The evaluation method in this embodiment of the invention may also have interactive functionality, allowing users to view the evaluation details, reasoning process, and data sources of specific indicators through drill-down and filtering operations, thereby improving the credibility and readability of the evaluation results.

[0112] In an exemplary implementation, the evaluation method may further include: continuously optimizing the evaluation system based on feedback data, with the specific implementation process as follows: (1) Feedback data collection: Collect multi-dimensional feedback data of the evaluation results. The feedback data may include the actual application effect data of the port application technology to be evaluated, the manual calibration opinions of industry experts, and the latest industry standard update data. (2) Graph and model optimization: Update the port application technology-specific knowledge graph based on feedback data, including adjusting indicator thresholds, strengthening / weakening the strength of entity associations, and adding the latest industry evaluation indicators; at the same time, optimize the weight calculation parameters and inference rules, such as adjusting the coefficients of α, β, and γ, and updating the similarity threshold of subgraph matching; (3) Verification of assessment results: The consistency of the assessment results with those of port industry experts is verified. The preset consistency threshold is, for example, 85%. When the consistency is lower than the threshold, a secondary optimization process is triggered until the consistency between the assessment results and the expert opinions meets the requirements, thus ensuring the reliability of the assessment.

[0113] Compared with existing technologies, the advantages of this invention are: 1) High accuracy and objectivity in evaluation: Based on a knowledge graph specific to port application technologies, the invention achieves indicator association matching and dynamic weight calculation, combined with reasoning evaluation through multi-algorithm fusion, avoiding subjective human intervention, resulting in more objective and accurate evaluation results; 2) Strong adaptability to port scenarios: The invention designs knowledge graphs, data preprocessing algorithms, and evaluation rules specifically for the port industry, adapting to the characteristics of multiple scenarios, high dynamism, and strong security requirements of ports, thus solving the problem of poor adaptability of general evaluation methods; 3) Significantly improved assessment efficiency: The entire process from data collection and preprocessing to assessment result output is automated. For example, the assessment cycle is shortened from 15-30 days in traditional manual assessment to 1-3 days, adapting to the assessment needs of rapid iteration of new port technologies; 4) Comprehensive assessment dimensions and strong practicality: It supports multi-dimensional assessment, horizontal comparative assessment and vertical evolution assessment, and outputs quantitative scores, level determination, risk warning and optimization suggestions, providing comprehensive support for port decision-making; 5) Dynamic iteration capability: It realizes continuous optimization of knowledge graph and assessment model based on feedback data, adapts to changes in industry standards, technology iteration and scenario requirements, and ensures the long-term effectiveness of assessment methods.

[0114] The following describes a port application technology evaluation system based on knowledge graphs provided by the present invention. The port application technology evaluation system based on knowledge graphs described below and the port application technology evaluation method based on knowledge graphs described above can be referred to and correspond to each other.

[0115] like Figure 3 As shown, this embodiment of the invention also provides a port application technology evaluation system based on a knowledge graph, used to implement the port application technology evaluation method based on a knowledge graph in any of the above embodiments. This port application technology evaluation system based on a knowledge graph may include: The matching module 310 is used to match the port application technology to be evaluated based on the port technology knowledge graph to obtain the target standard technology and its target evaluation index system corresponding to the port application technology to be evaluated; the target evaluation index system includes multiple evaluation indicators. The scoring module 320 is used to generate basic scores for each evaluation indicator in the target evaluation indicator system through knowledge graph reasoning, and to adjust the basic scores based on the correlation between the evaluation indicators and benefits to obtain individual scores for each evaluation indicator. The weight determination module 330 is used to determine the dynamic weight of each evaluation indicator based on the correlation between the target standard technology and the evaluation indicators, as well as the correlation between the target standard technology and the application scenario. The weighting module 340 is used to perform weighted calculations based on the individual scores of each evaluation indicator and their corresponding dynamic weights to obtain a comprehensive score.

[0116] This invention also provides an electronic device, which may include a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other via the communication bus. The memory stores a computer program that can be executed by the processor; when the processor runs the computer program, it can execute the knowledge graph-based port application technology evaluation method of any of the above embodiments.

[0117] Furthermore, the logical instructions in the aforementioned memory can be implemented as software functional units and sold or used as independent products, and can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory, random access memory, magnetic disks, or optical disks.

[0118] In another aspect, the present invention also provides a non-transitory computer-readable storage medium, wherein the computer storage medium stores instructions that, when executed, implement the knowledge graph-based port application technology evaluation method in any of the above embodiments.

[0119] Although the invention has been described herein in conjunction with various embodiments, those skilled in the art will understand and implement other variations of the disclosed embodiments by reviewing the accompanying drawings, the disclosure, and the appended claims in carrying out the claimed invention. In the claims, the word "comprising" does not exclude other components or steps, and "a" or "an" does not exclude a plurality. A single processor or other unit can implement several functions listed in the claims. While different dependent claims may recite certain measures, this does not mean that these measures cannot be combined to produce good results.

[0120] Although the invention has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made therein without departing from the spirit and scope of the invention. Accordingly, this specification and drawings are merely exemplary descriptions of the invention as defined by the appended claims, and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its spirit and scope. Thus, if such modifications and modifications of the invention fall within the scope of the claims and their equivalents, the invention is also intended to include such modifications and modifications.

Claims

1. A port application technology evaluation method based on knowledge graph, characterized in that, The methods include: Based on the port technology knowledge graph, the application technologies of the port to be evaluated are matched to obtain the target standard technologies and their target evaluation index systems corresponding to the application technologies of the port to be evaluated. The target evaluation index system includes multiple evaluation indicators; The basic scores of each evaluation indicator in the target evaluation indicator system are generated by reasoning through knowledge graph, and the basic scores are adjusted based on the correlation between the evaluation indicators and benefits to obtain the individual scores of each evaluation indicator. Based on the correlation between the target standard technology and the evaluation indicators, as well as the correlation between the target standard technology and the application scenario, the dynamic weight of each evaluation indicator is determined. The comprehensive score is obtained by weighting the individual scores of each evaluation indicator with their corresponding dynamic weights.

2. The port application technology evaluation method based on knowledge graphs according to claim 1, characterized in that, The basic scores for each evaluation indicator in the target evaluation indicator system are generated through knowledge graph reasoning, including: Based on the subgraph matching reasoning algorithm, the feature subgraph of the port application technology to be evaluated is matched with the target standard technology subgraph pre-stored in the port technology knowledge graph to obtain the basic score of each evaluation indicator; wherein, the feature subgraph of the port application technology to be evaluated is constructed from the feature data of the port application technology to be evaluated according to the structure of the port technology knowledge graph; the target standard technology subgraph contains the evaluation indicator system and the standard features of each evaluation indicator.

3. The port application technology evaluation method based on knowledge graphs according to claim 2, characterized in that, The basic scores are adjusted based on the correlation between evaluation indicators and benefits to obtain individual scores for each evaluation indicator, including: Based on the path reasoning algorithm, the implicit relationship between each evaluation indicator and the benefit is explored, and the basic score is corrected to obtain the corrected score; The fuzzy inference algorithm is used to process the uncertain feature data in the feature data of the port application technology to be evaluated, and the correction score is calibrated based on the processing result to output the individual score of each evaluation index.

4. The port application technology evaluation method based on knowledge graphs according to claim 1, characterized in that, Based on the relationships between standard technologies and evaluation indicators in the port technology knowledge graph, as well as the relationships between standard technologies and application scenarios, the dynamic weights of each evaluation indicator are determined, including: Based on the correlation between the port application technologies to be evaluated and the evaluation indicators, the entity correlation degree and path length between the port application technologies to be evaluated and the evaluation indicators are determined. The importance of a scenario is determined based on the correlation between the port application technologies and application scenarios to be evaluated. The dynamic weights of each evaluation index are calculated based on the entity correlation, the path length, and the scene importance.

5. The port application technology evaluation method based on knowledge graphs according to claim 4, characterized in that, Based on the entity relevance, the path length, and the scene importance, the dynamic weights of each evaluation indicator are calculated, including: Substituting the entity association degree, the path length, and the scene importance into the formula: ； The dynamic weights of each evaluation indicator are calculated; among them, The dynamic weight; The degree of entity association; This is the sum of the entity correlations of all evaluation indicators; The path length; The importance of the scene; This is the sum of the importance of all scenarios; , and The weighting coefficients and .

6. The port application technology evaluation method based on knowledge graphs according to claim 2, characterized in that, The method further includes: game calibration based on a large language model for multiple agents; the multiple agents include an efficiency-prioritizing agent, a cost-controlling agent, and a safety red line agent; Multi-agent game calibration based on large language models includes: The intelligent agent system uses the individual scores and dynamic weights of the indicators responsible for each intelligent agent as initial anchor values, and divides the feature subgraphs related to the evaluation dimensions of each intelligent agent from the port technology knowledge graph and assigns them to the corresponding intelligent agents; among them, the efficiency priority intelligent agent receives the individual scores and dynamic weights of the operation efficiency indicators, the cost control intelligent agent receives the individual scores and dynamic weights of the cost control indicators, and the safety red line intelligent agent receives the individual scores and dynamic weights of the safety compliance indicators. Each agent engages in multiple rounds of interactive debate based on its own set optimization goals, and conducts graph tracing and evidence retrieval based on its corresponding feature subgraph. If an agent disagrees with its own single-item score, it calls the port technology knowledge graph query interface based on the path nodes and relationships in the corresponding feature subgraph to extract relevant path nodes and evolutionary relationships as rebuttal evidence, and uses a dynamic scoring evolution mechanism to update its own quantitative score in real time. Based on the port multidimensional decision-making collaborative convergence algorithm, the global collaborative steady-state determination mechanism is used to monitor the scoring adjustment status of each agent in real time. When the global system fluctuation value of the agent system is less than the preset convergence threshold, the game is determined to have reached a collaborative steady state and the debate is terminated. The calibrated scores of the three evaluation indicators of efficiency, cost and safety after convergence are extracted.

7. The port application technology evaluation method based on knowledge graphs according to claim 2, characterized in that, Before matching the port application technologies to be evaluated based on the port technology knowledge graph, the method further includes: Obtain initial characteristic data of the port application technology to be evaluated; the initial characteristic data includes technical parameter data, on-site operation data, cost input data, safety and compliance data, and scenario adaptation data; The initial feature data is preprocessed to obtain the feature data; the preprocessing includes an adaptive denoising operation, a data completion operation, an outlier removal operation, and a normalization mapping operation performed sequentially.

8. The port application technology evaluation method based on knowledge graphs according to claim 7, characterized in that, The adaptive noise reduction operation is achieved by using a filtering threshold that adapts to port machinery vibration, tidal cycles, and loading / unloading operation pulse interference.

9. The port application technology evaluation method based on knowledge graphs according to claim 8, characterized in that, The data completion operation is implemented based on the time sequence association rules of port operation process.

10. A port application technology evaluation system based on knowledge graphs, characterized in that, include: The matching module is used to match the port application technology to be evaluated based on the port technology knowledge graph, and to obtain the target standard technology and its target evaluation index system corresponding to the port application technology to be evaluated. The target evaluation index system includes multiple evaluation indicators; The scoring module is used to generate basic scores for each evaluation indicator in the target evaluation indicator system through knowledge graph reasoning, and to adjust the basic scores based on the correlation between the evaluation indicators and benefits to obtain individual scores for each evaluation indicator. The weight determination module is used to determine the dynamic weight of each evaluation indicator based on the relationship between the target standard technology and the evaluation indicators, as well as the relationship between the target standard technology and the application scenario. The weighting module is used to calculate the comprehensive score by weighting the individual scores of each evaluation indicator with their corresponding dynamic weights.

Images