Two-way linkage database table and regulatory reporting form field synchronization construction method

Abstract

The present application relates to the technical field of database management, and more particularly to a two-way linkage database table and supervision reporting form field synchronization construction method, which is applied to a supervision data reporting scene; the present application proposes the following scheme: obtaining physical structure metadata of a database table and definition metadata of a supervision reporting form, combining metadata semantic analysis and semantic coding to form a semantic vector set, and constructing a two-way mapping relationship graph through relationship weaving; constructing a linkage propagation path according to a field change event, realizing adaptive two-way structure adjustment, generating a field synchronization construction scheme, and driving two-way mapping relationship graph evolution optimization through feedback learning; the present application innovatively combines a two-way mapping relationship graph and adaptive two-way structure adjustment, and realizes efficient and adaptive two-way field synchronization construction.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of database management, and particularly relates to a method for constructing fields of a database table and a supervision reporting form in a bidirectional linkage. BACKGROUND

[0002] Under the background of digital transformation of supervision reporting, the field synchronization construction between the database table and the supervision reporting form as a core data governance link needs to consider data consistency guarantee and synchronization efficiency improvement. The construction precision of the field mapping relationship and the synchronization response efficiency affect the quality of supervision data reporting. The traditional field maintenance method has problems such as insufficient semantic analysis and correlation analysis capability, lack of adaptive adjustment mechanism, and the like, and an efficient and adaptive bidirectional linkage field synchronization construction scheme is urgently needed. First, the fields of the database table and the supervision reporting form need to be maintained respectively. Due to the lack of effective semantic understanding and mapping mechanism, the maintenance work is repeated and inefficient. Second, in a complex business scenario, the fields change frequently, and the existing technology lacks adaptive bidirectional structure adjustment capability, which easily causes inconsistency between the database table and the supervision reporting form. In addition, the existing synchronization scheme lacks feedback learning and evolution optimization mechanism, and cannot continuously improve the accuracy and efficiency of field synchronization.

[0003] For example, the Chinese patent with the authorization announcement number CN113672683B discloses a distributed database metadata synchronization method based on SparkSQL. The method constructs a cluster node based on a Gossip network, takes a Cockroach node of a distributed database cluster as an internal node, takes a Driver node of SparkSQL as an external node, classifies messages according to keys of Gossip messages through a request observer and a response observer of the external node, filters out metadata events related to a user database and a data table, and then synchronizes the metadata information of the user database and the data table of the internal node to a SparkSQL data warehouse through a metadata event processor interface implemented by a Gossip component. In this way, the high-availability transaction processing capability of CockroachDB and the online analysis advantage of SparkSQL are combined to solve the metadata synchronization problem between the distributed database and the SparkSQL data warehouse.

[0004] For example, Chinese patent CN114625806B discloses a method for constructing temporal RDF and RDFSchema based on a temporal relational database. This method uses the MariaDB temporal relational database as a foundation, first constructing an element-level temporal RDF and RDFSchema graph model, then performing semantic recognition and classification on the temporal tables in the temporal database, subsequently mapping and constructing a temporal RDFSchema based on the semantic information of the table structure according to preset rules, and finally mapping timestamped tuple information to temporal RDF instances. This preserves the temporal semantics of the temporal relational database, supporting data reuse and sharing in the Semantic Web. The disclosed embodiment relies on preset temporal model definitions and table mapping rules to achieve semantic transformation of temporal data, but the correlation between the synchronization strategy and dynamic field changes is weak, lacking adaptive adjustment capabilities.

[0005] The existing technologies described above all suffer from the problems raised in this background: insufficient semantic understanding depth, static mapping relationships, and weak synchronization strategy associations. Therefore, they struggle to meet the core requirements of field synchronization construction accuracy, response speed, and data consistency in regulatory data reporting scenarios. To address these issues, this application provides a bidirectional, interactive method for synchronizing database tables and regulatory reporting form fields. Summary of the Invention

[0006] The technical problem this invention aims to solve is to address the shortcomings of existing technologies by providing a bidirectional, interconnected method for synchronizing database tables and regulatory reporting form fields. This method identifies business and regulatory semantic features through metadata semantic parsing and transforms these features into multi-dimensional semantic vectors through semantic encoding. Relationship weaving technology, combined with field change history records, is used to construct a bidirectional mapping relationship graph, optimizing the field synchronization path. For field change events, a parallel analysis algorithm determines the impact domain and generates a linked propagation path. If synchronization conflicts exist, a consistency assessment mechanism is triggered to prevent data inconsistency. This method can dynamically update the mapping relationship, improving the adaptability and accuracy of database table and regulatory reporting form field synchronization, providing data governance personnel with a real-time synchronization solution, optimizing maintenance efficiency, and improving data consistency.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] A method for synchronizing database tables and regulatory reporting form fields in a two-way linkage manner, applied in regulatory data reporting scenarios, includes:

[0009] Obtain the physical structure metadata of the database tables and the definition metadata of the regulatory reporting forms. Through metadata semantic parsing, identify business semantic features and regulatory semantic features, and generate a two-way semantic feature set.

[0010] The business semantic features and the regulatory semantic features in the bidirectional semantic feature set are converted into multi-dimensional semantic vectors through semantic coding, forming a semantic vector set;

[0011] According to the semantic vector set, combined with the field change history record, a multi-dimensional association relationship between the database table fields and the regulatory submission form fields is constructed through a relationship weaving technique, and a bidirectional mapping relationship graph is constructed;

[0012] A field change event is obtained, and according to the bidirectional mapping relationship graph, a corresponding influence domain of the field change event is determined through a parallel analysis algorithm, and a linkage propagation path is generated;

[0013] According to the linkage propagation path, a synchronization node is determined through consistency evaluation, and an updated linkage propagation path is obtained;

[0014] According to the updated linkage propagation path, a field synchronization construction scheme is generated through adaptive bidirectional structure adjustment.

[0015] The business semantic features and the regulatory semantic features are identified through metadata semantic analysis, and a bidirectional semantic feature set is generated, including:

[0016] The physical structure metadata of the database table is obtained, and through lexical analysis and syntax analysis, the field name, data type, constraint condition and relationship dependency are extracted, and the business semantic features are identified;

[0017] The definition metadata of the regulatory submission form is obtained, and through structured analysis, the form item identifier, data format, verification rule and business logic are extracted, and the regulatory semantic features are identified;

[0018] According to the business semantic features and the regulatory semantic features, a bidirectional semantic feature set is generated through a semantic matching algorithm.

[0019] The multi-dimensional association relationship between the database table fields and the regulatory submission form fields is constructed through a relationship weaving technique, and a bidirectional mapping relationship graph is constructed, including:

[0020] According to the semantic vectors of each field in the semantic vector set, the semantic similarity between the fields is calculated through a cosine similarity algorithm, forming a semantic similarity matrix;

[0021] According to the field change history record, the change frequency and influence range of the field are extracted, and the association strength between the fields is calculated through a statistical analysis algorithm, and a dynamic weight relationship is constructed;

[0022] According to the semantic similarity matrix and the dynamic weight relationship, a bidirectional mapping relationship graph is constructed through a graph neural network fusion modeling algorithm.

[0023] The field change event is classified according to the construction mode type, including database table mapping change, manual table structure change and SQL query logic change, wherein:

[0024] The database table mapping change is a field mapping relationship change caused by database table structure adjustment when a database table field mapping construction mode is adopted;

[0025] The manual table structure change is a database table structure change caused by supervision reporting form field maintenance when a manual table construction mode is adopted;

[0026] The SQL query logic change is a field generation logic change caused by multi-table association query statement adjustment when a SQL query construction mode is adopted.

[0027] According to the bidirectional mapping relationship graph, the influence domain corresponding to the field change event is determined by a parallel analysis algorithm to generate a linkage propagation path, comprising:

[0028] According to the field change event, a change source node is located in the bidirectional mapping relationship graph, and an influence node and a corresponding influence level are identified by a graph traversal algorithm in combination with the association strength in the dynamic weight relationship;

[0029] According to the bidirectional mapping relationship graph, the propagation path and the path length from the change source node to each influence node are calculated by a parallel analysis algorithm, and the propagation path influence weight is calculated in combination with the association strength;

[0030] According to the influence node, the corresponding influence level and the propagation path influence weight, an influence domain containing different levels is determined by classification aggregation, and the different levels include a core influence level and a secondary influence level;

[0031] According to the field change event and the corresponding influence domain, a linkage propagation path is generated by a differentiated propagation strategy.

[0032] According to the field change event and the corresponding influence domain, a linkage propagation path is generated by a differentiated propagation strategy, comprising:

[0033] When the field change event is a database table mapping change, a direct mapping propagation strategy is adopted, and a direct propagation path from the change source node to the target mapping field is generated according to the area of the core influence level in the influence domain;

[0034] When the field change event is a manual table structure change, a structure synchronization propagation strategy is adopted, and a multi-stage propagation path containing table structure creation, field attribute adjustment and constraint relationship update is generated according to different levels in the influence domain;

[0035] When the field change event is an SQL query logic change, a logic restructuring propagation strategy is adopted, and according to the area in the influence domain that is a secondary influence level, a recursive propagation path containing query statement parsing, multi-table association reconstruction and field generation logic update is generated;

[0036] According to the propagation path influence weight and influence level, the direct propagation path, the multi-stage propagation path and the recursive propagation path are prioritized and path optimized to generate a linkage propagation path.

[0037] According to the linkage propagation path, a synchronization node is determined through consistency evaluation to obtain an updated linkage propagation path, comprising:

[0038] According to the propagation node in the linkage propagation path, consistency detection is performed on the data format, data type and constraint condition of each propagation node through consistency evaluation to obtain a consistency score;

[0039] According to the consistency score, a conflict recognition algorithm is used to mark the propagation node with a consistency score less than a preset threshold as a conflict node, and the propagation node with a consistency score greater than or equal to the preset threshold is determined as a synchronization node;

[0040] According to the conflict node, the propagation branch corresponding to the conflict node is removed from the linkage propagation path, and according to the synchronization node and the remaining propagation branch in the linkage propagation path, the linkage propagation path is reconstructed to obtain an updated linkage propagation path.

[0041] The consistency detection on the data format, data type and constraint condition of each propagation node through consistency evaluation to obtain a consistency score, comprising:

[0042] According to the propagation node, the physical structure metadata of the corresponding database table and the definition metadata of the regulatory reporting form are determined, and the data format, data type and constraint condition of each field are extracted;

[0043] According to the data type of each field, a type compatibility algorithm is used to compare the compatibility of the data types of the database table fields and the regulatory reporting form fields, and a type compatibility score is calculated;

[0044] According to the data format of each field, a format consistency algorithm is used to verify the matching degree of the data format between fields, and a format consistency score is calculated;

[0045] According to the constraint condition of each field, a constraint compatibility algorithm is used to detect the compatibility of the non-empty constraint, uniqueness constraint and foreign key constraint of the field, and a constraint compatibility score is calculated;

[0046] According to the type compatibility score, the format consistency score and the constraint compatibility score, a consistency score of each propagation node is calculated by a weighted fusion algorithm.

[0047] According to the updated linkage propagation path, a field synchronization construction scheme is generated by adaptive bidirectional structure adjustment, which includes database table structure adjustment and regulatory reporting form structure adjustment.

[0048] According to the synchronization node in the updated linkage propagation path, a corresponding field change event is determined, and a structure difference analysis algorithm is used to identify the synchronization node for database table structure adjustment and regulatory reporting form structure adjustment.

[0049] When the field change event corresponding to the synchronization node is a database table mapping change, a mapping synchronization adjustment strategy is adopted, and the regulatory reporting form structure adjustment is realized by direct mapping update of the database table field to the regulatory reporting form field.

[0050] When the field change event corresponding to the synchronization node is a manual table structure change, a structure creation adjustment strategy is adopted, and the database table structure adjustment is realized by database table creation, field attribute setting and constraint relationship establishment according to the definition metadata of the regulatory reporting form field.

[0051] When the field change event corresponding to the synchronization node is a SQL query logic change, a logic reconstruction adjustment strategy is adopted, and the regulatory reporting form structure adjustment is realized by query result field parsing, dynamic field mapping and regulatory reporting form structure update according to the updated query logic.

[0052] According to the execution result of the adaptive bidirectional structure adjustment, a field synchronization construction scheme is generated by a synchronization state verification algorithm.

[0053] The field synchronization construction scheme further includes updating the field change history record by feedback learning, and driving the evolution and optimization of the field synchronization construction scheme, including:

[0054] According to the execution result of the field synchronization construction scheme, a synchronization success rate and a data consistency accuracy are calculated by statistical analysis and comparison verification, and synchronization construction effect data is generated.

[0055] According to the synchronization construction effect data, an evolution trend index is generated by multi-period comparative analysis, and the correlation strength in the dynamic weight relationship is updated by feedback learning according to the evolution trend index, to obtain an updated correlation strength.

[0056] According to the updated correlation strength, the corresponding semantic similarity matrix is recalculated, the dynamic weight relationship is updated, the network parameters are retrained through a graph neural network fusion modeling algorithm, and a reconstructed bidirectional mapping relationship graph is obtained.

[0057] According to the reconstructed bidirectional mapping relationship graph, a mapping relationship is established for a field pair with a correlation strength greater than a preset correlation strength threshold, and the mapping relationship is removed for a field pair with a correlation strength less than the preset correlation strength threshold, thereby generating an evolved bidirectional mapping relationship graph.

[0058] The evolved bidirectional mapping relationship graph is adaptively optimized through an evolutionary optimization algorithm based on a graph structure, and the optimization is stopped when the evolution trend index meets a preset convergence condition. BRIEF DESCRIPTION OF DRAWINGS

[0059] Figure 1 A step flowchart for the method of synchronously constructing the fields of the bidirectionally linked database table and the regulatory reporting form in the application is shown in the figure.

[0060] Figure 2 A step flowchart for constructing the bidirectional mapping relationship graph in the application is shown in the figure.

[0061] Figure 3 A step flowchart for generating the linkage propagation path in the application is shown in the figure. DETAILED DESCRIPTION

[0062] The technical solutions of the application will be described in detail below with reference to the accompanying drawings and specific embodiments, and it should be understood that the embodiments and specific features in the embodiments are detailed descriptions of the technical solutions of the application, rather than limitations of the technical solutions of the application.

[0063] In the following text, the term "and / or" is merely a description of the association relationship between the associated objects, indicating that there can be three relationships, for example, A and / or B, which can represent the three cases of A alone, A and B together, and B alone. In addition, the character " / " generally represents an "or" relationship between the associated objects before and after it.

[0064] The application embodiment provides a data structure synchronization construction method, which comprises:

[0065] Metadata information of a source data structure and a target data structure is obtained, and through semantic analysis processing, the characteristic attributes of both ends are identified to generate a bidirectional feature set;

[0066] The feature information in the bidirectional feature set is converted into a calculable multi-dimensional vector representation to form a feature vector set.

[0067] Based on the feature vector set, a multi-dimensional mapping relationship between the source end and the target end is constructed through correlation analysis combined with historical change records to generate a bidirectional relationship network;

[0068] A structure change trigger event is detected, and an influence range of the structure change trigger event is determined through parallel computing based on the bidirectional relationship network to generate a change propagation path;

[0069] Based on the change propagation path, an effective node is determined through consistency checking to obtain an optimized propagation path;

[0070] According to the optimized propagation path, a synchronization construction scheme is generated through adaptive structure updating.

[0071] For example, the embodiment takes the regulatory data reporting scenario as an example to specifically describe the above-mentioned bidirectional linkage database table and regulatory reporting form field synchronization construction method.

[0072] In the prior art, field synchronization between a database table and a regulatory reporting form usually adopts two methods. Among them, method one is to adopt a static mapping method, that is, a one-to-one correspondence between the fields of the database table and the fields of the regulatory reporting form is established through pre-defined field mapping rules. This method ensures that in the case of relatively stable business, basic field mapping function can be realized with low maintenance cost. However, this method has the problem of lag response to field changes, and the synchronization efficiency of the system is mainly affected by the complexity of the mapping rules and the frequency of business changes. The lag response affects the synchronization performance of the system in a dynamic business environment.

[0073] Method two is to manually maintain the field synchronization, which maintains the field correspondence between the database table and the regulatory reporting form through manual identification. Although this method has high flexibility in theory, in actual application, due to the need for a large amount of manual intervention, the maintenance cost is high, and when the business rules change frequently, there are still problems of low maintenance efficiency and easy data inconsistency. Therefore, the existing technology often cannot achieve perfect balance among improving synchronization accuracy, reducing maintenance cost and ensuring data consistency, especially in financial regulation and other application scenarios with high requirements for data quality. The existing field synchronization method is difficult to meet the dual needs of rapid response and efficient maintenance.

[0074] In order to realize efficient and stable synchronization of the database table and the regulatory reporting form in the regulatory data reporting scenario, and improve the adaptability of the synchronization system to complex business changes, and to quickly respond to field changes to maximize maintenance efficiency, the embodiment of the present application provides a bidirectional linkage database table and regulatory reporting form field synchronization construction method based on the above-mentioned data structure synchronization construction method, as shown in Figure 1 The execution process of the method includes:

[0075] S1: Obtain the physical structure metadata of the database table and the definition metadata of the regulatory reporting form, identify the business semantic features and the regulatory semantic features through semantic analysis of the metadata, and generate a two-way semantic feature set;

[0076] In this step, first, the physical structure metadata of the database table and the definition metadata of the regulatory reporting form are obtained, and the business semantic features and the regulatory semantic features are identified through metadata semantic analysis technology, and a two-way semantic feature set is generated. Compared with the traditional single field mapping method, this method realizes the two-way feature extraction of the business level and the regulatory level through metadata semantic analysis. This method significantly improves the depth and accuracy of field semantic understanding, and enhances the ability of the system to identify potential relationships between fields. By constructing a complete two-way semantic feature set, a reliable feature basis is provided for subsequent semantic coding, relationship weaving, and mapping relationship construction, thereby ensuring the accuracy and completeness of field mapping.

[0077] In one aspect, when obtaining the physical structure metadata of the database table, first, connect the database management system, access the data dictionary table, and obtain the physical structure definition of the database table. Then, use a lexical analyzer to perform word segmentation processing on the physical structure definition, and identify field definitions, type declarations, constraint statements, and other morphemes. Next, construct a syntax tree through a syntax analyzer, and extract structured information such as field names, data types, constraint conditions, and relationship dependencies from the syntax tree. Finally, organize the extracted structured information into standardized physical structure metadata, and identify business semantic features.

[0078] On the other hand, when obtaining the definition metadata of the regulatory reporting form, first, read the electronic definition file of the regulatory reporting form. Then, use a structured parser to perform hierarchical analysis on the definition file, and identify form structure and field attributes. Next, extract key information such as form item identifiers, data formats, validation rules, and business logic. Finally, convert the extracted information into standardized definition metadata, and identify regulatory semantic features.

[0079] Further, when generating the two-way semantic feature set, first, establish a feature mapping space for storing the business semantic features and the regulatory semantic features. Then, perform associated mapping of the business semantic features and the regulatory semantic features through a semantic matching algorithm. Next, establish an association index between the features to form a two-way queryable feature structure. Finally, generate a two-way semantic feature set containing complete business semantic features and regulatory semantic features.

[0080] S2: Convert the semantic features in the two-way semantic feature set into multi-dimensional semantic vectors through semantic coding, and form a semantic vector set;

[0081] It can be understood that the semantic features include business semantic features and regulatory semantic features.

[0082] In this step, the semantic features in the bidirectional semantic feature set are processed for semantic coding, which are converted into multi-dimensional semantic vectors to form a standardized semantic vector set. Compared with the traditional string matching method, this method realizes high-dimensional expression of features through semantic vectorization. This method significantly improves the richness and computability of feature expression, and enhances the system's ability to measure the semantic similarity of fields. By constructing a standardized semantic vector set, a mathematical foundation is provided for subsequent multi-dimensional association analysis and mapping relationship construction, thereby ensuring the scientificity and reliability of field mapping.

[0083] In one aspect, the business semantic features in the bidirectional semantic feature set are semantically coded. Specifically, first, the field name in the business semantic feature is taken as input, and a pre-trained word vector model is used to map the morphemes after tokenization of the field name to initial word vectors; then, based on the obtained initial word vectors, a context-aware encoder is used to generate context-enhanced vectors in combination with the context information such as the data type and constraint conditions of the field; next, the initial word vectors and the context-enhanced vectors are weighted and fused through an attention mechanism to obtain a business fusion vector; finally, the fusion vector is converted into a standardized business semantic vector through dimension normalization processing.

[0084] In another aspect, the regulatory semantic features in the bidirectional semantic feature set are semantically coded. Specifically, first, the form item identifier in the regulatory semantic feature is taken as input, and a domain-specific semantic encoder is used to convert it into a basic semantic vector; then, a rule vector is constructed based on the data format and verification rules in the regulatory semantic feature, and a rule coding network is used to convert the business logic into a logic vector; next, the basic semantic vector, the rule vector and the logic vector are fused through a multi-layer perceptron to obtain a regulatory fusion vector; finally, the fusion vector is converted into a standardized regulatory semantic vector through vector regularization processing.

[0085] Further, based on the standardized business semantic vectors and the standardized regulatory semantic vectors obtained in the above two aspects, a semantic vector set is formed. Specifically, first, a vector index space is established for uniformly storing the standardized business semantic vectors and the standardized regulatory semantic vectors; then, the two types of semantic vectors in the vector index space are processed through vector alignment technology to ensure their comparability in the same vector index space; next, a vector retrieval structure is constructed based on the aligned vectors to support efficient similarity calculation and nearest neighbor search; finally, all aligned semantic vectors with retrieval structure are organized into a structured vector set to form the semantic vector set.

[0086] S3: Based on the semantic vector set, combined with the field change history record, a multi-dimensional association relationship between the database table fields and the regulatory reporting form fields is constructed through relationship weaving technology, and a bidirectional mapping relationship graph is constructed;

[0087] In this step, based on the semantic vector set, combined with the field change history record, a multi-dimensional association relationship between the fields is constructed through relationship weaving technology, and a bidirectional mapping relationship graph is finally constructed. Compared with the traditional one-to-one static mapping method, this method realizes multi-dimensional and dynamic association relationship construction through relationship weaving technology. This method significantly improves the adaptability and accuracy of the mapping relationship, and enhances the system's ability to analyze the impact of field changes. By constructing a complete bidirectional mapping relationship graph, it provides a network foundation for subsequent impact domain analysis and linkage propagation, thereby ensuring the accuracy and traceability of field synchronization.

[0088] S4: Obtain the field change event, determine the impact domain corresponding to the field change event according to the bidirectional mapping relationship graph, and generate a linkage propagation path;

[0089] In this step, based on the bidirectional mapping relationship graph, the change impact domain is determined through parallel analysis algorithm, and the corresponding linkage propagation path is generated. Compared with the traditional serial processing method, this method realizes efficient impact domain identification through parallel analysis algorithm. This method significantly improves the processing efficiency and accuracy of change impact analysis, and enhances the adaptability of the system to complex change scenarios. By constructing a clear linkage propagation path, it provides a decision basis for subsequent consistency evaluation and structure adjustment, thereby ensuring the efficiency and accuracy of field synchronization.

[0090] S5: According to the linkage propagation path, determine the synchronization node through consistency evaluation, obtain the updated linkage propagation path, and generate a field synchronization construction scheme through adaptive bidirectional structure adjustment according to the updated linkage propagation path;

[0091] In this step, based on the linkage propagation path, the synchronization node is determined through consistency evaluation and the propagation path is updated, and finally the field synchronization construction scheme is generated through adaptive bidirectional structure adjustment. Compared with the traditional fixed synchronization strategy, this method realizes intelligent structure synchronization through consistency evaluation and adaptive adjustment. This method significantly improves the reliability and adaptability of the synchronization scheme, and enhances the system's ability to handle different types of changes. By generating a complete field synchronization construction scheme, it provides clear operation guidance for actual execution, thereby ensuring the structural consistency and data accuracy between the database table and the regulatory reporting form.

[0092] Please refer to Figure 2FIG. 3 is a flowchart of a step of constructing a bidirectional mapping relationship graph according to an embodiment of the present application, and the specific steps of S3 are as follows:

[0093] S3.1: According to the semantic vectors of each field in the semantic vector set, the semantic similarity between fields is calculated by a cosine similarity algorithm to form a semantic similarity matrix.

[0094] In this embodiment, the semantic similarity matrix is constructed by calculating the semantic similarity between fields using the semantic vector set as input. Compared with the traditional string matching method, this method realizes the accurate quantification of the semantic correlation degree between fields by calculating the cosine similarity in the vector space. This method significantly improves the accuracy and reliability of the similarity calculation, and provides a mathematical basis for the subsequent construction of the association relationship.

[0095] Specifically, first, the standardized business semantic vectors of the database table fields and the standardized regulatory semantic vectors of the regulatory reporting form fields are extracted from the semantic vector set. Since the database table fields and the regulatory reporting form fields come from different semantic parsing processes, the semantic vectors generated by them may have different dimension specifications, which will cause the inaccuracy of the cosine similarity calculation results.

[0096] The dimension information of the database table field semantic vectors and the regulatory reporting form field semantic vectors is obtained by a vector dimension detection algorithm. When it is detected that the dimensions of the two types of semantic vectors are inconsistent, the extracted semantic vectors are subjected to dimension alignment processing to ensure that the semantic vectors of fields from different sources have the same dimension specifications. Specifically, when the business semantic vector dimension is less than the regulatory semantic vector dimension, a zero padding strategy is used to expand the low-dimensional vector to a high-dimensional vector; when the business semantic vector dimension is greater than the regulatory semantic vector dimension, a truncation or average pooling strategy is used to compress the high-dimensional vector to a low-dimensional vector.

[0097] After completing the dimension alignment, the vector normalization algorithm is used to normalize the lengths of the semantic vectors to unit length to eliminate the influence of the vector amplitude difference on the similarity calculation. When the dimension of the semantic vector exceeds a preset high-dimensional threshold, the principal component analysis algorithm is used to reduce the dimension of the high-dimensional semantic vector to improve the calculation efficiency while maintaining the main semantic information.

[0098] Then, according to the above-processed semantic vectors, the semantic similarity between fields is calculated by a cosine similarity algorithm. Batch matrix operation is used to calculate the semantic similarity of multiple field pairs in parallel. A similarity threshold filtering mechanism is introduced, and the similarity threshold is set according to the business scenario, and only the field pairs with a similarity exceeding the similarity threshold are retained.

[0099] Finally, the calculated semantic similarity of all field pairs is organized into a two-dimensional matrix form to obtain a semantic similarity matrix. The row dimension of the matrix corresponds to the number of database table fields, the column dimension corresponds to the number of regulatory reporting form fields, and the matrix element represents the semantic similarity value of the corresponding field pair. Sparse matrix storage format is used to optimize memory usage, and the matrix is normalized to map the similarity value to a standard numerical interval.

[0100] S3.2: According to the field change history record, the change frequency and influence range of the field are extracted, the correlation strength between the fields is calculated by statistical analysis algorithm, and the dynamic weight relationship is constructed;

[0101] In this embodiment, the field change history record is taken as input, and the dynamic weight relationship between the fields is constructed by analyzing the change characteristics of the fields. Compared with the static structure analysis method, this method realizes the dynamic quantification of the field correlation relationship by introducing the historical change information. This method significantly improves the timeliness and adaptability of the correlation relationship, and provides a dynamic weight basis for subsequent graph construction.

[0102] First, according to the field change history record, the field change event sequence is extracted by the change event classification algorithm. The execution of the change event classification algorithm includes: extracting change record information from the field change history record, including change time, change type, change content, involved fields and other change basic data; analyzing the change basic data by constructing a mode recognition algorithm, and identifying the construction mode type according to the characteristic pattern of the change operation; classifying the field change event into three types of database table mapping change, manual table structure change and SQL query logic change, and generating a field change event sequence containing change time, change type and construction mode type identifier. Different types of field change events are given corresponding importance weight coefficients, among which SQL query logic change has the highest weight, manual table structure change is second, and database table mapping change has relatively low weight.

[0103] Then, according to the field change event sequence, the change frequency data is generated by the change mode recognition algorithm. The execution of the change mode recognition algorithm includes: extracting time sequence information from the field change event sequence, including the change time point, change interval, change type distribution and other time sequence characteristic data of each field; grouping and counting the time sequence characteristic data by time window algorithm, respectively calculating the database table mapping change frequency, manual table structure change frequency and SQL query logic change frequency; and weighting the change frequency of each type by weighting frequency calculation algorithm combined with importance weight coefficient and time decay factor to generate change frequency data containing weighted change frequency and change interval statistical characteristics.

[0104] Then, according to the field change event sequence and system structure information, the influence range data is generated by an influence range analysis algorithm. The execution of the influence range analysis algorithm includes: extracting change involved field information from the field change event sequence, including the source field, target field, associated field, and other change influence objects; extracting dependency relationship information from the system structure information, including the foreign key relationship of the database table, index dependency relationship, field reference relationship, query association relationship, and other structure dependency data; performing correlation analysis on the change influence objects and structure dependency data by a hierarchical influence propagation algorithm, constructing a table-level dependency network, a field-level dependency network, and a query-level dependency network, dividing the influence range into a direct influence layer, an indirect influence layer, and a potential influence layer, and generating influence range data including influence layer level, influence field list, and propagation path.

[0105] Finally, according to the change frequency data and the influence range data, the dynamic weight relationship is generated by a dynamic weight fusion algorithm. The execution of the dynamic weight fusion algorithm includes: extracting frequency weight information from the change frequency data, including the weighted change frequency of each field pair, change type weight distribution, and other frequency feature data; extracting range weight information from the influence range data, including the influence layer level, propagation strength, and dependency depth of each field pair; performing fusion calculation on the frequency feature data and the range feature data by a multi-dimensional weight calculation algorithm based on construction mode perception, setting a differentiated fusion strategy according to different construction mode types, calculating the correlation strength value between fields, and generating a dynamic weight relationship including the correlation strength value, construction mode type identifier, and weight calculation basis.

[0106] S3.3: According to the semantic similarity matrix and dynamic weight relationship, a bidirectional mapping relationship graph is constructed by a graph neural network fusion modeling algorithm.

[0107] In this embodiment, first, according to the semantic similarity matrix, a basic graph structure is generated by a graph structure initialization algorithm. Node basic data is extracted from the semantic similarity matrix, and database table fields and regulatory reporting form fields are taken as nodes of different types to generate a node set; a connection edge between nodes is established based on a similarity threshold to generate a semantic edge set; attribute features are extracted from field basic information to generate an initial feature vector of the node. The basic graph structure including the node set, the semantic edge set, and the initial feature vector of the node is obtained.

[0108] Then, according to the basic graph structure and the dynamic weight relationship, an enhanced graph structure is generated through an edge feature fusion algorithm. The correlation strength in the dynamic weight relationship is taken as the weight attribute of each edge in the semantic edge set; a comprehensive feature vector is calculated by combining the semantic similarity and the correlation strength, containing multi-dimensional features such as semantic similarity, correlation strength, time factor, and type compatibility; feature normalization processing is performed to ensure that the numerical ranges of different dimensional features are consistent. An enhanced graph structure containing a node set, a semantic edge set with multi-dimensional features, and a node initial feature vector is obtained.

[0109] Next, according to the enhanced graph structure, a trained graph convolutional neural network model and a node representation vector are generated through a graph neural network training algorithm, specifically including: constructing a multi-layer graph convolutional neural network model, and adaptively learning the importance weight between nodes through a graph attention mechanism; deep fusion of node features and relationship modeling are realized through multi-round iterative training; the final representation vector of the node is extracted from the trained network, and the trained graph neural network model and the node representation vector learned through deep learning are obtained.

[0110] Finally, according to the trained model and the node representation vector, a bidirectional mapping relationship graph is generated through a graph construction algorithm. Based on the node representation vector, the latent edge connection probability is calculated to identify the implicit field association relationship; the connection confidence score between field pairs is obtained through similarity calculation; a connection probability threshold is set to filter high-confidence edge connections, and a bidirectional mapping relationship graph is constructed. Finally, a graph structure containing the bidirectional mapping relationship between the database table fields and the regulatory reporting form fields is obtained.

[0111] Please refer to Figure 3 The step flow chart provided by the embodiment of the present application for generating a linkage propagation path is as follows:

[0112] S4.1: According to the field change event, locate the change source node in the bidirectional mapping relationship graph, and identify the influence nodes and corresponding influence levels through a graph traversal algorithm in combination with the correlation strength in the dynamic weight relationship;

[0113] In this embodiment, first, a field change event is acquired. The information sources of the field change event include database change records, regulatory reporting system operation records, and SQL query definition change records. According to the construction type, the field change event is divided into three categories: database table mapping change, manual table structure change, and SQL query logic change. Among them, the database table mapping change is the change of field mapping relationship caused by database table structure adjustment when the database table field mapping construction method is adopted, and the corresponding change information is obtained by recording the database structure change; the manual table structure change is the change of database table structure caused by the maintenance of regulatory reporting form field when the manual table construction method is adopted, and the corresponding change information is obtained by recording the reporting form maintenance operation; the SQL query logic change is the change of field generation logic caused by multi-table association query statement adjustment when the SQL query construction method is adopted, and the corresponding change information is obtained by recording the query statement modification.

[0114] Then, according to the type and content of the field change event, the change source node is located in the bidirectional mapping relationship graph. In this embodiment, the change type recognition mechanism is used to convert different types of change events into unified node query conditions, so as to realize accurate positioning in the graph. For the database table mapping change, the database field node is located in the graph through the table field identifier; for the manual table structure change, the regulatory field node is located in the graph through the reporting field identifier; for the SQL query logic change, the derived field node is located in the graph through the query output field identifier.

[0115] Next, based on the graph traversal algorithm, the influence node recognition is performed from the change source node. In this embodiment, an improved graph traversal algorithm is used, and through the hierarchical traversal control strategy and dynamic pruning strategy, the traversal efficiency and recognition accuracy are improved. In the traversal process, in combination with the correlation strength in the dynamic weight relationship, the influence degree of each reachable node is evaluated through the influence degree calculation method. The influence degree calculation method uses a cumulative influence calculation model to consider the correlation strength and propagation distance between nodes, and realizes the quantitative evaluation of the influence degree.

[0116] Finally, according to the influence degree evaluation result, the influence level is determined through the influence level division method. In this embodiment, a threshold adjustment strategy is designed to dynamically determine the division standard according to the characteristics of the change type and the structure characteristics of the graph, and the identified influence nodes are labeled as the corresponding influence levels. The threshold adjustment strategy optimizes the threshold parameter setting by analyzing the historical change data and the influence propagation mode, and improves the accuracy of the level division.

[0117] S4.2: According to the bidirectional mapping relationship graph, the propagation path and its path length from the change source node to each influence node are calculated through the parallel analysis algorithm, and the propagation path influence weight is calculated in combination with the correlation strength;

[0118] In this embodiment, first, a parallel propagation path search algorithm is adopted to calculate the propagation paths from the change source node to each of the affected nodes according to the node connection relationship in the bidirectional mapping relationship graph, and a propagation path set is obtained. Specifically, this embodiment groups the affected nodes according to node types and graph distribution characteristics by designing a multi-thread propagation path search strategy, allocates independent computing threads to each group, and realizes parallel processing of the propagation path search. In the propagation path search process, a hybrid search strategy combining depth-first search and breadth-first search is adopted, and by setting the upper limit of the propagation path length and the cycle detection mechanism, the generation of invalid propagation paths is avoided, and the search efficiency is improved. For node pairs with multiple reachable propagation paths, all different propagation paths are identified and retained by a propagation path deduplication algorithm, ensuring the integrity of the propagation path analysis, and finally obtaining the deduplicated propagation paths.

[0119] Then, the propagation path feature of each identified propagation path is calculated, and the propagation path feature is calculated according to the path structure information in the propagation path set. Specifically, this embodiment calculates the number of edges on the propagation path according to the edge connection relationship on the propagation path by a propagation path length calculation method, obtains the path length as a measurement index of propagation distance, and records the complete node sequence on the propagation path including node type, node identifier and node position information in the propagation path according to the node information in the bidirectional mapping relationship graph, obtains the propagation path structure information, and provides detailed structure information for subsequent propagation path analysis. Through a propagation path complexity evaluation method, the structural complexity index of the propagation path is calculated according to the propagation path structure information, considering the diversity of node types and the complexity of connection modes in the propagation path.

[0120] Then, the propagation path influence weight of each propagation path is calculated in combination with the association strength in the dynamic weight relationship, and the propagation path influence weight is obtained according to the path length, the propagation path structure information and the association strength. Specifically, this embodiment adopts a propagation path influence weight aggregation algorithm to calculate the propagation path influence weight by analyzing the association strength of each edge on the propagation path. The weighted aggregation function considers the cumulative effect and attenuation effect of edge weight, and by designing a weight attenuation factor, the gradual weakening characteristics of influence in the propagation path propagation process are simulated. For propagation paths containing different types of nodes, the node type weight adjustment mechanism is used to correct the propagation path influence weight according to the importance of node types, and the final propagation path influence weight is obtained.

[0121] S4.3: According to the influence node, the corresponding influence level and the propagation path influence weight, the influence domain containing different levels is determined by classification aggregation;

[0122] In this embodiment, according to the influence node, the corresponding influence level and the propagation path influence weight, the influence domain containing different levels is constructed by a classification aggregation algorithm. The different levels include a core influence level, a secondary influence level and a boundary influence level. This method realizes the fine management of the influence range through the multi-level influence domain division, and significantly improves the rationality of the influence domain division.

[0123] Firstly, according to the influence node, an influence node feature table is generated by a feature information collection algorithm. Feature information is extracted from each influence node, including the influence level, the shortest path length from the influence node to the change source node, the maximum propagation path influence weight, the node type identifier and the node dependency relationship; an influence node feature table is established, in which each row corresponds to an influence node and each column corresponds to a feature attribute; data cleaning, missing value processing and abnormal value detection are performed to obtain a complete influence node feature table.

[0124] Then, according to the influence level and the propagation path influence weight in the influence node feature table, a hierarchical classification rule set is generated by a hierarchical classification rule design algorithm. It is set that when the influence level of an influence node exceeds the preset core level threshold and its propagation path influence weight exceeds the preset core weight threshold, the node is classified as a core influence level; it is set that when the influence node does not meet the core influence level determination condition, but its influence level exceeds the preset minimum level threshold, the node is classified as a secondary influence level; when the influence node neither meets the core influence level determination condition nor meets the secondary influence level determination condition, but is still within the propagation path coverage, the node is classified as a boundary influence level. The hierarchical classification rule set containing the core influence level determination rule, the secondary influence level determination rule and the boundary influence level determination rule is obtained.

[0125] Next, according to the hierarchical classification rule set, the nodes in the influence node feature table are classified and aggregated by a classification aggregation algorithm to generate a hierarchical grouping result. According to the determination conditions in the hierarchical classification rule set, each influence node is assigned to a level; according to the propagation path influence weight values of the nodes in each level, the nodes are sorted in descending order; a similarity-based aggregation method is used to aggregate nodes with similar propagation characteristics into the same group. Through similarity measurement calculation and aggregation determination, the hierarchical grouping result containing the node assignment of each level and the sorting within the group is obtained.

[0126] Then, according to the node distribution in the hierarchical grouping result, an optimized hierarchical grouping result is generated by a classification aggregation optimization algorithm. It is checked whether there are nodes with unreasonable classification, including hierarchical boundary nodes and isolated nodes. The hierarchical boundary nodes are re-evaluated to determine their final attribution level. The isolated nodes are re-aggregated to a more suitable level to obtain an optimized and adjusted hierarchical grouping result.

[0127] Finally, according to the level information and node organization relationship in the optimized hierarchical grouping result, an influence domain containing different levels is generated by an influence domain determination algorithm. The classification aggregation result is organized into a structured influence domain description, including level identification, node set, feature statistical information and level relationship description. A level description document is generated for each level. The correlation relationship mapping between levels is established. A processing strategy identifier is assigned to each level to obtain an influence domain containing core influence levels, secondary influence levels and boundary influence levels. In this embodiment, according to the influence nodes, corresponding influence levels and propagation path influence weights, an influence domain containing different levels is constructed by classification aggregation technology, including core influence levels, secondary influence levels and boundary influence levels. Compared with the single influence domain division method, this method realizes fine management of the influence range through multi-level influence domain division. This method significantly improves the rationality of influence domain division and provides a basis for subsequent differentiated processing.

[0128] S4.4: According to the field change event and the corresponding influence domain, a linkage propagation path is generated by a differentiated propagation strategy.

[0129] In this embodiment, based on the divided influence domain, a corresponding propagation strategy is selected according to the type of the field change event to generate the final linkage propagation path. This method adapts to the characteristics of different types of changes through differentiated strategies, significantly improving the pertinence and practicality of the propagation path.

[0130] First, for the input field change event, a type recognition algorithm is used to determine its belonging to a change category. A change event type classification model is established. Based on the source object type, operation type identifier and influence range characteristics of the event, the change event is classified into a predefined change type. The change event type includes: database table mapping change, manual table structure change, SQL query logic change and other main types to obtain the specific type identifier of the change event.

[0131] Then, based on the identified change event type, a corresponding differential propagation strategy is selected from a preconfigured strategy library by a strategy selection algorithm. A strategy mapping mechanism is adopted to establish a one-to-one correspondence between the change event type and the propagation strategy. For database table mapping change, a direct mapping propagation strategy is adopted, the mapping relationship of the change source field in the database table is identified, a point-to-point propagation path from the change source node to each target mapping field is constructed, the propagation weight of each direct propagation path is calculated, and a direct propagation path description containing the path identifier, the starting node, the ending node, and the propagation weight is obtained.

[0132] For manual table structure change, a structure synchronization propagation strategy is adopted, a comprehensive propagation path containing multiple processing stages is generated according to different levels of information in the influence domain, including a table structure creation stage, a creation propagation path of a new table structure is generated based on the nodes of the core influence level; a field attribute adjustment stage, a synchronous adjustment propagation path of the field attribute is generated based on the nodes of the secondary influence level; a constraint relationship update stage, an update propagation path of the constraint relationship is generated based on the nodes of the boundary influence level. Each processing stage is assigned an execution priority and a dependency relationship identifier, and a multi-stage propagation path is obtained.

[0133] For SQL query logic change, a logic reconstruction propagation strategy is adopted, a composite propagation path containing query logic reconstruction is generated based on the node information in the secondary influence level, including query statement parsing processing, syntax parsing and semantic analysis are performed on the affected SQL query statement; multi-table association reconstruction processing, the association logic and connection conditions between tables are reconstructed; field generation logic update processing, the generation logic of the calculation field and the derived field is updated; a recursive analysis algorithm is used to process complex nested query and subquery structures, and a recursive propagation path of logic reconstruction is obtained.

[0134] Finally, based on the generated various propagation paths, a final linkage propagation path is generated by priority sorting and path optimization algorithm. According to the factors such as propagation path influence weight, influence level and path complexity, various propagation paths are comprehensively sorted. A multi-factor weighted scoring algorithm is used to calculate the priority score of each propagation path; path optimization processing is performed, including redundant path elimination and loop detection, a path similarity analysis algorithm is used to identify redundant paths with repeated functions, a loop detection algorithm in graph theory is used to identify and process possible path loops, and the optimized propagation paths are organized into a structured linkage propagation path.

[0135] The specific steps of S5 are as follows:

[0136] S5.1: According to the propagation nodes in the linkage propagation path, consistency detection is performed on the data format, data type and constraint condition of each propagation node through consistency evaluation, and a consistency score is obtained;

[0137] In this embodiment, a comprehensive consistency evaluation is performed on each propagation node in the linkage propagation path, and data compatibility and structural consistency between nodes are ensured through a multi-dimensional detection mechanism.

[0138] First, according to the propagation nodes in the linkage propagation path, a node metadata set is generated through a metadata extraction algorithm. Taking the propagation nodes as input data, a propagation node parsing operation is performed to parse and obtain the basic attribute information of the nodes according to the identification information and node type of each propagation node; a database metadata query operation is performed to query the corresponding database table physical structure metadata according to the database table identification of the propagation node, including table structure definition, field attribute configuration and index information; a regulatory metadata query operation is performed to obtain the regulatory reporting form definition metadata according to the regulatory reporting form identification of the propagation node, including field definition specification, data requirement specification and verification rule configuration; a field attribute extraction operation is performed to extract the complete attribute description of each field, including data format specification, data type definition and constraint condition configuration; a field mapping relationship establishment operation is performed to establish a corresponding relationship between the database table fields and the regulatory reporting form fields, and a node metadata set containing complete field attributes and mapping relationships is obtained.

[0139] Then, according to the data types of the fields, a type compatibility score is calculated through a type compatibility algorithm. A data type extraction operation is performed to extract the data type definitions of the corresponding field pairs from the node metadata set, including basic data types, precision parameters and length limits; a type compatibility rule matching operation is performed to find the compatibility rules between the database table field types and the regulatory reporting form field types according to the predefined data type compatibility rule library; a type conversion feasibility analysis operation is performed to evaluate the feasibility and safety of converting from the database table field type to the regulatory reporting form field type; a complex type structure parsing operation is performed to recursively analyze the internal structure and constituent elements of the type using a type structure parsing algorithm; a type compatibility scoring operation is performed to calculate the type compatibility score for each field pair, and a type compatibility score reflecting the data type compatibility degree of the database table fields and the regulatory reporting form fields is obtained.

[0140] Next, based on the data format of each field, a format consistency score is calculated using a format consistency algorithm. This involves: extracting the format pattern definitions for corresponding field pairs from the node metadata set, including the decimal places and thousands separator configuration for numeric formats, the encoding method and case sensitivity for character formats, and the display mode for date formats; performing a format pattern matching operation, using pattern matching techniques from the format consistency algorithm to compare the format specifications of database table fields with those of regulatory reporting forms; assessing the complexity of format conversion and the risk of data loss; verifying the feasibility of format conversion, validating the degree of data format matching between fields and the feasibility of conversion; and finally, calculating a format consistency score for each field pair to obtain a score reflecting the degree of data format matching between the database table fields and the regulatory reporting form fields.

[0141] Then, based on the constraints of each field, a constraint compatibility score is calculated using a constraint compatibility algorithm. This involves: extracting constraint definitions for corresponding field pairs from the node metadata set, including NOT NULL constraints, uniqueness constraints, foreign key constraints, and check constraints; performing NOT NULL constraint compatibility checks to compare the consistency of null value handling strategies and assess the likelihood of constraint conflicts and data migration risks; performing uniqueness constraint compatibility checks to analyze whether the uniqueness requirements of database table fields and regulatory reporting form fields are consistent; performing foreign key constraint compatibility checks to check the existence of related tables and the compatibility of related fields; performing check constraint compatibility checks to compare the logical rules of constraints and assess the consistency of constraint rules; and finally, performing constraint compatibility scoring to calculate a constraint compatibility score for each field pair, resulting in a score reflecting the degree of compatibility between the constraints of the database table fields and the regulatory reporting form fields.

[0142] Finally, based on the type compatibility score, format consistency score, and constraint compatibility score, a weighted fusion algorithm is used to calculate the consistency score for each propagation node. The process involves determining the weights of the scoring dimensions using the analytic hierarchy process (AHP) to establish the weight coefficients for type compatibility, format consistency, and constraint compatibility; calculating the overall consistency score at the field level by weighting and fusing the three-dimensional scores for each field pair using a weighted fusion algorithm; calculating the overall consistency score for each propagation node based on the overall consistency score and the field importance weights; dynamically optimizing the scoring weights based on historical synchronization success rate data and node performance; and standardizing the consistency scores to ensure comparability and usability, resulting in a consistency score that reflects the overall compatibility of each propagation node.

[0143] S5.2: According to the consistency score, through the conflict identification algorithm, the propagation node with a consistency score less than a preset threshold is marked as a conflict node, and the propagation node with a consistency score greater than or equal to the preset threshold is determined as a synchronization node;

[0144] In this embodiment, the synchronization nodes and the conflict nodes are distinguished through the conflict identification mechanism, and the node classification information is provided for subsequent path optimization. First, based on historical synchronization experience and business fault tolerance requirements, a preset threshold of the consistency score is determined through a threshold optimization algorithm. Specifically, the historical synchronization operation data and the business requirement configuration are taken as input data, a historical data statistical analysis operation is performed, the distribution relationship between the synchronization success rate and the consistency score is statistically analyzed according to the historical synchronization operation record, and a score-success rate mapping model is established. An ROC curve analysis operation is performed, the ROC curve analysis method is adopted, the synchronization success rate is taken as the true positive rate, and the false positive rate is taken as the false positive rate, the ROC curve is drawn, the optimal threshold point is determined, and the relationship between the synchronization success rate and the false positive rate is balanced. A business fault tolerance quantification operation is performed, the tolerance degree of the business to data inconsistency is quantified according to the business importance level and the data quality requirement, and a business fault tolerance evaluation model is established. A dynamic threshold adjustment operation is performed, the preset threshold is dynamically adjusted according to the particularity of the current business scene and the real-time performance requirement, the threshold optimization algorithm is adopted to adapt to the demand changes of different business environments, and the optimal preset threshold adapting to the current business environment and historical experience is obtained.

[0145] Secondly, according to the preset threshold, the propagation nodes are classified and processed through the conflict identification algorithm to distinguish the conflict nodes and the synchronization nodes. The consistency score of each propagation node and the preset threshold determined in the foregoing step are taken as input data, a node traversal operation is performed, all the propagation nodes in the linkage propagation path are traversed, and the consistency score and the node identification information of each node are extracted. The conflict identification algorithm is adopted to compare the size relationship between the consistency score of each node and the preset threshold, and the node classification marking is performed according to the comparison result. A conflict node marking operation is performed, for the propagation node with a consistency score less than the preset threshold, the propagation node is marked as a conflict node, and the conflict identification, the conflict discovery time and the conflict severity of the node and other attribute information are recorded. A synchronization node determination operation is performed, for the propagation node with a consistency score greater than or equal to the preset threshold, the propagation node is determined as a synchronization node, and the synchronization priority, the synchronization reliability level and the expected synchronization performance of the node and other attribute information are marked. A node classification result set containing the classification information of the conflict nodes and the synchronization nodes is obtained.

[0146] Further, according to the consistency score of the conflict node, the specific conflict type and conflict severity of the conflict node are analyzed by a conflict type identification algorithm. The conflict node set and detailed score data obtained in the foregoing steps are taken as input data, and according to the specific values of the type compatibility score, the format consistency score and the constraint compatibility score of the conflict node, the main conflict dimension and the secondary conflict dimension are identified. According to the abnormal pattern of the dimension score, the conflict type is determined by using the conflict type identification algorithm. According to the deviation degree of the consistency score from the preset threshold and the conflict influence range, the conflict severity is calculated by using the conflict severity evaluation model. According to the conflict type and the node metadata information, the root cause analysis algorithm is used to trace the root cause of the conflict, including data source difference, business rule change, system version incompatibility and the like. Finally, according to the conflict type, severity and root cause, the detailed conflict analysis result is obtained.

[0147] Finally, according to the detailed conflict analysis result, a conflict processing strategy matching algorithm is used to generate targeted processing suggestions and solutions for each conflict node. According to the pre-constructed conflict type knowledge base, the pattern matching technology is used to match the identified conflict node with the standard conflict pattern in the knowledge base, and similar historical conflict cases are identified. According to the conflict pattern matching result, the processing strategy retrieval operation is performed to retrieve the corresponding processing method and solution from the processing strategy library, including data conversion strategy, constraint adjustment scheme, format standardization method and the like. For the identified complex mixed conflict, a composite conflict decomposition operation is performed, and a multi-level conflict decomposition algorithm is used to decompose the composite conflict into multiple independent single conflicts, and corresponding processing strategies are formulated respectively, so as to improve the operability and success rate of conflict resolution. According to the conflict severity, processing complexity and business influence range, a processing priority sorting operation is performed to determine the priority order of the processing tasks of each conflict node, and the allocation efficiency of the processing resources is optimized.

[0148] S5.3: According to the conflict node, the propagation branch corresponding to the conflict node is removed from the linkage propagation path, and according to the synchronization node and the remaining propagation branch in the linkage propagation path, an updated linkage propagation path is obtained by reconstructing the linkage propagation path;

[0149] In this embodiment, based on the conflict node identification result, a reliable updated linkage propagation path is generated through a path reconstruction mechanism to ensure the successful execution of subsequent synchronization operations. First, according to the conflict node and the bidirectional mapping relationship graph, a conflict branch identification operation is performed to obtain a set of propagation branches that need to be removed. A graph traversal algorithm is used to analyze the upstream and downstream dependency relationships of the conflict node in the bidirectional mapping relationship graph, and all propagation path branches affected by the conflict node are identified, including direct association branches and indirect dependency branches. Specifically, a depth-first search algorithm is used to start from the conflict node and traverse along the associated edges in the bidirectional mapping relationship graph, record all propagation path segments passing through the conflict node, and form a conflict branch identification set.

[0150] Then, according to the conflict branch identification set and the connectivity constraints of the linkage propagation path, a propagation branch removal operation is performed to obtain the residual path structure after removing the conflict branches. A branch isolation algorithm is used to accurately locate and remove the corresponding propagation path segments based on the branch information in the conflict branch identification set, ensuring that the removal operation does not damage the overall connectivity and logical integrity of the path. For the key path nodes identified in the conflict branch identification set, a path rerouting mechanism is used to find alternative connection methods based on the alternative paths in the bidirectional mapping relationship graph, maintaining the continuity of the propagation path. Further, the removal information of each branch in the conflict branch identification set is recorded, including the removal reason, the impact range and the potential recovery scheme, providing a reference basis for subsequent path optimization.

[0151] Next, according to the synchronization node and the residual path structure, a path reconstruction operation is performed to obtain the candidate updated linkage propagation path. First, based on the available nodes and connection relationships in the residual path structure, a path reconstruction search space is constructed to form a candidate path node set and a candidate connection edge set. Then, a shortest path algorithm is used to recalculate the optimal connection path between the synchronization nodes, taking the synchronization nodes as the starting point and the termination point of path reconstruction, and finding the optimal propagation path in the search space composed of the candidate path node set and the candidate connection edge set using Dijkstra algorithm. Based on the dynamic weight relationship in the bidirectional mapping relationship graph, considering multiple factors such as path length, propagation delay and resource consumption, a preliminary candidate propagation path set is generated. Next, the preliminary candidate propagation path set is subjected to path optimization processing, a path redundancy mechanism is introduced to establish a backup path for the key propagation path, and the multi-dimensional association relationship in the bidirectional mapping relationship graph is used to expand the candidate propagation path set, improving the fault tolerance and stability of the system. Optionally, a path load balancing algorithm is used to distribute propagation tasks among multiple optional paths in the candidate propagation path set, and load optimization is achieved by dynamically adjusting the propagation weight, finally obtaining the candidate updated linkage propagation path.

[0152] Finally, according to the candidate updated linkage propagation path, a path integrity verification operation is performed to obtain the updated linkage propagation path. The connectivity of all synchronization nodes is verified using a breadth-first search algorithm, a topological sorting sequence is constructed based on the association strength in the bidirectional mapping graph to check the dependency relationship, and the path execution efficiency and resource consumption are evaluated by a complexity analysis method. According to the results of the above verification, the path pruning and merging techniques are used to optimize and adjust the problematic path segments, and finally the updated linkage propagation path that has passed the integrity verification is output.

[0153] S5.4: According to the updated linkage propagation path, a field synchronization construction scheme is generated through adaptive bidirectional structure adjustment.

[0154] In this embodiment, based on the updated linkage propagation path, the adaptive bidirectional structure adjustment mechanism is used to realize the coordinated execution of database table structure adjustment and supervision reporting form structure adjustment, and finally a complete field synchronization construction scheme is generated. The adaptive bidirectional structure adjustment mechanism can adopt corresponding adjustment strategies according to different field change event types to ensure the accuracy and consistency of bidirectional structure adjustment.

[0155] First, based on the updated linkage propagation path, the structure difference analysis algorithm is used to select the synchronization nodes that need to be adjusted. Specifically, the structural feature vector of the database table field is extracted from the metadata information of the synchronization node, including field name, data type, length limit, non-empty constraint and other attribute information; the structural feature vector of the supervision reporting form field is extracted from the metadata information of the synchronization node, including field identification, data format, verification rule, mandatory attribute and other configuration information; the database table field structural feature vector and the supervision reporting form field structural feature vector are compared one by one through the feature vector comparison algorithm, the structural similarity score between each field pair is calculated, and a structure difference evaluation matrix is generated with the database table field as the row and the supervision reporting form field as the column. Traverse each synchronization node in the linkage propagation path, compare the current structure state of the database table field and the supervision reporting form field based on the structure difference evaluation matrix, and identify the synchronization nodes with inconsistent structures. The structure difference includes field existence difference, attribute configuration difference and constraint setting difference, etc.

[0156] Then, according to different field change event types, corresponding adaptive adjustment strategies are adopted. For database table mapping changes, a mapping synchronization adjustment strategy is adopted to update the mapping from the database table fields to the regulatory reporting form fields, thereby realizing regulatory reporting form structure adjustment. Specifically, based on the mapping relationship information in the synchronization node, the field mapping configuration that needs to be updated is identified, including field name mapping, data type mapping, and format conversion mapping, etc. Through a mapping rule update algorithm, new field mapping configuration is generated to ensure that database table field changes can be accurately reflected in the regulatory reporting form. Further, mapping consistency verification is performed to verify the correctness and integrity of the mapping configuration through data sample testing.

[0157] For manual table structure changes, a structure creation adjustment strategy is adopted to adjust the database table structure according to the definition metadata of the regulatory reporting form fields through database table creation, field property setting, and constraint relationship establishment. In a preferred embodiment, based on the specification requirements of the regulatory reporting form fields, corresponding database table structure definitions are generated, including table name determination, field definition, primary key setting, index creation, etc. A structure creation script generation algorithm is used to automatically generate SQL scripts for database table creation to ensure the specification and integrity of the table structure. Optionally, a structure optimization suggestion mechanism is introduced to provide table structure optimization suggestions based on performance optimization and storage efficiency considerations.

[0158] For SQL query logic changes, a logic restructuring adjustment strategy is adopted to adjust the regulatory reporting form structure according to the updated query logic through query result field parsing, dynamic field mapping, and regulatory reporting form structure updating. Specifically, the changed SQL query statement is parsed and analyzed to extract the field structure information of the query result, including field name, data type, calculation logic, etc. Through a dynamic mapping algorithm, the mapping relationship between the query result fields and the regulatory reporting form fields is established to support the mapping processing of complex calculation fields and aggregation fields. Further, the structure definition of the regulatory reporting form is updated according to the mapping relationship to ensure that the form can correctly receive and process query result data.

[0159] Finally, based on the execution results of adaptive bidirectional structure adjustment, a field synchronization construction scheme is generated through a synchronization state verification algorithm. Specifically, a synchronization state evaluation model is established according to the execution of structure adjustment, and the construction of the synchronization state evaluation model includes: extracting the execution state data of the synchronization operation from the structure adjustment execution results, including the adjustment success rate, execution time, error information and other execution condition information; extracting the structure state data from the adjusted database table and the supervision reporting form, including the field mapping relationship, data type matching degree, constraint consistency and other structure information; through a multi-dimensional verification algorithm, the execution state data and the structure state data are comprehensively analyzed, the success rate, data consistency and performance indicators of the synchronization operation are calculated, and the synchronization state evaluation results are generated. Based on the synchronization state evaluation results, quality inspection is carried out through a multi-dimensional verification mechanism, which includes structure integrity verification, data compatibility verification, performance efficiency verification and the like, to ensure that the synchronization construction scheme meets the preset quality standards and reliability requirements. According to the synchronization state evaluation results that pass the verification, a field synchronization construction scheme is generated, which includes field mapping configuration, synchronization rule definition and exception handling strategy. Optionally, a synchronization construction report is generated, which records in detail the adjustment process, execution results and potential risks, providing a reference for subsequent maintenance and optimization.

[0160] S5.5: updating the field change history record through feedback learning, and driving evolution and optimization of the field synchronization construction scheme;

[0161] In this embodiment, based on the execution effect of the field synchronization construction scheme, the bidirectional mapping relationship graph is continuously optimized through a feedback learning mechanism, and the adaptive evolution of the field synchronization construction scheme is realized. Through multi-cycle effect evaluation and relationship adjustment, the accuracy and efficiency of synchronization construction are continuously improved, forming a self-optimizing closed-loop system.

[0162] Specifically, first, according to the execution results of the field synchronization construction scheme, synchronization success rate and data consistency accuracy are calculated through statistical analysis and comparison verification, and synchronization construction effect data is generated. In one embodiment, a synchronization effect evaluation model is established to quantitatively evaluate the execution quality of synchronization construction from multiple dimensions. For the calculation of synchronization success rate, the ratio of the number of successfully completed fields to the total number of synchronization fields is calculated, and the distribution of failure causes and failure types is recorded. For the calculation of data consistency accuracy, through a data sampling verification mechanism, the data content consistency of the database table fields and the supervision reporting form fields before and after synchronization is compared, and the data matching degree and error rate are calculated. Further, performance indicators in the synchronization execution process are collected, including execution time, resource consumption, concurrent processing capacity and the like, to form a comprehensive synchronization construction effect data set.

[0163] Then, according to the synchronization construction effect data, the evolution trend index is generated through multi-period comparative analysis. Specifically, a time series analysis model is established to analyze the trend of the effect data of multiple consecutive synchronization periods, identify the change trend of the synchronization success rate, the improvement degree of data consistency, and the optimization direction of performance indicators, etc. In a preferred embodiment, a sliding window algorithm is used to calculate the change rate of the effect indicators in different time windows, and a trend fitting algorithm is used to predict the future evolution direction. Optionally, an anomaly detection mechanism is introduced to identify abnormal fluctuations and mutation points in the effect data, analyze the abnormal reasons and develop corresponding optimization strategies.

[0164] Next, according to the evolution trend index, the correlation strength in the dynamic weight relationship is updated through feedback learning to obtain the updated correlation strength. In an embodiment, a feedback learning algorithm is established to adjust the correlation strength weight between fields based on the correlation analysis of synchronization construction effect and correlation strength. For field pairs with high synchronization success rate, their correlation strength is enhanced to reflect their stable and reliable mapping relationship; for field pairs with high synchronization failure rate, their correlation strength is reduced to avoid the continuous impact of error mapping. Specifically, a gradient descent algorithm is used to optimize the correlation strength parameters, taking the synchronization construction effect as the objective function, and realizing the precise optimization of the correlation strength through iterative adjustment. Further, a regularization mechanism is introduced to prevent overfitting and ensure the generalization ability and stability of the correlation strength adjustment.

[0165] Then, according to the updated correlation strength, the corresponding semantic similarity matrix is recalculated, and the dynamic weight relationship is updated. Specifically, based on the adjusted correlation strength weight, the cosine similarity calculation is performed again to generate the updated semantic similarity matrix. Through matrix normalization processing, the reasonable distribution and comparability of the similarity values are ensured. At the same time, the weight coefficients in the dynamic weight relationship are updated to reflect the latest field correlation pattern and importance distribution.

[0166] Next, the network parameters are retrained through the graph neural network fusion modeling algorithm to obtain the reconstructed bidirectional mapping relationship graph. In a preferred embodiment, based on the updated semantic similarity matrix and dynamic weight relationship, the graph neural network model is retrained. An incremental learning algorithm is used to integrate new feedback information on the basis of maintaining existing knowledge, avoiding the problem of catastrophic forgetting. The network weights and bias parameters are adjusted through the backpropagation algorithm to optimize the representation ability and prediction accuracy of the graph. Optionally, an ensemble learning method is used to combine multiple graph neural network models with different structures to improve the robustness and accuracy of graph construction.

[0167] Further, according to the reconstructed bidirectional mapping relationship graph, a mapping relationship is established for a field pair with a correlation strength greater than a preset correlation strength threshold, and the mapping relationship is removed for a field pair with a correlation strength less than the preset correlation strength threshold, to generate an evolved bidirectional mapping relationship graph. Specifically, a dynamic correlation strength threshold is set, which is adaptively adjusted based on historical evolution effects and current business requirements. Through the threshold filtering mechanism, high-quality mapping relationships are retained, and low-reliability mapping relationships are removed, to ensure the simplicity and effectiveness of the graph. In an implementation, a threshold optimization algorithm is used to dynamically adjust the correlation strength threshold based on the connectivity and coverage of the graph, to balance the integrity and accuracy of the graph.

[0168] Finally, the evolved bidirectional mapping relationship graph is adaptively optimized by a graph structure-based evolution optimization algorithm, and the optimization is stopped when the evolution trend index meets a preset convergence condition. In an implementation, a graph structure optimization model is established to optimize the structure layout and connection mode of the graph based on the topological properties and node importance of the graph. A community discovery algorithm is used to identify functional modules and correlation clusters in the graph, to improve the organization and interpretability of the graph through modular optimization. Specifically, the convergence condition includes a stability threshold of the evolution trend index, a limit on the number of optimization iterations, and a lower limit on the effect improvement amplitude, and other judgment criteria. When the effect improvement amplitude of consecutive multiple optimization cycles is less than a preset threshold, or the evolution trend index reaches a stable state, it is determined that the optimization process converges, and further evolution optimization operation is stopped.

[0169] Optionally, an evolution history recording mechanism is established to record in detail the parameter adjustment, effect change, and decision basis of each evolution optimization, to form an evolution knowledge base. Through evolution mode analysis, effective optimization strategies and adjustment rules are identified to provide experience guidance for subsequent evolution optimization. Further, a multi-objective optimization mechanism is introduced to find an optimal balance point among multiple objectives such as synchronization success rate, data consistency, and execution efficiency, to achieve comprehensive performance optimization.

[0170] Through the above systematic feedback learning and evolution optimization implementation scheme, continuous improvement of the bidirectional mapping relationship graph and adaptive optimization of the field synchronization construction scheme are realized, which significantly improves the learning ability and adaptability of the system, and provides intelligent technical support for long-term data governance in the regulatory data reporting scenario.

[0171] Those skilled in the art will appreciate that embodiments of the application can be devised for a method, a system, or a computer program product. Accordingly, the present application can be embodied in the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application can take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage devices, etc.) embodying computer readable program code.

[0172] The present application is described in reference to the flowchart and / or block diagrams of the method, apparatus (system) and computer program product according to embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagrams, and a combination of flows and / or blocks in the flowchart and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general purpose computer, a special purpose computer, an embedded processor, or other programmable data processing apparatus to produce a machine, so that the instructions, which are executed via the processor of the computer or other programmable data processing apparatus, generate a means for implementing the functions specified in the flowchart and / or block diagrams of the flowchart and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 means for functionally implementing the

[0173] These computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer readable memory produce an article of manufacture including instructions means which implement the function specified in the flowchart and / or block diagrams of the flowchart and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 means for functionally implementing the

[0174] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process so that the instructions which are executed on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart and / or block diagrams of the flowchart and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 means for functionally implementing the

[0175] The above description is only preferred embodiments of the present application, the protection scope of the present application is not limited to the above-mentioned embodiments, any technical scheme falling within the idea of the present application is within the protection scope of the present application. It should be noted that, for ordinary skilled in the art, some improvements and refinements without departing from the principle of the present application, these improvements and refinements should also be considered as the protection scope of the present application.

Claims

1. A method for synchronously constructing database tables and regulatory reporting form fields through bidirectional linkage, characterized in that: The method includes: Obtain the physical structure metadata of the database tables and the definition metadata of the regulatory reporting forms. Through metadata semantic parsing, identify business semantic features and regulatory semantic features, and generate a two-way semantic feature set. The business semantic features and regulatory semantic features in the bidirectional semantic feature set are transformed into multidimensional semantic vectors through semantic encoding, forming a semantic vector set. Based on the semantic vector set and combined with the field change history, a multidimensional association relationship between database table fields and regulatory reporting form fields is constructed using relation weaving technology, and a bidirectional mapping relationship graph is built. Obtain field change events, determine the influence domain corresponding to the field change events based on the bidirectional mapping relationship graph, and generate linkage propagation paths through parallel analysis algorithms; Based on the aforementioned linkage propagation path, synchronization nodes are determined through consistency assessment to obtain the updated linkage propagation path. Based on the updated linkage propagation path, a field synchronization construction scheme is generated through adaptive bidirectional structural adjustment. The process of constructing a multidimensional association between database table fields and regulatory reporting form fields using relational weaving technology, and building a bidirectional mapping relationship graph, includes: Based on the semantic vectors of each field in the semantic vector set, the semantic similarity between fields is calculated using the cosine similarity algorithm, forming a semantic similarity matrix; Based on the field change history, the change frequency and impact range of the fields are extracted. Through statistical analysis algorithms, the correlation strength between fields is calculated, and a dynamic weight relationship is constructed. Based on the semantic similarity matrix and dynamic weight relationship, a bidirectional mapping relationship graph is constructed using a graph neural network fusion modeling algorithm; The step of determining the influence domain corresponding to the field change event and generating a linkage propagation path based on the bidirectional mapping relationship graph and using a parallel analysis algorithm includes: Based on the field change event, locate the change source node in the bidirectional mapping relationship graph, and identify the influencing node and its corresponding influence level by combining the association strength in the dynamic weight relationship and the graph traversal algorithm. Based on the bidirectional mapping relationship graph, the propagation path and its length from the change source node to each affected node are calculated using a parallel analysis algorithm. Combined with the association strength, the influence weight of the propagation path is calculated. Based on the influencing nodes, their corresponding influence levels, and the influence weights of the propagation paths, influence domains containing different levels are determined through classification and aggregation. These different levels include core influence levels and secondary influence levels. Based on the field change events and their corresponding impact domains, a coordinated propagation path is generated using a differentiated propagation strategy.

2. The method for synchronously constructing database tables and regulatory reporting form fields in a two-way linkage manner according to claim 1, characterized in that, The process involves semantic parsing of metadata to identify business semantic features and regulatory semantic features, generating a bidirectional semantic feature set, including: Obtain the physical structure metadata of the database table, and extract field names, data types, constraints, and relational dependencies through lexical and syntactic analysis to identify business semantic features; Obtain the definition metadata of the regulatory reporting form, and extract the form item identifier, data format, validation rules and business logic through structured parsing to identify regulatory semantic features; Based on business semantic features and regulatory semantic features, a semantic matching algorithm is used to perform association mapping and generate a two-way semantic feature set.

3. The method for synchronously constructing database tables and regulatory reporting form fields in a two-way linkage manner according to claim 1, characterized in that, The field change events are categorized based on the construction method, including database table mapping changes, manual table structure changes, and SQL query logic changes, among which: When the database table mapping is changed to a database table field mapping construction method, the field mapping relationship changes due to the adjustment of the database table structure. When the manual table creation structure is changed to the manual table creation method, the database table structure changes are caused by the maintenance of fields in the regulatory reporting form. When the SQL query logic is changed to use the SQL query construction method, the field generation logic is changed due to the adjustment of the multi-table join query statement.

4. The method for synchronously constructing database tables and regulatory reporting form fields in a two-way linkage manner according to claim 1, characterized in that, The step of generating a coordinated propagation path based on the field change event and its corresponding impact domain, using a differentiated propagation strategy, includes: When the field change event is a database table mapping change, a direct mapping propagation strategy is adopted. Based on the core impact level area in the impact domain, a direct propagation path from the change source node to the target mapping field is generated. When the field change event is a manual table structure change, a structure synchronization propagation strategy is adopted to generate a multi-stage propagation path that includes table structure creation, field attribute adjustment and constraint relationship update according to different levels in the influence domain. When the field change event is an SQL query logic change, a logic reconstruction propagation strategy is adopted. Based on the areas in the influence domain that are secondary influence levels, a recursive propagation path is generated that includes query statement parsing, multi-table association reconstruction, and field generation logic update. Based on the influence weight and influence level of the propagation path, the direct propagation path, multi-stage propagation path, and recursive propagation path are prioritized and optimized to generate a coordinated propagation path.

5. The method for synchronously constructing database tables and regulatory reporting form fields in a two-way linkage manner according to claim 4, characterized in that, The step of determining the synchronization node through consistency evaluation based on the linkage propagation path, and obtaining the updated linkage propagation path, includes: Based on the propagation nodes in the aforementioned linkage propagation path, a consistency assessment is conducted to detect the consistency of the data format, data type, and constraints of each propagation node, resulting in a consistency score. Based on the consistency score, a conflict identification algorithm is used to mark propagation nodes with consistency scores less than a preset threshold as conflict nodes, and to determine propagation nodes with consistency scores greater than or equal to the preset threshold as synchronization nodes. Based on the conflicting node, the propagation branch corresponding to the conflicting node is removed from the linkage propagation path. Based on the synchronization node and the remaining propagation branches in the linkage propagation path, the linkage propagation path is reconstructed to obtain the updated linkage propagation path.

6. The method for synchronously constructing database tables and regulatory reporting form fields in a two-way linkage manner according to claim 5, characterized in that, The consistency assessment involves checking the consistency of data format, data type, and constraints at each propagation node to obtain a consistency score, including: Based on the propagation node, determine the physical structure metadata of the corresponding database table and the definition metadata of the regulatory reporting form, and extract the data format, data type and constraints of each field; Based on the data type of each field, a type compatibility algorithm is used to compare the data type compatibility between the database table fields and the regulatory reporting form fields, and a type compatibility score is calculated. Based on the data format of each field, the degree of matching between data formats is verified by a format consistency algorithm, and a format consistency score is calculated. Based on the constraints of each field, the constraint compatibility algorithm is used to detect the compatibility of the field's non-null constraints, uniqueness constraints, and foreign key constraints, and the constraint compatibility score is calculated. Based on the type compatibility score, format consistency score, and constraint compatibility score, a consistency score for each propagation node is calculated using a weighted fusion algorithm.

7. The method for synchronously constructing database tables and regulatory reporting form fields in a two-way linkage manner according to claim 5, characterized in that, Based on the updated linkage propagation path, an adaptive bidirectional structural adjustment is used to generate a field synchronization construction scheme. This adaptive bidirectional structural adjustment includes adjustments to the database table structure and the regulatory reporting form structure, wherein: Based on the synchronization nodes in the updated linkage propagation path, the corresponding field change events are determined, and the synchronization nodes that make database table structure adjustments and regulatory reporting form structure adjustments are identified through the structural difference analysis algorithm. When the field change event corresponding to the synchronization node is a database table mapping change, a mapping synchronization adjustment strategy is adopted. The structure adjustment of the regulatory reporting form is achieved by directly mapping and updating the database table fields to the regulatory reporting form fields. When the field change event corresponding to the synchronization node is a manual table structure change, a structure creation and adjustment strategy is adopted. Based on the definition metadata of the fields in the regulatory reporting form, the database table structure is adjusted through database table creation, field attribute setting, and constraint relationship establishment. When the field change event corresponding to the synchronization node is a change in SQL query logic, a logic restructuring and adjustment strategy is adopted. Based on the updated query logic, the regulatory reporting form structure is adjusted through query result field parsing, dynamic field mapping, and regulatory reporting form structure update. Based on the execution result of the adaptive bidirectional structure adjustment, a field synchronization construction scheme is generated through a synchronization state verification algorithm.

8. The method for synchronously constructing database tables and regulatory reporting form fields in a two-way linkage manner according to claim 7, characterized in that, The field synchronization construction scheme also includes updating the field change history through feedback learning to drive the evolution and optimization of the field synchronization construction scheme, including: Based on the execution results of the field synchronization construction scheme, the synchronization success rate and data consistency accuracy are calculated through statistical analysis and comparative verification, and synchronization construction effect data is generated. Based on the synchronous construction effect data, an evolutionary trend index is generated through multi-period comparative analysis. Based on the evolutionary trend index, the correlation strength in the dynamic weight relationship is updated through feedback learning to obtain the updated correlation strength. Based on the updated association strength, the corresponding semantic similarity matrix is ​​recalculated, the dynamic weight relationship is updated, and the network parameters are retrained through a graph neural network fusion modeling algorithm to obtain the reconstructed bidirectional mapping relationship graph. Based on the reconstructed bidirectional mapping relationship graph, a mapping relationship is established for field pairs with a correlation strength greater than a preset correlation strength threshold, and the mapping relationship is removed for field pairs with a correlation strength less than the preset correlation strength threshold, thereby generating an evolved bidirectional mapping relationship graph. The evolved bidirectional mapping relationship graph is adaptively optimized using a graph-based evolutionary optimization algorithm, and the optimization stops when the evolutionary trend index meets the preset convergence condition.

Images