A method, system, device and medium for full-link component adaptive mapping and optimization fused with digital map coding

Abstract

The application discloses a kind of fusion digital map encoding full-link component adaptive mapping and optimization method, system, equipment and medium, including to the feature of acquisition component is weighted digital coding, constructs component relationship diagram and generates node representation, carries out component adaptive matching based on node representation, determines mapping relationship and priority, generates candidate invocation path, combines constraint, target execution path is screened with weight, collects running feedback, dynamically updates feature weight, mapping priority and path evaluation weight, based on updated parameter re-executes whole process, by graph structure modeling, fusion structure and semantic features, improve the matching precision, the introduction of dependency consistency correction can avoid link breakage, by multi-objective cost evaluation screening optimal path, constructs running feedback closed loop mechanism, realizes parameter adaptive update and rule back-annotation, solve the problem that static model cannot respond dynamic change, to improve the running efficiency, stability and explainability of system.

Description

Technical Field

[0001] This invention relates to the field of component mapping and optimization technology, specifically to a method, system, device, and medium for end-to-end component adaptive mapping and optimization that integrates digital graph coding. Background Technology

[0002] In existing component mapping and optimization techniques, there is a lack of unified graph modeling capabilities for multidimensional dependencies between components. Matching is mainly based on manually defined dependency matrices or preset templates, assuming that component connections are fixed and linearly separable. However, in real complex systems, component dependencies exhibit multi-level, non-linear, and dynamic migration characteristics, which are affected by factors such as service environment, call sequence, and data flow direction. If traditional static rule mapping is used, component redundancy and dependency breakage may occur, thus failing to form an effective full-link mapping.

[0003] However, existing component mapping only builds models for fixed topologies or static processes, lacking the ability to automatically adjust mapping weights and path structures based on feedback. When component feature distributions shift or dependent paths are reconstructed, the model cannot be updated or repaired in a timely manner, and accuracy will continue to decline. Furthermore, existing methods usually treat matching and optimization as two independent stages, lacking a unified optimization objective and closed-loop feedback mechanism. When some components experience performance degradation or communication delays, the system cannot dynamically adjust the link structure or update and optimize.

[0004] In addition, traditional algorithms only output matching results or process paths, lacking an interpretable explanation of the process of forming mapping relationships and the basis for path selection. This makes it difficult for system maintainers to understand the mapping generation logic. Furthermore, mapping models depend on the component characteristics and topology of specific systems. During migration, they need to be retrained or have parameters manually adjusted. If components change frequently or new modules are launched, static mapping models cannot be updated and learned online, making it difficult to meet the needs of automatic learning and evolution of component mapping relationships in complex systems. Summary of the Invention

[0005] In view of the above-mentioned problems, the present invention provides a method, system, device and medium for adaptive mapping and optimization of end-to-end components that integrates digital graph coding.

[0006] Therefore, the technical problem solved by the present invention is that the existing technology has problems such as simple mapping structure, static model, lack of global adaptive optimization and feedback, which makes it difficult to adapt to the application requirements of complex systems.

[0007] To address the aforementioned technical problems, this invention provides the following technical solution: a method for adaptive mapping and optimization of end-to-end components integrating digital graph coding, comprising: Collect feature information of system components, extract features, and perform weighted digital encoding based on preset feature weights; Based on the digitally encoded features and component dependencies, a component relationship graph is constructed, and node representations are generated using graph encoding technology. Based on the node representation, adaptive component matching is performed to determine the component mapping relationship and mapping priority parameters; Based on the mapping relationship, candidate call paths are generated, and target execution paths are selected by combining preset system operation constraints and path evaluation weights. Components are invoked according to the target execution path, running feedback information is collected, and feature weights, mapping priority parameters and path evaluation weights are dynamically updated based on the running feedback information. Based on the updated weights and parameters, the weighted digitization encoding, component adaptive matching, and target execution path selection are re-executed.

[0008] As a preferred embodiment of the end-to-end component adaptive mapping and optimization method integrating digital graph coding described in this invention, the weighted digital coding includes: The feature information is standardized to map features of different dimensions to a unified numerical range, and component feature vectors are generated by combining preset feature weights.

[0009] As a preferred embodiment of the end-to-end component adaptive mapping and optimization method integrating digital graph coding described in this invention, the step of constructing a component relationship graph based on the digitally encoded features and component dependencies, and generating node representations using graph coding technology, includes: Construct a component relationship graph containing a set of nodes and a set of edges based on component dependencies; The feature similarity between component pairs with dependencies is calculated based on the component feature vectors, and the edge weights are determined by combining the dependency strength and feature similarity. A weighted adjacency matrix is ​​constructed based on the edge weights and the set of nodes. If there is an edge between node pairs, the matrix element takes the corresponding edge weight; otherwise, the matrix element is zero. The component feature vector is used as the initial feature of the node, and the feature is aggregated and updated using graph coding technology based on the weighted adjacency matrix to generate the node representation.

[0010] As a preferred embodiment of the end-to-end component adaptive mapping and optimization method integrating digital graph coding described in this invention, the component adaptive matching includes: Determine the source component set and the target component set from the system components; Traverse the source component set and the target component set to generate candidate matching pairs; The attribute similarity of the candidate matching pairs is calculated based on the component feature vectors, and the structural similarity of the candidate matching pairs is calculated based on the node representations. The initial matching score is obtained by weighting the attribute similarity and structural similarity. Based on the consistency of the dependency direction of the matched associated components on both sides of the candidate matching pair, a dependency consistency correction value is generated; The initial matching score is corrected based on the dependency consistency correction value to obtain the corrected matching score, and the component mapping relationship and mapping priority parameters are determined based on the corrected matching score.

[0011] As a preferred embodiment of the end-to-end component adaptive mapping and optimization method integrating digital graph coding described in this invention, the step of generating candidate call paths based on the mapping relationship and filtering target execution paths by combining preset system operation constraints and path evaluation weights includes: Based on the component mapping relationship and the component dependency relationship, generate candidate call paths that satisfy preset system operation constraints; The execution latency, resource consumption, and path stability of the candidate call paths are statistically analyzed. The execution latency, resource consumption, and path stability are weighted and calculated based on preset path evaluation weights to obtain a comprehensive cost. The candidate call path with the lowest overall cost is determined as the target execution path.

[0012] As a preferred embodiment of the end-to-end component adaptive mapping and optimization method integrating digital graph coding described in this invention, the step of dynamically updating the feature weights, mapping priority parameters, and path evaluation weights based on the operational feedback information includes: Calculate a feedback evaluation value based on the operation feedback information, and compare the feedback evaluation value with a preset threshold. Adjust the mapping priority parameter based on the comparison result; The path evaluation weight and the feature weight are adjusted based on the operational feedback evaluation value.

[0013] As a preferred embodiment of the end-to-end component adaptive mapping and optimization method integrating digital graph coding described in this invention, the step of re-executing weighted digital coding, component adaptive matching, and target execution path filtering based on updated weights and parameters includes: A structured rule set is generated based on the adjusted mapping priority parameters, path evaluation weights, and feature weights. The structured rule set is then merged with the historical rule set, the rule base is updated, and a back injection is performed.

[0014] This invention provides an end-to-end component adaptive mapping and optimization system that integrates digital graph coding.

[0015] To address the aforementioned technical problems, this invention provides the following technical solution: a full-link component adaptive mapping and optimization system integrating digital graph coding, comprising: The feature encoding module is used to collect feature information of system components, extract features, and perform weighted digital encoding based on preset feature weights. The graph model construction module is used to construct a component relationship graph based on the digitally encoded features and component dependencies, and to generate node representations using graph encoding technology. The component matching module is used to perform adaptive component matching based on the node representation, and determine the component mapping relationship and mapping priority parameters; The path optimization module is used to generate candidate call paths based on the mapping relationship, and filter target execution paths by combining preset system operation constraints and path evaluation weights. The parameter update module is used to call components according to the target execution path, collect running feedback information, and dynamically update feature weights, mapping priority parameters, and path evaluation weights based on the running feedback information. The closed-loop evolution module is used to re-execute weighted digital encoding, component adaptive matching, and target execution path selection based on the updated weights and parameters.

[0016] The present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, characterized in that the processor executes the computer program to implement the steps of the described method for adaptive mapping and optimization of end-link components incorporating digital graph coding.

[0017] The present invention provides a computer-readable storage medium having a computer program stored thereon, characterized in that, when the computer program is executed by a processor, it implements the steps of the aforementioned method for adaptive mapping and optimization of end-link components incorporating digital graph coding.

[0018] The beneficial effects of this invention are as follows: By generating component feature vectors and node representations through weighted digital encoding and graph encoding techniques, joint modeling of multi-dimensional features of components is achieved. In component adaptive matching, dependency consistency correction values ​​are generated to improve component matching accuracy and mapping integrity. In addition, by calculating the comprehensive cost to filter the target execution path, the global optimal path search under multi-objective constraints can be achieved. Based on the running feedback information, the feature weights, mapping priority parameters, and path evaluation weights are dynamically updated, which can enhance the system's self-learning and self-correction capabilities. A structured rule set is generated according to the adjusted parameters, the rule base is updated, and back-injection is performed, thereby constructing a closed-loop optimization mechanism, providing an interpretable explanation of the mapping basis, supporting cross-scenario migration and incremental learning, and thus ensuring execution efficiency and system operation stability. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart illustrating an overall process for an end-to-end component adaptive mapping and optimization method that integrates digital graph coding, as provided in one embodiment of the present invention. Detailed Implementation

[0021] To make the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0022] Example 1, referring to Figure 1 This is one embodiment of the present invention, which provides a method for adaptive mapping and optimization of end-to-end components that integrates digital graph coding, including the following steps S1 to S6: In step S1, feature information of system components is collected, features are extracted, and weighted digital encoding is performed based on preset feature weights; Furthermore, system components can include hardware or software components such as servers, microservices, containers, and databases. The runtime records of all components constitute the original component set, specifically represented as follows: in, Indicates the total number of components. Indicates the first Each component.

[0023] For each It contains multiple feature fields, which extract raw features from basic attributes, interface definitions, runtime logs and dependency descriptions. For example, it extracts component type, version number, number of interfaces and number of parameters from basic attributes, call frequency, average execution time, number of exceptions and resource utilization from runtime logs, and upstream dependency number, downstream call number and dependency level depth from dependency descriptions. Furthermore, we obtain the first The original feature sequences of each component are specifically represented as follows: in, Indicates the first The original values ​​of each component in the k-th feature dimension This represents the number of feature dimensions.

[0024] By extracting quantifiable features from the basic attribute information, interface information, runtime behavior information, and dependency description information of components, and converting them into a unified numerical representation, component feature vectors and component feature matrices are obtained. These are then input into subsequent stages for graph structure modeling.

[0025] In this embodiment, weighted digital encoding includes: standardizing feature information, mapping features of different dimensions to a unified numerical range, and generating component feature vectors by combining preset feature weights. For example, for the k-th feature dimension, the th The standardization result of each component is denoted as ; Specifically, the range standardization method is used for calculation, which is expressed as follows: and This represents the minimum and maximum values ​​of all components in the k-th feature dimension.

[0026] Further, obtain the preset feature weight set, denoted as... , in, The weight represents the k-th feature dimension, used to indicate the relative importance of the k-th component feature field in behavioral analysis; In addition, the field weights are initially set by experts based on their experience with the business importance and security risks of the components, and are then determined after being corrected based on the ability of each field in the historical operation logs to distinguish abnormal behavior. Preferably, for the weight set Normalization is performed so that the sum of the weights of each dimension is 1.

[0027] Multiplying the standardized result by the corresponding weights dimension by dimension, we obtain the first... Feature vectors of each component: in, Indicates the first Numerical feature representation of each component; Furthermore, the feature vectors of all components are arranged in component order to construct a component feature matrix: Furthermore, using a matrix Each column is used as a sample sequence for the corresponding feature dimension. The correlation coefficient between any two feature dimensions is calculated to obtain the feature correlation matrix: in, Indicates the first The feature dimension and the first The degree of correlation between the feature dimensions The Pearson correlation coefficient is preferred for calculation, but other conventional correlation coefficient calculation methods may also be used.

[0028] In step S1, the feature correlation matrix As output data, it is passed to the next stage as one of the input data for graph structure modeling, and can be used for subsequent feature relationship analysis and interpretable output.

[0029] In step S2, a component relationship graph is constructed based on the digitally encoded features and component dependencies, and node representations are generated using graph encoding technology; It should be noted that in step S2, the main focus is on component dependencies and component feature matrices. We construct a component relationship graph, and then use a graph neural network to extract the structural semantic representation of each component.

[0030] In an optional implementation, step S2 above includes the following sub-steps A1-A4: A1: Construct a component relationship graph containing a set of nodes and a set of edges based on component dependencies; Specifically, based on the call relationships, dependencies, and data interaction relationships between components, a directed graph can be constructed through system log parsing, configuration file reading, or service tracing systems, and represented as follows: in, Represents a set of nodes, nodes Corresponding components , Represents a set of edges, if the component With components If there are calls or dependencies, then in the node With nodes Establish edges between ; By processing in the above way, system components can be mapped to nodes in a graph structure, and dependencies between components can be mapped to edges in the graph, thereby realizing a graph representation of component dependencies.

[0031] A2: Calculate the feature similarity between component pairs that have dependencies based on the component feature vectors, and determine the edge weights by combining the dependency strength and feature similarity; Furthermore, for component pairs with dependencies... The cosine similarity is used to calculate the similarity between the feature vectors of the components, specifically expressed as follows: in, and Representation Component and The component feature vector, and Represents the magnitude of a vector.

[0032] Furthermore, the strength of the dependency With feature similarity Weighted combination yields edge weights: in, Representation Component For components The dependency strength can be obtained from historical operation logs. It is preferable to normalize the calculation based on the call frequency, data interaction volume, or dependency tightness per unit time. The more frequent the calls or the larger the data interaction volume, the higher the dependency strength. and These are weighting coefficients, which can be configured to emphasize the degree of dependence strength and feature similarity according to the application scenario; By obtaining edge weights, we can simultaneously reflect the structural dependency strength and attribute semantic similarity between components, thereby avoiding mapping bias caused by relying on only a single dimension.

[0033] A3: Construct a weighted adjacency matrix based on the edge weights and the set of nodes. If there is an edge between node pairs, the matrix elements are the corresponding edge weights; otherwise, the matrix elements are zero. Specifically, constructing a weighted adjacency matrix This is manifested in: If node With nodes If there is an edge, then If no edge exists, then .

[0034] A4: Use the component feature vector as the initial feature of the node, and use graph coding technology to perform feature aggregation and update based on the weighted adjacency matrix to generate the node representation; Specifically, using component feature vectors As the initial feature of the nodes, the weighted adjacency matrix As input to a graph structure, graph convolutional networks or neighbor aggregation networks can be used to learn the representations of nodes; For example, the node representation is updated using graph convolution; The update process involves summing the node's own features with the features of its neighboring nodes according to edge weights, then performing a linear transformation and activation function to obtain the node's structural representation. Node representation It also includes the node's own attribute characteristics, neighbor node characteristics, and upstream and downstream dependency information.

[0035] After graph convolution or neighbor aggregation with a preset number of layers, the representation set of all nodes is obtained: in, Indicates the first The structural semantic representation of each component in the component relationship diagram.

[0036] By modeling the structure shown in the diagram and generating node representations, we have achieved joint modeling of component structure, semantics, and temporal features. Compared with the traditional approach that only matches attribute features, this solution can integrate neighbor node information and topological dependencies by generating node representations, thereby improving the structural awareness capability of component matching.

[0037] Furthermore, by calculating edge weights and combining dependency strength with feature similarity, the accuracy of the modeling can be ensured, providing a robust feature representation foundation for subsequent end-to-end mapping and optimization.

[0038] In step S3, adaptive matching of components is performed based on the node representation to determine the component mapping relationship and mapping priority parameters; It should be noted that in step S3, the source component and the target component are further matched and scored based on the component attribute features and structural semantic representation, and a mapping relationship is generated under dependency constraints.

[0039] In an optional implementation, step S3 above includes the following sub-steps B1-B6: B1: Determine the source component set and target component set from the system components; Based on system migration, refactoring, or mapping requirements, the system components to be matched are divided into source and target sides; The source component set is represented as , The target component set is represented as ; Each source component and target components All components have been processed in steps S1 and S2, and each has a corresponding component feature vector and node representation, thereby clarifying the matching range and object.

[0040] B2: Traverse the source component set and the target component set to generate candidate matching pairs; Furthermore, for each source component Traverse all target components Calculate attribute similarity and structural similarity to form candidate matching pairs. , Preferably, to reduce computational load, a coarse screening can be performed first based on basic attributes such as component type or number of interfaces. Then, candidate matching pairs can be generated for component pairs that only meet the coarse screening criteria. .

[0041] B3: Calculate the attribute similarity of candidate matching pairs based on component feature vectors, and calculate the structural similarity of candidate matching pairs based on node representations; For each candidate matching pair Calculate the following two similarities respectively: First, attribute similarity is based on the component feature vector generated in step S1. and The cosine similarity is used for calculation, and is expressed as follows: This reflects the degree of similarity between components in terms of static properties, runtime behavior, and other characteristics; Secondly, structural similarity is based on the node representations generated in step S2. and The cosine similarity is used for calculation, and is expressed as: This reflects the degree of similarity of components in terms of structural semantics such as topology and upstream / downstream dependencies; In addition, by jointly calculating the two similarities mentioned above, we can avoid matching bias caused by relying on only a single dimension, thereby avoiding matching errors where the attributes are similar but the structural positions are different.

[0042] B4: The initial matching score is obtained by weighting the attribute similarity and structural similarity. For each candidate matching pair We perform weighted calculations to determine the initial matching score, which is expressed as: in, The balancing coefficient is a preset parameter used to adjust the relative importance of structural similarity and attribute similarity. In scenarios requiring high structural stability, a larger value can be set. value.

[0043] B5: Generate a dependency consistency correction value based on the consistency of the dependency direction of the matched related components on both sides of the candidate matching pair; To avoid dependency chain breaks caused by matching based solely on the local features of a single component, global dependency constraints are implemented, and candidate matching pairs are checked. The matching status of neighboring components on both sides; Furthermore, if the source component If the upstream or downstream related components have already formed a match with the corresponding components on the target side, and the upstream and downstream directions after the match are consistent, the consistency correction value of the candidate matching pair can be increased accordingly; if the directions are inconsistent or the dependency chain is broken, the consistency correction value can be decreased.

[0044] Furthermore, relying on consistency correction values The determination is made using dynamic calculation, based on the consistency of the upstream and downstream directions of the matched components on both sides of the candidate matching pair. Preferably, relying on consistency correction values This is expressed as a correction amount that is positively correlated with the number of items in the same direction and negatively correlated with the number of items in the opposite direction or the number of broken dependency chains, and is used to correct the initial matching score; Specifically, the system counts the number of candidate matching pairs among the matched related components on both sides that can maintain consistency in the upstream and downstream directions. The number of things that are inconsistent with the direction of existence To judge, it is reflected in: like > If so, a positive correction value is generated to improve the initial matching score; like < If so, a negative correction value is generated to reduce the initial matching score.

[0045] For example, if the source component In the process, there are a total of 3 matched related components. Among them, 2 components are consistent with the upstream and downstream directions of the component corresponding to the target side, and 1 component is inconsistent in direction. This can improve the consistency correction value of the candidate matching pair. Conversely, if the number of consistent components decreases or there is a broken dependency chain, the consistency correction value of the candidate matching pair can be reduced to ensure the structural coherence of the mapping results.

[0046] B6: Correct the initial matching score based on the dependency consistency correction value to obtain the corrected matching score, and determine the component mapping relationship and mapping priority parameters based on the corrected matching score; The initial matching score is adjusted based on the dependency consistency correction value, and is expressed as follows: in, The dependency constraint correction coefficient is a preset parameter used to control the weight of dependency consistency on the final score. Furthermore, the matching score matrix is ​​represented as: For each source component The target component with the highest score that meets a preset threshold is selected from all target components as the mapping result. If multiple target components have similar scores, the highest score is retained. Each target component is used as a candidate mapping relationship, and a priority order is generated according to the score. A mapping priority parameter is generated according to the score, which is used to characterize the confidence of the mapping relationship. The higher the score, the higher the priority.

[0047] By integrating attribute features and structural semantic representations, multi-dimensional adaptive matching of components is achieved. Furthermore, by introducing a dependency consistency correction mechanism, the problem of dependency chain breakage that is prone to occur in traditional matching methods is effectively avoided. In this way, the coherence and accuracy of the mapping relationship in the entire link are improved, providing highly reliable component alignment results for subsequent path optimization.

[0048] In step S4, candidate call paths are generated based on the mapping relationship, and target execution paths are selected by combining preset system operation constraints and path evaluation weights. It should be noted that in step S4, the input component relationship diagram is... The node represents a set. Based on component mapping relationships and component relationship graphs, multiple feasible full-link call paths are generated, and the target execution path with the best overall performance is evaluated and selected to provide a specific scheduling scheme for the actual operation of the system.

[0049] In an optional implementation, step S4 above includes the following sub-steps C1-C4: C1: Generate candidate call paths that satisfy preset system operation constraints based on component mapping relationships and component dependencies; Based on the mapping relationship between the source component and the target component determined in step S3, and combined with the directed edge dependency order in the component relationship graph constructed in step S2, starting from the business initiation component, the system expands layer by layer along the dependency direction to generate multiple feasible call paths. The preset system operation constraints mainly include dependency order constraints and no-circular dependency constraints. The generated paths need to follow the component relationship graph. The directed edges in the path are subject to the order of dependencies, meaning the path cannot contain closed loops, i.e., there are no cyclic dependencies.

[0050] Preferably, a depth-first search or a breadth-first search is used to generate candidate paths, and paths that do not satisfy the dependency order or have circular dependencies are eliminated during the search process. Each candidate path is represented as follows: in, Indicates the first Candidate paths, This indicates the component nodes contained in the path. Indicates the number of nodes contained in the path; The above processing ensures that the generated candidate paths are topologically connected and conform to business logic.

[0051] C2: Statistically analyze the execution latency, resource consumption, and path stability of candidate call paths; For each candidate call path Based on historical operation logs or real-time monitoring systems, performance metrics are calculated, specifically including execution latency. Resource consumption and path stability ; Among them, execution delay It is obtained by summing the average execution time of each component in the path. Resource consumption It is obtained by accumulating the resource usage of each component in the path. Path stability It is calculated from the historical success rate or failure rate of each component in the path.

[0052] C3: Combine the preset path evaluation weights to perform a weighted calculation of execution latency, resource consumption, and path stability to obtain the comprehensive cost; Furthermore, regarding candidate paths The overall cost is calculated and expressed as: in, Here, represents the path evaluation weights, and represents the preset parameters that satisfy . , It can be dynamically updated based on operational feedback in step S5 to achieve adaptive optimization; In addition, the path evaluation weights can be configured according to the emphasis of the actual business scenario.

[0053] C4: Determine the candidate call path with the lowest overall cost as the target execution path.

[0054] From all candidate paths, the path with the lowest overall cost is selected as the target execution path, that is, the optimal execution path: Furthermore, follow the target execution path The order of each node in the graph, combined with the dependencies in the component relationship graph, generates a component scheduling sequence to guide the order of subsequent component calls.

[0055] By setting up a multi-objective weighted evaluation mechanism, the problems of resource imbalance or stability degradation caused by traditional single-index optimization can be avoided. Furthermore, by combining preset constraints to generate candidate paths, the business feasibility of the paths can be ensured. By comparing and minimizing the comprehensive cost, the target path is selected, realizing the globally optimal path of the entire process, thereby improving the efficiency and stability of system operation.

[0056] In step S5, components are invoked according to the target execution path, running feedback information is collected, and feature weights, mapping priority parameters, and path evaluation weights are dynamically updated based on the running feedback information. It should be noted that in step S5, the main task is to collect the actual running data of the target execution path, evaluate the performance of the current mapping and path scheme, and dynamically adjust the feature weights, mapping priorities and path evaluation parameters so that the system can adapt to component performance degradation or dependent path reconstruction and complete the optimization.

[0057] In an optional implementation, step S5 above includes the following sub-steps D1-D3: D1: Calculate the feedback evaluation value based on the operation feedback information, and compare the feedback evaluation value with the preset threshold; The operational feedback information includes average execution latency. Execution success rate Number of abnormalities or intensity of abnormalities Indicators such as; Preferably, operational feedback information can be obtained through system logs, monitoring probes, or link tracing tools.

[0058] Analyze the operational feedback data within the current execution cycle and unify different metrics into a single feedback evaluation value: in, The weights corresponding to the feedback indicators are preset parameters. If the feedback evaluation value... The larger the value, the worse the performance of the current mapping relationship or path scheme.

[0059] It should be further noted that the preset threshold It can be set based on the statistical distribution of historical operating data or the service quality requirements of the business, through preset thresholds. This allows us to determine whether the current performance is acceptable, thereby quantifying the quality of the operational feedback and providing a basis for subsequent parameter adjustments.

[0060] D2: Adjust the mapping priority parameter based on the comparison results; Furthermore, for the mapping relationships involved in the current execution path, based on the feedback evaluation value... Its priority is adjusted, preferably when Greater than the preset threshold When, reduce the priority of the current mapping relationship, when Less than the preset threshold When the operation is good, the priority of the current mapping relationship is increased.

[0061] Preferably, let the current mapping relationship be set. The original priority is Feedback evaluation value The corrected priority It can be determined in the following way: in, The preset adjustment coefficient is, and >0.

[0062] Furthermore, when When, reduce the priority of the current mapping relationship, when At that time, the priority of the current mapping relationship is increased, thereby realizing the linear adjustment of the mapping priority according to the running performance.

[0063] For example, suppose ; when hour, That is, reduce the priority of the mapping relationship; when hour, That is, increase the priority of this mapping relationship; After correction, the priority of all mapping relationships can be normalized to keep them within a uniform numerical range. This method makes subsequent component matching more inclined to select mapping relationships with stable historical performance.

[0064] D3: Adjust the path evaluation weight and feature weight based on the operational feedback evaluation value; Furthermore, based on the feedback evaluation value Weight parameters in path evaluation Adjustments will be made; specifically, when execution latency remains consistently high, the weight of the latency item will be increased. When resource usage fluctuates significantly, increase the weight of the corresponding resource item. When the anomaly rate increases, the weight corresponding to the stability term is increased. .

[0065] Furthermore, based on the results of the operational feedback analysis, the feature weights in step S1 are adjusted. Update the system and generate corresponding weight adjustments based on the mapping deviations or path performance issues reflected in the feedback. For example, if the updated rule includes a rule that increases the matching score when the component types are consistent, then the weight of the feature item corresponding to the component type is increased; if it includes a rule that decreases the matching score when the upstream and downstream directions are inconsistent, then the weight of the feature item corresponding to the consistency of the dependency direction is decreased. Preferably, let the first Each feature dimension is denoted as a feature term. Its original weight is The adjustment amount generated by rule feedback is The updated feature weights can then be expressed as: Among them, weight adjustment amount It can be determined based on feedback rules or historical optimization experience.

[0066] Preferably, in step S6, the generated structured rules can also be used to extract corresponding feature identifiers based on the feature terms involved, and generate... .

[0067] Furthermore, the updated mapping priority and path evaluation parameters can be written into the parameter set and input into the next stage for component matching and path search, thereby forming a closed-loop optimization that enables component feature encoding to adaptively highlight feature dimensions that are more important to the current operating environment.

[0068] By dynamically updating the parameters driven by runtime feedback in the above manner, the problem that traditional static mapping models cannot respond to component performance degradation and dependency path reconstruction is solved. By adjusting the mapping priority, path weight, and feature weight in a differentiated manner, the parameters of the entire link can be optimized collaboratively, thereby improving the mapping stability, path rationality, and global optimization performance of the system in dynamic evolution scenarios.

[0069] In step S6, the weighted digital encoding, component adaptive matching, and target execution path filtering are re-executed based on the updated weights and parameters; It should be noted that in step S6, the optimized mapping relationship and path selection results are converted into a rule set, and the updated rules and parameters are fed back to the preceding stage to form a closed-loop optimization mechanism.

[0070] In an optional implementation, a structured rule set is generated based on the adjusted mapping priority parameters, path evaluation weights, and feature weights. The structured rule set is then merged with the historical rule set, the rule base is updated, and a back injection is performed. Furthermore, the mapping relationships, path selection results, and feedback correction results obtained within the current period are organized into rule entries to form a rule set, represented as follows: , in, Indicates the first The rules include component matching priority rules, dependency consistency correction rules, path filtering rules, and parameter adjustment rules. Furthermore, this will include the set of rules added or revised in this round. With historical rules set The merged set yields the updated set of rules, specifically represented as follows: Update the rule set The mapping priority parameters and path evaluation weights are fed back to the preceding stages. Specifically, when fed back to step S1, they are used to adjust the feature weights; when fed back to step S3, they are used to adjust the mapping priority; and when fed back to step S4, they are used to adjust the path evaluation parameters.

[0071] In the next round of processing, the updated rules and parameters are called in the previous stage to re-execute feature modeling, component matching and path search, realizing a closed-loop optimization process of running, feedback, updating and running again; For example, in a certain round of processing, step S1 completes candidate component matching based on the current feature weights, step S3 determines that a certain mapping relationship has a low priority based on the mapping execution result, and step S4 finds that "consistency in upstream and downstream directions" has a significant impact on the final result based on the path execution result. Based on the above feedback, a new rule is added in step S6: "Increase the matching score when the upstream and downstream directions are consistent", and the weight of the feature item corresponding to "consistency of dependency direction" is increased. Furthermore, in the next round of processing, step S1 directly recalculates the matching score using the updated feature weights, step S3 reselects the mapping relationship using the updated mapping priority, and step S4 re-evaluates the path using the updated path evaluation parameters.

[0072] By implementing the above methods, rule updates and closed-loop back-injection are achieved, solving the problem that traditional static mapping models cannot respond to the dynamic evolution of the system. The operational feedback is transformed into structured rules and back-injected into the preceding stages, realizing the adaptive adjustment of system parameters and knowledge accumulation. This improves the response speed, adaptability, and global optimization performance of the mapping system in complex engineering environments, enabling the system to evolve automatically.

[0073] Example 2 is an embodiment of the present invention, which provides an example of a full-link component adaptive mapping and optimization method integrating digital graph coding, based on the previous embodiment, with a distributed power dispatch management platform as the application scenario.

[0074] In addition, the power dispatch management platform includes multiple business components such as user request access service, generation dispatch service, load management service, metering and settlement service, alarm notification service, resource monitoring service, and data storage service. These components work together through interface calls, message transmission, and data dependencies to complete business processes such as power dispatch, load allocation, metering and settlement, and abnormal alarms.

[0075] In this embodiment, assuming that the source system needs to be migrated or reconstructed to the target system, an adaptive mapping relationship needs to be established between the source components and the target components to determine the target execution path that meets the operational constraints.

[0076] The specific processing procedure is as follows: In step S1, component features are extracted and digitally encoded. Specifically, it collects basic information, interface definition information, runtime log information, and dependency description information of each component in the source system and the target system; For example, for the power generation dispatch service component, its component type is dispatch class, version number is V3.2, number of interfaces is 12, average number of parameters is 5, average daily call frequency is 23,000 times, average execution time is 180ms, number of exceptions is 7, resource utilization rate is 63%, number of upstream dependencies is 3, number of downstream calls is 5, and dependency level depth is 4.

[0077] Furthermore, the aforementioned original features are extracted for each component to form a component original feature sequence; Subsequently, features of different dimensions are mapped to a unified numerical range using range standardization, and then component feature vectors are generated by combining preset feature weights. For example, the standardized features of the power generation dispatch service are multiplied dimension by dimension to obtain the digital feature representation of the component. Then, the same processing is performed on all components to construct the component feature matrix. Simultaneously, using each column of the feature matrix as a sample sequence, the correlation between each feature dimension is calculated, thus obtaining the feature correlation matrix. .

[0078] Through the processing in step S1, we can obtain the feature representation of all components after unified numericalization, thus providing the basic input for subsequent graph structure modeling and component matching.

[0079] In step S2, graph structure modeling and node representation generation are performed; Specifically, in obtaining the component feature matrix Then, a component relationship graph is constructed based on the calling relationships, dependencies, and data interaction relationships between components. .

[0080] For example, a user requests access to the service and calls the generation dispatch service, which in turn calls the load management service, which in turn calls the metering and settlement service and the alarm notification service, thus establishing corresponding directed edges in the graph.

[0081] Furthermore, for component pairs with dependencies, edge weights are calculated by combining dependency strength and feature similarity. For example, if the generation dispatch service and the load management service call each other frequently and have a large amount of data interaction, then their dependency strength is high. If the two components are similar in terms of interface size, runtime latency, and resource consumption, then their corresponding edge weights are larger, and a weighted adjacency matrix can be constructed based on the edge weights of all component pairs. .

[0082] Furthermore, using component feature vectors as initial node features and weighted adjacency matrices... As input to the graph structure, graph convolutional network technology is used to aggregate and update the nodes in the graph, so that each node representation can not only reflect the component's own attributes, but also include the characteristics of its neighboring nodes and upstream and downstream dependency information. After graph convolution with a preset number of layers, the set of node representations corresponding to each component can be obtained. Thus, a component representation result that simultaneously includes attribute semantics and structural semantics can be obtained.

[0083] In step S3, component matching and dependency consistency mapping are performed; Specifically, the source-side system component set is used as the object to be mapped, and the target-side system component set is used as the matching object; For example, the source-side system contains components such as power generation dispatch service A, load management service A, and alarm notification service A, while the target-side system contains components such as dispatch control component B, load analysis component B, and alarm push component B.

[0084] Furthermore, for each source component, all target components are traversed, and attribute similarity and structural similarity are calculated respectively to generate candidate matching pairs; For example, the cosine similarity of the feature vectors of the power generation dispatch service A and the dispatch control component B is calculated to obtain the attribute similarity, and then the cosine similarity of their node representations is calculated to obtain the structural similarity. Finally, the two are weighted together to form an initial matching score.

[0085] Furthermore, to avoid relying solely on local features of a single component for matching, it is also necessary to calculate a dependency consistency correction value for candidate matching pairs.

[0086] For example, if the upstream component of the source component power generation scheduling service A, the user request access service A has already formed a match with the target side access control component B, and the upstream and downstream directional relationships on both sides are consistent, then the consistency correction value of the candidate matching pair from power generation scheduling service A to scheduling control component B is increased; conversely, if the directions are inconsistent or there is a broken dependency chain, the consistency correction value is decreased, and then the initial matching score is corrected according to the correction value to obtain the corrected matching score.

[0087] Furthermore, for each source component, the target component with the highest score that exceeds a preset threshold is selected as the final mapping result; if multiple target components have similar scores, the top one is retained. Each candidate mapping relationship is selected and prioritized according to its score. For example, the final mapping relationships can be obtained from generation dispatch service A to dispatch control component B, load management service A to load analysis component B, alarm notification service A to alarm push component B, etc.

[0088] In step S4, candidate links are generated and optimal paths are searched. Specifically, based on the aforementioned component mapping relationship and component relationship diagram Generate candidate call paths; For example, for a power grid load adjustment request, a candidate execution path can be formed: from access control component B to dispatch control component B to load analysis component B to metering and settlement component B to alarm push component B. Other candidate paths can also be formed by combining different mapping relationships.

[0089] Furthermore, for each candidate path, the execution latency, resource consumption, and path stability are calculated separately. Among them, the execution latency can be obtained by summing the average execution time of each component in the path, the resource consumption can be obtained by summing the CPU, memory or bandwidth usage of each component in the path, and the path stability can be obtained by converting the historical success rate or failure rate of each component in the path.

[0090] Furthermore, the comprehensive cost of each candidate path is calculated based on the preset evaluation weights, and the path with the smallest comprehensive cost is selected as the target execution path. For example, after calculation, it is found that the path from access control component B to scheduling control component B to load analysis component B to metering and settlement component B has a lower overall cost, and thus it is determined as the target execution path. Then, combined with the sequential dependencies of the components in the path, a corresponding scheduling sequence is generated to guide the subsequent component call order.

[0091] In step S5, parameter updates are performed based on runtime feedback. Specifically, after the system actually runs according to the optimal path, the system collects the running feedback information within the current execution cycle, including average execution latency, execution success rate, number of exceptions, number of retries, and resource fluctuations. For example, within a certain scheduling period, the average execution latency was 240ms, the execution success rate was 92%, and the number of exceptions was 4. Unify different feedback indicators into a single feedback evaluation value. If the feedback evaluation value is large, it indicates that the current mapping relationship or path scheme is performing poorly; if the feedback evaluation value is small, it indicates that the performance is good.

[0092] Furthermore, the priority of the mapping relationships involved in the current execution path is adjusted based on the feedback evaluation value; For example, when the feedback evaluation value is higher than a preset threshold, the priority of the power generation dispatch service A to the dispatch control component B is reduced, and when the feedback evaluation value is lower than the preset threshold, its priority is increased.

[0093] Furthermore, the weights of each item in the path evaluation are adjusted based on the feedback information; If the execution latency is high in multiple consecutive cycles, the weight of the latency item will be increased. If resources fluctuate significantly, increase the weight of the resource consumption item. If the anomaly rate increases, the weight of the stability term is increased.

[0094] After the above processing, the updated mapping priority and path evaluation weight will be written into the parameter set for the next stage of component matching and path search.

[0095] In step S6, rule updates and closed-loop injection are performed. Specifically, based on the mapping relationships, path selection results, and feedback correction results obtained in this round of processing, structured rule entries are generated to form a rule set; For example, the following rules can be generated: When the component types are the same and the difference in the number of interfaces is less than a preset threshold, the matching score is increased. When the upstream and downstream directions are consistent, increase the dependency consistency correction value; When the execution latency is continuously higher than the threshold, the latency weight in path evaluation is increased.

[0096] Furthermore, the set of rules newly added or revised in this round is merged with the set of historical rules to obtain an updated rule base. The updated rule set, mapping priority parameters, and path evaluation weights are then fed back to the aforementioned stages. Specifically, when feedback is sent to stage one, the corresponding feature items are extracted from the rules, and the corresponding weight adjustment amount is generated to update the feature weights; When feedback is sent to stage three, it is used to adjust the priority order of candidate mapping relationships; When feedback is sent to stage four, the evaluation parameters for adjusting the overall path cost are used.

[0097] For example, if it is found in a certain round of processing that the consistency of upstream and downstream directions significantly improves the accuracy of mapping results, then a new rule is added to the rule base to improve the matching score when the upstream and downstream directions are consistent, and the weight of the feature items corresponding to the consistency of the dependent direction is increased.

[0098] In the next round of processing, the system can directly call the updated rules and parameters to re-execute feature modeling, component matching, and path search, forming a closed-loop optimization process of running, feedback, updating, and running again.

[0099] In summary, this solution does not rely on pre-specified component mapping relationships or fixed paths. Instead, it starts from basic component information, interface information, runtime logs, and dependencies to gradually generate component feature representations, relationship graph structures, candidate mapping relationships, target execution paths, and feedback update rules. In this way, it is possible to adaptively establish and dynamically optimize component mapping relationships in complex and heterogeneous software systems, thereby improving the accuracy of component matching, the rationality of path scheduling, and the stability of system operation.

[0100] Example 3, referring to Figure 1 As one embodiment of the present invention, this embodiment provides a full-link component adaptive mapping and optimization system integrating digital graph coding, comprising: The feature encoding module is used to collect feature information of system components, extract features, and perform weighted digital encoding based on preset feature weights. The graph model construction module is used to construct a component relationship graph based on the digitally encoded features and component dependencies, and to generate node representations using graph encoding technology. The component matching module is used to perform adaptive component matching based on node representation, and to determine the component mapping relationship and mapping priority parameters. The path optimization module is used to generate candidate call paths based on mapping relationships and filter target execution paths by combining preset system operation constraints and path evaluation weights. The parameter update module is used to call components according to the target execution path, collect runtime feedback information, and dynamically update feature weights, mapping priority parameters, and path evaluation weights based on the runtime feedback information. The closed-loop evolution module is used to re-execute weighted digital encoding, component adaptive matching, and target execution path selection based on the updated weights and parameters.

[0101] This embodiment also provides an electronic device applicable to a method for adaptive mapping and optimization of end-to-end components that integrates digital graph coding, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the method for adaptive mapping and optimization of end-to-end components that integrates digital graph coding as proposed in the above embodiment.

[0102] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements a full-link component adaptive mapping and optimization method integrating digital graph coding as proposed in the above embodiments.

[0103] The storage medium proposed in this embodiment and the method for implementing an adaptive mapping and optimization of end-link components with integrated digital graph coding proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0104] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0105] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for adaptive mapping and optimization of end-to-end components integrating digital graph coding, characterized in that: include, Collect feature information of system components, extract features, and perform weighted digital encoding based on preset feature weights; Based on the digitally encoded features and component dependencies, a component relationship graph is constructed, and node representations are generated using graph encoding technology. Based on the node representation, adaptive component matching is performed to determine the component mapping relationship and mapping priority parameters; Based on the mapping relationship, candidate call paths are generated, and target execution paths are selected by combining preset system operation constraints and path evaluation weights. Components are invoked according to the target execution path, running feedback information is collected, and feature weights, mapping priority parameters and path evaluation weights are dynamically updated based on the running feedback information. Based on the updated weights and parameters, the weighted digitization encoding, component adaptive matching, and target execution path selection are re-executed.

2. The end-to-end component adaptive mapping and optimization method integrating digital graph coding as described in claim 1, characterized in that, The weighted digitization encoding includes: The feature information is standardized to map features of different dimensions to a unified numerical range, and component feature vectors are generated by combining preset feature weights.

3. The end-to-end component adaptive mapping and optimization method integrating digital graph coding as described in claim 2, characterized in that, The step of constructing a component relationship graph based on the digitized encoded features and component dependencies, and generating node representations using graph encoding technology, includes: Construct a component relationship graph containing a set of nodes and a set of edges based on component dependencies; The feature similarity between component pairs with dependencies is calculated based on the component feature vectors, and the edge weights are determined by combining the dependency strength and feature similarity. A weighted adjacency matrix is ​​constructed based on the edge weights and the set of nodes. If there is an edge between node pairs, the matrix element takes the corresponding edge weight; otherwise, the matrix element is zero. The component feature vector is used as the initial feature of the node, and the feature is aggregated and updated using graph coding technology based on the weighted adjacency matrix to generate the node representation.

4. The end-to-end component adaptive mapping and optimization method integrating digital graph coding as described in claim 3, characterized in that, The adaptive matching of the components includes: Determine the source component set and the target component set from the system components; Traverse the source component set and the target component set to generate candidate matching pairs; The attribute similarity of the candidate matching pairs is calculated based on the component feature vectors, and the structural similarity of the candidate matching pairs is calculated based on the node representations. The initial matching score is obtained by weighting the attribute similarity and structural similarity. Based on the consistency of the dependency direction of the matched associated components on both sides of the candidate matching pair, a dependency consistency correction value is generated; The initial matching score is corrected based on the dependency consistency correction value to obtain the corrected matching score, and the component mapping relationship and mapping priority parameters are determined based on the corrected matching score.

5. The end-to-end component adaptive mapping and optimization method integrating digital graph coding as described in claim 4, characterized in that, The process of generating candidate call paths based on the mapping relationship, and filtering target execution paths by combining preset system operation constraints and path evaluation weights, includes: Based on the component mapping relationship and the component dependency relationship, generate candidate call paths that satisfy preset system operation constraints; The execution latency, resource consumption, and path stability of the candidate call paths are statistically analyzed. The execution latency, resource consumption, and path stability are weighted and calculated based on preset path evaluation weights to obtain a comprehensive cost. The candidate call path with the lowest overall cost is determined as the target execution path.

6. The end-to-end component adaptive mapping and optimization method integrating digital graph coding as described in claim 5, characterized in that, The step of dynamically updating the feature weights, mapping priority parameters, and path evaluation weights based on the operational feedback information includes: Calculate a feedback evaluation value based on the operation feedback information, and compare the feedback evaluation value with a preset threshold. Adjust the mapping priority parameter based on the comparison result; The path evaluation weight and the feature weight are adjusted based on the operational feedback evaluation value.

7. The end-to-end component adaptive mapping and optimization method integrating digital graph coding as described in claim 4, characterized in that, The process of re-executing weighted digital encoding, adaptive component matching, and target execution path filtering based on the updated weights and parameters includes: A structured rule set is generated based on the adjusted mapping priority parameters, path evaluation weights, and feature weights. The structured rule set is then merged with the historical rule set, the rule base is updated, and a back injection is performed.

8. A full-link component adaptive mapping and optimization system integrating digital graph coding, using the method described in any one of claims 1 to 7, characterized in that, include: The feature encoding module is used to collect feature information of system components, extract features, and perform weighted digital encoding based on preset feature weights. The graph model construction module is used to construct a component relationship graph based on the digitally encoded features and component dependencies, and to generate node representations using graph encoding technology. The component matching module is used to perform adaptive component matching based on the node representation, and determine the component mapping relationship and mapping priority parameters; The path optimization module is used to generate candidate call paths based on the mapping relationship, and filter target execution paths by combining preset system operation constraints and path evaluation weights. The parameter update module is used to call components according to the target execution path, collect running feedback information, and dynamically update feature weights, mapping priority parameters, and path evaluation weights based on the running feedback information. The closed-loop evolution module is used to re-execute weighted digital encoding, component adaptive matching, and target execution path selection based on the updated weights and parameters.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the end-to-end component adaptive mapping and optimization method that integrates digital graph coding as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the end-to-end component adaptive mapping and optimization method that integrates digital graph coding as described in any one of claims 1 to 7.

Images