A method, system, device and medium for intelligent mapping and process reconstruction of atomized components based on multi-layer graph embedding

Abstract

The application discloses a kind of based on multi-layer graph embedding atomization component intelligent mapping and process reconstruction method, system, equipment and medium, belong to component reconstruction technical field, comprising: component feature matrix is constructed by multi-source feature extraction, and based on this, multi-layer dependency graph is established to learn component embedding representation;Similarity matching and mapping are carried out using embedding, and the mapping result is optimized in combination with dependency consistency constraint, and then the process dependency topology is constructed and optimized to eliminate redundancy and cycle;Adaptively update dependency graph by detecting historical changes, and introduce expert knowledge constraint to correct and jointly optimize the result, and feed back the revised knowledge to knowledge base.The application realizes unified modeling and deep fusion of component structure, semantics and execution layer dependency relationship by multi-layer graph embedding technology, improves the precision and panoramic understanding ability of component mapping;Ensure the timeliness and accuracy of mapping and reconstruction result.

Description

Technical Field

[0001] This invention relates to the field of component reconfiguration technology, and specifically to a method, system, device, and medium for intelligent mapping and process reconfiguration of atomic components based on multi-layer graph embedding. Background Technology

[0002] Despite the deepening research on software component-based refactoring and automated process mapping in recent years, encompassing various methods from static dependency analysis and feature matching to machine learning and knowledge graph-based approaches, current mainstream technologies still have many limitations, as follows: Most existing methods generally lack the ability to uniformly model multi-level dependencies and cross-level structural features. For example, existing research usually abstracts the calling relationship between components into a single-level directed graph or a flat dependency table, which can only describe direct input-output relationships and cannot effectively characterize multi-level coupling, cross-level dependencies, and hierarchical semantic mapping between functional blocks in the process. In real-world scenarios, such as complex enterprise-level business systems, a process node often depends on multiple sub-components, and these components may have multi-hop dependencies and upstream and downstream feedback paths.

[0003] Existing technologies in most studies rely on static code analysis or fixed semantic labels, assuming that the semantics of component functions remain stable during the training or matching phases. However, modern software systems are characterized by continuous iteration and module evolution; component interface parameters, dependency topologies, and even execution logic are dynamically updated with version changes.

[0004] Existing technologies are insufficient for cross-scenario migration and large-scale online system adaptation. For example, current process mapping tools mostly adopt offline parsing and periodic synchronization strategies, which cannot meet the real-time refactoring requirements under high concurrency, cross-module, and distributed architectures. When the system iterates rapidly and components are frequently replaced, the static model cannot capture the generation of new dependencies and the degradation of old dependencies, resulting in frequent problems such as mapping delays, refactoring failures, or inconsistent execution paths. Summary of the Invention

[0005] In view of the above-mentioned problems, the present invention provides a method, system, device and medium for intelligent mapping and process reconfiguration of atomic components based on multi-layer graph embedding.

[0006] Therefore, the technical problem solved by this invention is: This invention aims to solve the problems of single dependency modeling of software components, fragmented feature expression and lack of evolution perception in the prior art, and provides an atomic component intelligent mapping and process reconstruction method that can unify the modeling of multi-layer dependency structure, integrate multi-dimensional component features, and have dynamic evolution perception and adaptive adjustment capabilities, so as to overcome the technical problems of low mapping accuracy, poor adaptability and inability to adapt to the continuous evolution of components in traditional methods.

[0007] To address the aforementioned technical problems, this invention provides the following technical solution: a method for intelligent mapping and process refactoring of atomic components based on multi-layer graph embedding, comprising, By parsing and extracting features from source code, configuration files, and dynamic call logs, a multi-layer component feature matrix containing structural, semantic, and behavioral features is constructed.

[0008] Based on the multi-layer component feature matrix, a three-layer dependency graph of structure, semantics and behavior is constructed, and a hierarchical embedding representation of the components is generated by learning through a graph embedding network.

[0009] Similarity matching and mapping between components are performed based on hierarchical embedding representation, and local dependency consistency constraints are introduced to optimize the mapping results.

[0010] Based on the mapping results and the original dependencies, the process dependency topology is constructed and optimized, and the optimized flowchart is generated by minimizing path redundancy and circular dependencies.

[0011] Changes in the multi-layer dependency graph are detected based on historical data, and the dependency graph and weights are updated adaptively.

[0012] By introducing expert knowledge constraints, the mapping and process refactoring results are corrected and jointly optimized, and the corrected knowledge is fed back to the knowledge base.

[0013] As a preferred embodiment of the atomic component intelligent mapping and process reconstruction method based on multi-layer graph embedding described in this invention, the construction of a multi-layer component feature matrix including structural, semantic, and behavioral features includes, Static analysis of the source code and interface configuration is performed to extract the static features of the components.

[0014] Extract behavioral characteristic statistics of components based on dynamic call logs.

[0015] The static features and behavioral features are concatenated and normalized to obtain a unified multi-layer component feature matrix.

[0016] As a preferred embodiment of the atomic component intelligent mapping and process reconstructing method based on multi-layer graph embedding described in this invention, the construction of a three-layer dependency graph (structure, semantics, and behavior), and the generation of hierarchical embedding representations of components through a graph embedding network, includes: Based on the calling relationships, semantic dependencies, and dynamic behavior associations between components, dependency graphs are constructed in three dimensions: structure, semantics, and behavior.

[0017] By using graph embedding networks, the vector representation of a node in each layer is iteratively learned by aggregating the neighbor feature information of nodes in the dependency graph of each layer.

[0018] The embedding vectors learned by the components at the structural, semantic, and behavioral layers are concatenated to generate a comprehensive hierarchical embedding representation.

[0019] As a preferred embodiment of the atomic component intelligent mapping and process refactoring method based on multi-layer graph embedding described in this invention, the step of performing similarity matching and mapping between components based on hierarchical embedding representation, and introducing local dependency consistency constraints to optimize the mapping results includes, Based on the hierarchical embedding representation and the structural information of the unified dependency graph, the overall matching degree between components is calculated.

[0020] Based on the comprehensive matching degree, the optimal matching criterion is used to find the best mapping component in the target system for the component in the source system, and form an initial mapping relationship.

[0021] Apply local dependency consistency constraints to the initial mapping relationship, optimize and output the final component mapping matrix.

[0022] As a preferred embodiment of the atomic component intelligent mapping and process refactoring method based on multi-layer graph embedding described in this invention, the step of constructing and optimizing the process dependency topology structure, and generating an optimized flowchart by minimizing path redundancy and cyclic dependencies, includes: Based on the component execution sequence and mapping relationship, a new process dependency matrix is ​​constructed to represent the reconstructed process topology.

[0023] The design optimization objective is used to solve the optimization objective through a graph optimization strategy, generating a flowchart that achieves optimal path efficiency and structural consistency.

[0024] The beneficial effects of this preferred technical solution are that, by establishing a clear process dependency matrix and setting a solvable graph optimization objective, the originally complex process refactoring problem is transformed into a structured mathematical optimization problem. It can systematically eliminate path redundancy and circular dependencies introduced by component mapping, not only ensuring the functional equivalence of the refactored process but also significantly improving the efficiency of execution paths and the clarity of the topology at the graph structure level.

[0025] As a preferred embodiment of the intelligent mapping and process refactoring method for atomic components based on multi-layer graph embedding described in this invention, the step of detecting changes in the multi-layer dependency graph based on historical data and adaptively updating the dependency graph and weights includes: Component nodes that have experienced structural drift are detected by comparing the rate of change of node centrality in historical and current dependency graphs.

[0026] If the changes in a node exceed a predetermined threshold, an adaptive adjustment and update of the weights of the node and its associated dependent edges will be triggered.

[0027] Record the evolution process and generate an evolution report, outputting an updated multi-layer dependency graph.

[0028] The beneficial effect of this preferred technical solution is that by monitoring the temporal changes in node centrality to quantify structural drift, it achieves proactive and quantitative perception of the evolution of dependency graphs, thereby automatically identifying changes in key components and triggering precise updates, avoiding the computational overhead caused by frequent global reconstruction.

[0029] As a preferred embodiment of the atomic component intelligent mapping and process refactoring method based on multi-layer graph embedding described in this invention, the step of correcting and jointly optimizing the mapping and process refactoring results, and feeding the corrected knowledge back to the knowledge base includes: The prior knowledge of experts regarding component dependencies is formalized into constraints to evaluate and correct automatically generated mappings and process results.

[0030] The process refactoring goals, mapping consistency constraints, and expert constraints are jointly optimized to obtain the final corrected dependencies.

[0031] New and valid dependencies generated during the optimization process will be updated in the expert knowledge base.

[0032] The beneficial effect of this preferred technical solution is that by formalizing expert knowledge into computable constraints and integrating it into the joint optimization framework, it can not only correct deviations that may occur in the automated process in real time, but also ensure that the output conforms to the domain logic.

[0033] This invention provides an intelligent mapping and process reconfiguration system for atomic components based on multi-layer graph embedding.

[0034] To address the aforementioned technical problems, this invention provides the following technical solution: an intelligent mapping and process reconstruction system for atomic components based on multi-layer graph embedding, comprising: a component feature extraction and representation module, a multi-layer graph construction and embedding learning module, a component matching and mapping module, a process reconstruction and optimization module, an evolution perception and update module, and an expert feedback and closed-loop optimization module.

[0035] The component feature extraction and representation module constructs a multi-layer component feature matrix containing structural, semantic, and behavioral features by parsing and extracting features from source code, configuration files, and dynamic call logs.

[0036] The multi-layer graph construction and embedding learning module constructs a three-layer dependency graph of structure, semantics and behavior based on the multi-layer component feature matrix, and learns to generate hierarchical embedding representations of components through a graph embedding network.

[0037] The component matching and mapping module performs similarity matching and mapping between components based on hierarchical embedding representation, and introduces local dependency consistency constraints to optimize the mapping results.

[0038] The process reconstruction and optimization module constructs and optimizes the process dependency topology based on the mapping results and the original dependency relationships, and generates an optimized process flow by minimizing path redundancy and circular dependencies.

[0039] The evolution perception and update module detects changes in the multi-layer dependency graph based on historical data and adaptively updates the dependency graph and weights.

[0040] The expert feedback and closed-loop optimization module introduces expert knowledge constraints to correct and jointly optimize the mapping and process reconstruction results, and feeds the corrected knowledge back to the knowledge base.

[0041] The present invention provides a computer device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the described method for intelligent mapping and process refactoring of atomic components based on multi-layer graph embedding.

[0042] The present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the described method for intelligent mapping and process refactoring of atomic components based on multi-layer graph embedding.

[0043] The beneficial effects of this invention are as follows: By using multi-layer graph embedding technology, a unified modeling and deep fusion of component structure, semantics, and execution layer dependencies are achieved, improving the accuracy of component mapping and the ability to understand the whole picture; combined with dynamic evolution perception and adaptive update mechanism, this invention can track and respond to changes in dependency structure in real time, ensuring the timeliness and accuracy of mapping and reconstruction results; at the same time, the embedded interpretability module provides an intuitive evolution map and mapping visualization, enhancing the transparency and credibility of the reconstruction process, thereby achieving highly robust, interpretable, and adaptive intelligent component mapping and process reconstruction in a dynamically changing software environment. Attached Figure Description

[0044] 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.

[0045] Figure 1 The above is a general flowchart of an atomized component intelligent mapping and process reconstruction method based on multi-layer graph embedding, which is provided as an embodiment of the present invention.

[0046] Figure 2 This is a schematic diagram of a scheme for an intelligent mapping and process refactoring system for atomic components based on multi-layer graph embedding, provided as an embodiment of the present invention. Detailed Implementation

[0047] 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.

[0048] Example 1, referring to Figure 1 This is one embodiment of the present invention, which provides a method for intelligent mapping and process refactoring of atomic components based on multi-layer graph embedding, including: S1. By parsing and extracting features from source code, configuration files, and dynamic call logs, a multi-layer component feature matrix containing structural, semantic, and behavioral features is constructed.

[0049] S2. Based on the multi-layer component feature matrix, construct a three-layer dependency graph of structure, semantics and behavior, and learn to generate hierarchical embedding representations of components through graph embedding network.

[0050] S3. Similarity matching and mapping between components are performed based on hierarchical embedding representation, and local dependency consistency constraints are introduced to optimize the mapping results.

[0051] S4. Based on the mapping results and the original dependencies, construct and optimize the process dependency topology, and generate an optimized flowchart by minimizing path redundancy and circular dependencies.

[0052] S5. Detect changes in the multi-layer dependency graph based on historical data and adaptively update the dependency graph and weights.

[0053] S6. Introduce expert knowledge constraints to correct and jointly optimize the mapping and process refactoring results, and feed the corrected knowledge back to the knowledge base.

[0054] This invention achieves the characterization and mapping of complex panoramic dependencies by embedding the structure, semantics, and behavior dependencies of a unified modeling component into a multi-layer graph. Combined with a dynamic evolution perception mechanism, it ensures the accuracy and timeliness of mapping and reconstruction in continuous iteration. At the same time, it introduces expert knowledge constraints and closed-loop optimization, which improves the rationality and interpretability of the results and enables self-learning and continuous evolution.

[0055] Example 2, an embodiment of the present invention, provides a method for intelligent mapping and process refactoring of atomic components based on multi-layer graph embedding, based on the previous embodiment, including: In S1, the source code set is represented as The set of configuration files and interface definitions is represented as The dynamic call log collection is represented as .

[0056] Constructing a multi-layer component feature matrix that includes structural, semantic, and behavioral features includes steps A1-A3: A1. Static analysis of the source code and interface configuration is performed to extract the static features of the components.

[0057] A2. Extract behavioral characteristic statistics of components based on dynamic call logs.

[0058] This includes call frequency, execution time, and context information.

[0059] A3. The static features and behavioral features are concatenated and normalized to obtain a unified multi-layer component feature matrix.

[0060] In this application's implementation, the static analysis in A1 refers to Abstract Syntax Tree (AST) parsing, specifically, performing AST parsing and feature extraction on the source code and interface description: in, This represents the static feature representation vector corresponding to the i-th atomized component, which is used to characterize the comprehensive features of the component at the structural and semantic layers. It is the basic input for subsequent multi-layer dependency graph construction and graph embedding learning. This represents a feature extraction function based on an abstract syntax tree, used to extract the static structure and semantic information of components from source code or interface descriptions; This represents the source code or interface description text corresponding to the i-th component. Represents structural feature encoding. This represents the encoding of semantic features. This indicates a splicing operation.

[0061] Extract dynamic behavior statistics from the call logs: in, This indicates a function for extracting behavioral features based on call logs, used to extract runtime behavior statistics of components from dynamic call logs. () indicates a component Call frequency statistics within a given time window are used to reflect the intensity of component usage; () indicates a component The execution time-related statistics are used to characterize the runtime performance features of the component; () indicates a component The contextual information features associated during the invocation process are used to reflect the invocation environment and interaction relationships of the component.

[0062] The resulting multi-layer feature representation matrix is ​​obtained by combining the features: in, ( ) represents the feature normalization function, used to normalize the static feature vectors of a component. Scale unification is performed to eliminate the influence of different feature dimensions on subsequent modeling; n represents the number of atomic components involved in modeling; d represents the dimension of the fused component feature vector.

[0063] Output the multi-layer component feature matrix F and the component attribute set. (Includes semantic tags and interface descriptions).

[0064] In one alternative implementation, static analysis can involve performing unified lexical and syntactic parsing on source code files and interface configuration files to generate an abstract syntax tree (AST), then extracting structural features, and finally concatenating the two types of features by mapping them to vectors respectively.

[0065] In this embodiment, structural features include, but are not limited to, dependencies of classes, methods, and functions, call hierarchy, and code block complexity; semantic features include, but are not limited to, identifier naming vectors, comment keywords, interface parameter types and constraints; and encoders include, but are not limited to, models based on graph neural networks or Transformers.

[0066] In another alternative implementation, static analysis can also be based on a pre-trained language model to directly embed the source code and configuration text, extract their deep semantic and syntactic information, and then separate the structurally and semantically related feature vectors through specific pooling or attention mechanisms, thereby achieving efficient feature extraction without explicitly constructing a complete AST.

[0067] Furthermore, the construction of a three-layer dependency graph (structure, semantics, and behavior) in S2, and the learning of hierarchical embedding representations of components through a graph embedding network, includes steps B1-B3: B1. Based on the calling relationships, semantic dependencies, and dynamic behavior associations between components, construct dependency graphs in three dimensions: structure, semantics, and behavior.

[0068] In S2, the results of S1 need to be input, including the component feature matrix F and the retrieved relation table. Semantic dependency matrix .

[0069] Construct a three-layer dependency graph: structural layer semantic layer Behavioral layer : in, This represents a collection of nodes composed of atomic components, where each node corresponds to a component entity. This represents a set of structural dependency edges, used to describe the structural dependencies between components based on code structure, call relationships, or configuration relationships. This represents a set of semantic layer dependency edges, used to describe the relationships formed between components at the functional semantics, interface semantics, or business meaning level. This represents a set of behavioral layer dependency edges, used to describe the behavioral dependencies formed by components during runtime based on invocation behavior, timing relationships, or interaction patterns.

[0070] B2. Using graph embedding networks, the vector representation of nodes in each layer is iteratively learned by aggregating the neighbor feature information of nodes in the dependency graph of each layer.

[0071] Specifically, graph embedding learning is performed on multi-layer graphs: in, Let i represent the set of neighbors of node i. The normalization coefficient is... Let l be the weight matrix of the l-th layer. This represents the feature representation vector of node j at layer l. This represents the feature representation vector of node i after the neighbor information is aggregated and transformed at the (l+1)th layer.

[0072] B3. The embedding vectors learned by the component at the structural, semantic, and behavioral layers are concatenated to generate a comprehensive hierarchical embedding representation.

[0073] Aggregation yields a hierarchical embedding representation of the components: in, This represents the dependency graph of component i in the structural layer. The structural layer feature representation learned by the Chinese Economic Graph Embedding Network; This represents the semantic layer dependency graph of component i. The semantic layer feature table obtained by the Chinese Economic Graph Embedding Network; This represents the dependency graph of component i in the behavior layer. The behavioral layer feature representations learned by the Chinese Economic Graph Embedding Network.

[0074] Output multi-layer embedding vector set and unified dependency graph .

[0075] In this embodiment, the similarity matching and mapping in S3 involves calculating the structural-semantic comprehensive similarity to perform automatic matching and mapping between components. Specifically, in the hierarchical embedding representation space, the mapping of atomic components across different systems or versions is achieved through similarity measurement and structural matching algorithms. The result of input S2 includes a multi-layer embedding set. Unified Dependency Graph Target system component set to be mapped Calculate any component pair The structure-semantic comprehensive similarity is used to generate a mapping matrix based on the optimal matching criterion, and local dependency consistency constraints are introduced to output the component mapping matrix and the matching confidence report.

[0076] In one alternative implementation, similarity matching and mapping can be performed by calculating component embedding vector pairs using cosine similarity or Euclidean distance. The semantic similarity between components is calculated and weighted by combining it with local structural features extracted from the unified dependency graph (Gall) to obtain a structure-semantic comprehensive similarity matrix. Subsequently, the Hungarian algorithm or the maximum weight matching algorithm is used to find the source system component set and the target system component set under the one-to-one mapping constraint. The system calculates the global optimal match between the similarity scores and the constraint satisfaction, while also introducing local dependency consistency constraints. The final output is a mapping matrix and a matching confidence report calculated based on similarity scores and constraint satisfaction.

[0077] In another alternative implementation, similarity matching and mapping can also be based on end-to-end joint learning and matching of graph neural networks (GNNs). The component embedding and the dependency graph structure are input into the graph matching network. The probability of correspondence between components is directly calculated through message passing and attention mechanisms. Dependency consistency constraints are naturally incorporated during training or inference, thereby simultaneously optimizing similarity measurement and structural alignment. Finally, a mapping matrix is ​​directly generated with the probability of the network output as the confidence level.

[0078] Furthermore, S3 performs similarity matching and mapping between components based on hierarchical embedding representation, and introduces local dependency consistency constraints to optimize the mapping results, including steps C1-C3: C1. Calculate the overall matching degree between components based on the hierarchical embedding representation and the structural information of the unified dependency graph.

[0079] C2. Based on the overall matching degree, the optimal matching criterion is used to find the best mapping component in the target system for the component in the source system, and form an initial mapping relationship.

[0080] C3. Apply local dependency consistency constraints to the initial mapping relationship, optimize and output the final component mapping matrix.

[0081] In the embodiments of this application, the comprehensive matching degree between the components in C1 is calculated as the structural-semantic comprehensive similarity. Specifically, the similarity of the comprehensive embedding vectors between any two components is calculated, and combined with their path similarity in the unified dependency graph, the structural-semantic comprehensive similarity is obtained.

[0082] in, This represents the weighting coefficient, used to balance the contribution ratio of component embedding similarity and path dependency similarity in the overall similarity calculation; This represents the cosine similarity function, used to measure the hierarchical embedding representations of component i and component j. and The degree of similarity in the feature space; () indicates the global dependency graph The path similarity between node i and node j is used to characterize the degree of association between components in the dependency topology.

[0083] Generate the mapping matrix based on the optimal matching criterion: Introduce local dependency consistency constraints: in, This represents the set of neighboring nodes of component i in the dependency graph or its corresponding neighborhood feature representation, used to characterize the local dependencies of component i; This represents the set of neighbor nodes of component j corresponding to component i in the dependency graph, or its corresponding neighborhood feature representation; This represents the set of component mapping relationships, containing all component pairs (i,j) that are determined to be mutually mapped.

[0084] Output component mapping matrix and matching confidence report .

[0085] Furthermore, in S4, the process dependency topology is built and optimized by minimizing path redundancy and circular dependencies to generate an optimized flowchart, including steps D1-D2: D1. Based on the component execution sequence and mapping relationship, construct a new process dependency matrix to represent the reconstructed process topology.

[0086] In S4, the result mapping matrix of S3 needs to be input. Unified Dependency Graph and component execution sequence .

[0087] Specifically, the build process dependency matrix is ​​represented as follows: in, This indicates a conditional statement used to determine whether a dependency exists between component i and component j.

[0088] D2. Design optimization objectives, solve for the optimization objectives using graph optimization strategies, and generate a flowchart that achieves optimal path efficiency and structural consistency.

[0089] Specifically, the graph optimization strategy minimizes path redundancy and circular dependencies: in, The connection weight from component i to component j in the process dependency graph represents the relationship between them and indicates whether there is a process dependency. This represents the regularization weight coefficient, used to balance the relative impact of process similarity constraints and circular dependency penalties on the overall optimization objective; The process dependency matrix represents the process dependency matrix. The transpose of the matrix; F represents the matrix norm, which measures the degree of difference between the process dependency matrix and its transpose to characterize the strength of circular dependencies in the process.

[0090] Generate the optimal flowchart structure .

[0091] Output the optimized flowchart Dependency Integrity Report .

[0092] In this embodiment, detecting changes in the multi-layer dependency graph in S5 involves modeling the temporal evolution of the dependency graph to detect component changes and structural drift. Specifically, this includes inputting an optimization flowchart. Historical dependency graph sequence and update time window Calculate the rate of change of node centrality using a change threshold. Determine structural drift, update dependency edge weights, and record evolution logs to output the updated multi-layer dependency graph and component evolution report.

[0093] In one alternative implementation, detecting changes to the multi-level dependency graph can be based on the current optimized flowchart. With historical graph sequence China recently The dependency graph within the window is compared, and the rate of change of various centrality indicators of each component in the graph over time is calculated; when the rate of change of the critical centrality of any node exceeds a preset drift threshold... When the local structure of the node is drifted, it is marked as such. At the same time, by analyzing the increase or decrease of dependent edges or the change of weights, the edge weights in the current multi-layer dependency graph are updated, and all changes are recorded in the evolution log. Finally, the updated dependency graph and the component evolution report containing change summary and impact assessment are output.

[0094] In another alternative implementation, detecting changes to the multi-level dependency graph can also involve comparing the historical graph sequence with the current graph. Treating it as a time-series graph data stream, it applies dynamic graph embedding or time-series graph neural network (T-GNN) for modeling, automatically learns the temporal change patterns of nodes and edges, identifies abnormal changes by predicting the deviation between the graph state in the next time step and the actual state, and uses an attention mechanism to locate the key nodes and edges that cause drift, thereby directly updating the dependency graph structure and generating a component evolution report that describes the evolution trend and potential risks in detail.

[0095] Furthermore, S5's method of detecting changes in the multi-layer dependency graph based on historical data and adaptively updating the dependency graph and weights includes steps E1-E3: E1. Detect component nodes that have experienced structural drift by comparing the rate of change of node centrality in the historical and current dependency graphs.

[0096] The rate of change of nodal centrality is expressed as follows: in, This indicates that node i is at time t. The centrality index value at time t is used to characterize the importance of a node in the dependency graph; This represents the centrality index value of node i at time t.

[0097] E2. If the changes in a node exceed a predetermined threshold, an adaptive adjustment and update of the weights of the node and its associated dependent edges will be triggered.

[0098] By changing the threshold Determine structural drift: Update the dependency edge weights and record the evolution log: in, .

[0099] E3. Record the evolution process and generate an evolution report, outputting the updated multi-layer dependency graph.

[0100] Updated multi-level dependency graph and component evolution report .

[0101] Furthermore, in S6, the mapping and process reconstruction results are corrected and jointly optimized, and the corrected knowledge is fed back to the knowledge base, including steps F1-F3: F1. Formalize the expert's prior knowledge about component dependencies into constraints, which are used to evaluate and correct automatically generated mappings and process results.

[0102] Specifically, formalizing prior knowledge into constraints requires inputting the results of S5, including the updated dependency graph. Expert knowledge base and component evolution report .

[0103] Constructing expert constraint functions: in, The value represents the expert constraint weight coefficient, which is used to characterize the confidence or importance of experts in the consistency constraints between component pairs (i,j).

[0104] F2. Jointly optimize the process refactoring goals, mapping consistency constraints, and expert constraints to obtain the final corrected dependencies.

[0105] Jointly optimize the overall objective function: in, This represents the process dependency optimization loss function, used to minimize path redundancy and circular dependencies in the process dependency topology; This represents a local dependency consistency constraint, used to constrain the consistency of mapped components in the local dependency structure, where... For the corresponding weighting coefficients, This represents the expert knowledge constraint function, used to introduce prior constraints from domain experts on component mapping relationships; The weight coefficient represents the local consistency constraint term, which is used to adjust the degree of influence of the constraint on the overall optimization objective; This represents the weighting coefficient of the expert knowledge constraint term, used to adjust the degree of influence of expert constraints in the overall optimization process.

[0106] F3. Update the new effective dependencies generated during the optimization process to the expert knowledge base.

[0107] The corrected dependencies have been updated in the knowledge base: in, ( ) represents the new dependency weights obtained after process refactoring and joint optimization.

[0108] Output the updated knowledge base and intelligent closed-loop report .

[0109] Example 3, referring to Figure 2 This is one embodiment of the present invention, which provides an intelligent mapping and process reconstruction system for atomic components based on multi-layer graph embedding, including: a component feature extraction and representation module, a multi-layer graph construction and embedding learning module, a component matching and mapping module, a process reconstruction and optimization module, an evolution perception and update module, and an expert feedback and closed-loop optimization module.

[0110] The component feature extraction and representation module parses and extracts features from source code, configuration files, and dynamic call logs to construct a multi-layer component feature matrix containing structural, semantic, and behavioral features.

[0111] The multi-layer graph construction and embedding learning module constructs a three-layer dependency graph of structure, semantics, and behavior based on a multi-layer component feature matrix, and learns to generate hierarchical embedding representations of components through a graph embedding network.

[0112] The component matching and mapping module performs similarity matching and mapping between components based on hierarchical embedding representation, and introduces local dependency consistency constraints to optimize the mapping results.

[0113] The process refactoring and optimization module constructs and optimizes the process dependency topology based on the mapping results and the original dependency relationships, and generates an optimized process flow by minimizing path redundancy and circular dependencies.

[0114] The evolution perception and update module detects changes in the multi-layer dependency graph based on historical data and adaptively updates the dependency graph and weights.

[0115] The expert feedback and closed-loop optimization module introduces expert knowledge constraints to correct and jointly optimize the mapping and process reconstruction results, and feeds the corrected knowledge back to the knowledge base.

[0116] This embodiment also provides an electronic device applicable to a method for intelligent mapping and process refactoring of atomic components based on multi-layer graph embedding, 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 intelligent mapping and process refactoring of atomic components based on multi-layer graph embedding proposed in the above embodiment.

[0117] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements an atomic component intelligent mapping and process refactoring method based on multi-layer graph embedding as proposed in the above embodiment.

[0118] The storage medium proposed in this embodiment and the method for implementing an atomic component intelligent mapping and process refactoring based on multi-layer graph embedding 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.

[0119] 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.

[0120] 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 intelligent mapping and process refactoring of atomic components based on multi-layer graph embedding, characterized in that: include, By parsing and extracting features from source code, configuration files, and dynamic call logs, a multi-layer component feature matrix containing structural, semantic, and behavioral features is constructed. Based on the multi-layer component feature matrix, a three-layer dependency graph of structure, semantics and behavior is constructed, and a hierarchical embedding representation of the components is learned and generated through a graph embedding network. Similarity matching and mapping between components are performed based on hierarchical embedding representation, and local dependency consistency constraints are introduced to optimize the mapping results; Based on the mapping results and the original dependencies, the process dependency topology is constructed and optimized, and the optimized flowchart is generated by minimizing path redundancy and circular dependencies. Changes in the multi-layer dependency graph are detected based on historical data, and the dependency graph and weights are updated adaptively. By introducing expert knowledge constraints, the mapping and process refactoring results are corrected and jointly optimized, and the corrected knowledge is fed back to the knowledge base.

2. The method for intelligent mapping and process refactoring of atomic components based on multi-layer graph embedding as described in claim 1, characterized in that: The construction of a multi-layer component feature matrix, which includes structural, semantic, and behavioral features, includes: Static features of components are extracted by performing static analysis on the source code and interface configuration; Extract behavioral characteristic statistics of components based on dynamic call logs; The static features and behavioral features are concatenated and normalized to obtain a unified multi-layer component feature matrix.

3. The method for intelligent mapping and process refactoring of atomic components based on multi-layer graph embedding as described in claim 2, characterized in that: The construction of a three-layer dependency graph (structure, semantics, and behavior) and the generation of hierarchical embedding representations of components through a graph embedding network include, Based on the calling relationships, semantic dependencies, and dynamic behavior associations between components, dependency graphs are constructed in three dimensions: structure, semantics, and behavior. By using graph embedding networks, the vector representation of nodes in each layer is iteratively learned by aggregating the neighbor feature information of nodes in the dependency graph of each layer. The embedding vectors learned by the components at the structural, semantic, and behavioral layers are concatenated to generate a comprehensive hierarchical embedding representation.

4. The method for intelligent mapping and process refactoring of atomic components based on multi-layer graph embedding as described in claim 3, characterized in that: The method of performing similarity matching and mapping between components based on hierarchical embedding representation, and optimizing the mapping results by introducing local dependency consistency constraints, includes: Based on the structural information of the hierarchical embedding representation and the unified dependency graph, the overall matching degree between components is calculated; Based on the comprehensive matching degree, the optimal matching criterion is used to find the best mapping component in the target system for the component in the source system, and to form an initial mapping relationship. Apply local dependency consistency constraints to the initial mapping relationship, optimize and output the final component mapping matrix.

5. The method for intelligent mapping and process refactoring of atomic components based on multi-layer graph embedding as described in claim 4, characterized in that: The process of constructing and optimizing the process dependency topology, and generating an optimized flowchart by minimizing path redundancy and circular dependencies, includes... Based on the component execution sequence and mapping relationship, a new process dependency matrix is ​​constructed to represent the reconstructed process topology; The design optimization objective is used to solve the optimization objective through a graph optimization strategy, generating a flowchart that achieves optimal path efficiency and structural consistency.

6. The method for intelligent mapping and process refactoring of atomic components based on multi-layer graph embedding as described in claim 4, characterized in that: The method of detecting changes in the multi-layer dependency graph based on historical data and adaptively updating the dependency graph and weights includes: Component nodes that have experienced structural drift are detected by comparing the rate of change of node centrality in the historical and current dependency graphs. If the changes in a node exceed a predetermined threshold, an adaptive adjustment and update of the weights of the node and its associated dependent edges will be triggered. Record the evolution process and generate an evolution report, outputting an updated multi-layer dependency graph.

7. The method for intelligent mapping and process refactoring of atomic components based on multi-layer graph embedding as described in claim 4, characterized in that: The process of correcting and jointly optimizing the mapping and process reconstruction results, and feeding the corrected knowledge back to the knowledge base, includes: The prior knowledge of experts about component dependencies is formalized into constraints to evaluate and correct automatically generated mappings and process results; The process refactoring goals, mapping consistency constraints, and expert constraints are jointly optimized to obtain the final corrected dependencies; New and valid dependencies generated during the optimization process will be updated in the expert knowledge base.

8. A system for intelligent mapping and process refactoring of atomic components based on multi-layer graph embedding, employing the method for intelligent mapping and process refactoring of atomic components based on multi-layer graph embedding as described in any one of claims 1 to 7, characterized in that, include: The module includes component feature extraction and representation, multi-layer graph construction and embedding learning, component matching and mapping, process reconstruction and optimization, evolution perception and update, and expert feedback and closed-loop optimization. The component feature extraction and representation module constructs a multi-layer component feature matrix containing structural, semantic, and behavioral features by parsing and extracting features from source code, configuration files, and dynamic call logs. The multi-layer graph construction and embedding learning module constructs a three-layer dependency graph of structure, semantics and behavior based on the multi-layer component feature matrix, and learns to generate hierarchical embedding representations of components through a graph embedding network. The component matching and mapping module performs similarity matching and mapping between components based on hierarchical embedding representation, and introduces local dependency consistency constraints to optimize the mapping results; The process reconstruction and optimization module constructs and optimizes the process dependency topology based on the mapping results and the original dependency relationship, and generates an optimized process flow by minimizing path redundancy and circular dependencies. The evolution perception and update module detects changes in the multi-layer dependency graph based on historical data and adaptively updates the dependency graph and weights. The expert feedback and closed-loop optimization module introduces expert knowledge constraints to correct and jointly optimize the mapping and process reconstruction results, and feeds the corrected knowledge back to the knowledge base.

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 atomic component intelligent mapping and process refactoring method based on multi-layer graph embedding 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 atomic component intelligent mapping and process refactoring method based on multi-layer graph embedding as described in any one of claims 1 to 7.

Images