Data change prediction method and device, computer equipment and storage medium

Abstract

The invention relates to a data change prediction method and device, computer equipment and a storage medium. The method comprises the following steps: acquiring source data of each source database contained in a measurement database; wherein the source data comprises at least two table files, and each table file comprises at least two fields; constructing a target association graph of the measurement database according to the source data; wherein the target association graph comprises nodes, connecting edges between the nodes and edge weights of the connecting edges; the nodes represent table files or fields, and connection edges between different nodes represent association relationships between different table files and / or fields; performing feature extraction on the target association graph and each piece of source data to obtain multi-dimensional features of the measurement database; and according to the multi-dimensional features and the target association diagram, predicting data change information of the measurement database in a future time period. By adopting the method, the data change information can be efficiently predicted in real time.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a data change prediction method, apparatus, computer equipment, and storage medium. Background Technology

[0002] Data change capture, as a core supporting technology for real-time database synchronization, aims to accurately identify data operations such as insertion, update, and deletion in the data source, ensuring real-time consistency of data between metering systems, which is of crucial significance for ensuring stable system operation.

[0003] Existing data change capture technologies mainly fall into two categories: one is traditional log-based tools that obtain data changes by parsing database transaction logs; the other is trigger- or polling-based methods that capture changes by setting triggers in database tables to respond to changes in real time, or by periodically polling data tables to compare differences. However, in scenarios with large-scale, highly correlated databases, existing technologies result in low accuracy in data change prediction and high synchronization latency, failing to meet the real-time requirements of core business functions such as data scheduling and equipment failure response. Summary of the Invention

[0004] Therefore, it is necessary to provide a data change prediction method, apparatus, computer equipment, and storage medium to address the aforementioned technical problems, which can efficiently and in real time predict data change information.

[0005] Firstly, this application provides a data change prediction method, including:

[0006] Obtain source data from each source database contained in the measurement database; wherein, the source data includes at least two table files, and each table file includes at least two fields;

[0007] Based on the source data, a target association graph of the measurement database is constructed; wherein, the target association graph includes nodes, connecting edges between nodes, and edge weights of the connecting edges; nodes represent table files or fields, and connecting edges between different nodes represent the association relationships between different table files and / or fields;

[0008] Feature extraction is performed on the target association graph and each of the source data to obtain the multidimensional features of the metrology database;

[0009] Based on the multidimensional features and the target association graph, the data change information of the measurement database in future time periods is predicted.

[0010] In one embodiment, constructing the target association graph of the metering database based on each of the source data includes:

[0011] Extract the table file and the fields contained in the table file from each of the source data;

[0012] Based on the extracted table files and fields, construct the nodes in the target association graph;

[0013] Based on the historical business association information between different fields and / or different table files, as well as the historical linkage information and historical change association information between different table files, the connection edges between different nodes in the target association graph are constructed, and the edge weights of the connection edges between different nodes are determined.

[0014] In one embodiment, the connection edge includes a foreign key connection edge, a business logic connection edge, and a change association connection edge;

[0015] The step of constructing connection edges between different nodes in the target association graph based on historical business association information between different fields and / or different table files, as well as historical linkage information and historical change association information between different table files, and determining the edge weights of the connection edges between different nodes includes:

[0016] Based on the historical linkage information between different table files, foreign key connection edges are constructed between nodes in the target association graph corresponding to different table files, and the edge weights of the foreign key connection edges are determined.

[0017] Based on the historical business association information between different fields and / or different table files, construct business logic connection edges between nodes in the target association graph corresponding to different fields and / or different table files, and determine the edge weights of the business logic connection edges.

[0018] Based on the historical change association information between different table files, construct change association connection edges between nodes in the target association graph corresponding to different table files, and determine the edge weight of the change association connection edges.

[0019] In one embodiment, the historical linkage information includes the number of linkage changes;

[0020] The step of constructing foreign key connection edges between nodes in the target association graph corresponding to different table files based on historical linkage information between different table files, and determining the edge weights of the foreign key connection edges, includes:

[0021] For any two table files, after obtaining the historical linkage information between the two table files, construct the foreign key connection edges between the nodes in the target association graph corresponding to the two table files;

[0022] The linkage frequency between the two table files is determined based on the ratio between the number of linkage changes between the two table files and the total number of data changes in the metering database during the historical period.

[0023] The edge weight of the foreign key connection edge is determined based on the linkage frequency.

[0024] In one embodiment, the multidimensional features include basic features, topological features, temporal features, and linkage features;

[0025] The step of extracting features from the target association graph and each of the source data to obtain the multidimensional features of the metrology database includes:

[0026] The source data are encoded to obtain the basic features;

[0027] The topological features are obtained by using a graph embedding algorithm to extract features from the target association graph.

[0028] The time-series features and linkage features are obtained by performing time-series feature analysis on each of the source data.

[0029] In one embodiment, the step of predicting data change information of the metering database in future time periods based on the multidimensional features and the target association graph includes:

[0030] Multidimensional features are fused to obtain fused features;

[0031] Based on a graph attention network, attention features are determined according to the fusion features and the target association graph;

[0032] Based on the graph convolutional network, convolutional features are determined according to the attention features and the target association graph;

[0033] The convolutional features are subjected to a nonlinear transformation to obtain global features;

[0034] By analyzing the global features, the data change information of the metering database in future time periods can be obtained.

[0035] In one embodiment, the data change information includes data change type, associated change topology information, and topology synchronization priority;

[0036] The analysis of the global features to obtain the data change information of the metering database in future time periods includes:

[0037] Based on a preset classifier, the global features are mapped to category probabilities that conform to a preset probability distribution, and the data change type is determined according to the category probabilities.

[0038] The global features and historical data change information are jointly analyzed to obtain the association change topology information; wherein, the association change topology information includes nodes in the target association graph that have changed in the future time period;

[0039] The topology synchronization priority is determined based on the global features and the node importance of each node in the target association graph; wherein, the topology synchronization priority includes the priority of each node in the target association graph for synchronization updates.

[0040] Secondly, this application also provides a data change prediction device, comprising:

[0041] The data acquisition module is used to acquire source data from each source database included in the measurement database; wherein, the source data includes at least two table files, and each table file includes at least two fields;

[0042] The graph construction module is used to construct a target association graph of the measurement database based on the source data. The target association graph includes nodes, connecting edges between nodes, and edge weights of the connecting edges. Nodes represent table files or fields, and connecting edges between different nodes represent the association relationships between different table files and / or fields.

[0043] The feature extraction module is used to extract features from the target association graph and each of the source data to obtain the multidimensional features of the metrology database;

[0044] The information prediction module is used to predict data change information of the measurement database in future time periods based on the multidimensional features and the target association graph.

[0045] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0046] Obtain source data from each source database contained in the measurement database; wherein, the source data includes at least two table files, and each table file includes at least two fields;

[0047] Based on the source data, a target association graph of the measurement database is constructed; wherein, the target association graph includes nodes, connecting edges between nodes, and edge weights of the connecting edges; nodes represent table files or fields, and connecting edges between different nodes represent the association relationships between different table files and / or fields;

[0048] Feature extraction is performed on the target association graph and each of the source data to obtain the multidimensional features of the metrology database;

[0049] Based on the multidimensional features and the target association graph, the data change information of the measurement database in future time periods is predicted.

[0050] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:

[0051] Obtain source data from each source database contained in the measurement database; wherein, the source data includes at least two table files, and each table file includes at least two fields;

[0052] Based on the source data, a target association graph of the measurement database is constructed; wherein, the target association graph includes nodes, connecting edges between nodes, and edge weights of the connecting edges; nodes represent table files or fields, and connecting edges between different nodes represent the association relationships between different table files and / or fields;

[0053] Feature extraction is performed on the target association graph and each of the source data to obtain the multidimensional features of the metrology database;

[0054] Based on the multidimensional features and the target association graph, the data change information of the measurement database in future time periods is predicted.

[0055] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, performs the following steps:

[0056] Obtain source data from each source database contained in the measurement database; wherein, the source data includes at least two table files, and each table file includes at least two fields;

[0057] Based on the source data, a target association graph of the measurement database is constructed; wherein, the target association graph includes nodes, connecting edges between nodes, and edge weights of the connecting edges; nodes represent table files or fields, and connecting edges between different nodes represent the association relationships between different table files and / or fields;

[0058] Feature extraction is performed on the target association graph and each of the source data to obtain the multidimensional features of the metrology database;

[0059] Based on the multidimensional features and the target association graph, the data change information of the measurement database in future time periods is predicted.

[0060] The aforementioned data change prediction method, apparatus, computer equipment, and storage medium acquire source data from various source databases included in the metrology database. Each source data includes at least two table files, and each table file includes at least two fields. Based on the source data, a target association graph of the metrology database is constructed. This target association graph includes nodes, connecting edges between nodes, and edge weights. Nodes represent table files or fields, and connecting edges between different nodes represent the relationships between different table files and / or fields. Feature extraction is performed on the target association graph and each source data to obtain multidimensional features of the metrology database. Based on these multidimensional features and the target association graph, data change information of the metrology database in future time periods is predicted. This scheme presents the table files and fields in the source database as nodes, uses connecting edges and edge weights to represent relationships, transforming the originally scattered metrology data into a visualized association graph that clearly shows the data flow path and dependencies. By combining the relationship features of the target association graph with the multidimensional attributes of the multi-source data, data change information is predicted, hidden patterns in data changes are uncovered, and the prediction results are more consistent with reality. Attached Figure Description

[0061] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0062] Figure 1 This is a flowchart illustrating a data change prediction method in one embodiment;

[0063] Figure 2 This is a flowchart illustrating the process of constructing a target association graph for a measurement database in one embodiment.

[0064] Figure 3 This is a flowchart illustrating the process of determining the edge weights of the connecting edges between different nodes in one embodiment.

[0065] Figure 4 This is a flowchart illustrating the process of determining multidimensional features of a measurement database in one embodiment.

[0066] Figure 5 This is a schematic diagram of a process for predicting data change information in one embodiment;

[0067] Figure 6 This is a flowchart illustrating the process of determining data change information in one embodiment;

[0068] Figure 7 This is a flowchart illustrating the data change prediction method in another embodiment;

[0069] Figure 8 This is a structural block diagram of a data change prediction device in one embodiment;

[0070] Figure 9 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0071] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0072] The data change prediction method provided in this application can be applied to application environments that capture data changes in a metering database. This data change prediction method can be executed by a computer device, which can be a server or a terminal with powerful computing capabilities.

[0073] In one exemplary embodiment, such as Figure 1 As shown, a data change prediction method is provided. Taking the application of this method to a server as an example, the specific steps include:

[0074] S101, Obtain source data from each source database contained in the measurement database.

[0075] The metrology database can be a database system specifically designed for storing, managing, processing, and analyzing metrology-related data. It centers on professional data in the metrology field and integrates various information such as metrology standards, traceability of measurement values, measuring instruments, test results, and calibration data. It serves as the core carrier for the digitalization and intelligentization of metrology work. In this embodiment, the metrology database can be a company's database. The source data includes at least two table files, each containing at least two fields.

[0076] Optionally, source data from each source database can be acquired in real time using methods such as direct sensor connection or synchronous database replication. For example, assume the metering data volume includes... Each source database contains [number] source databases. Zhang's business table, i.e., the table file, is denoted as... The structure of each table file is as follows: ,in, Contains a collection of fields Historical Change Log (Records insert / update / delete operations and related table linkage records for the past 12 months), business association rules wait.

[0077] Furthermore, the acquired data undergoes data cleaning and standardization. Specifically, firstly, abnormal data is filtered out. A rule engine can be used to remove change records from historical change logs that contain format errors (such as missing operation type identifiers) or logical conflicts (such as the same record having both INSERT and DELETE operations), retaining only valid logs. Next, format standardization is performed. Valid logs and business association rules are integrated into a standardized log dataset: "Operation Type (INSERT / UPDATE / DELETE) - Table Name - Field Change Values ​​Before and After - Timestamp - Related Table ID - Association Rule Type," supporting adaptation to multiple database formats such as MySQL, PostgreSQL, and Oracle. Finally, for the identifiable related table IDs in the standardized log dataset, reverse lookups using business association rules are used to complete the dataset.

[0078] S102, Based on the source data, construct the target association graph of the measurement database.

[0079] The target association graph includes nodes, connecting edges between nodes, and edge weights of the connecting edges; nodes represent table files or fields, and connecting edges between different nodes represent the association relationships between different table files and / or fields.

[0080] Optionally, table files and core fields from the source data can be used as nodes in the target association graph. Table nodes can be labeled with "table-source database" attribution, and field nodes can be labeled with "field-table-source database" attribution. Furthermore, the relationships between nodes can be identified based on business logic and data association rules, and connection edges between nodes can be constructed. Further, edge weights are used to quantify the strength of associations between nodes. In this embodiment, edge weights can be calculated using a multi-dimensional weighted method, combining business characteristics and data statistical characteristics.

[0081] Furthermore, graph databases or visualization tools can be used to draw the target association graph, with nodes representing table files or fields and weighted directed edges representing associations, labeling the attributes of nodes and connecting edges. The structural data of the target association graph is stored in a graph database to support subsequent graph structure analysis and feature extraction.

[0082] For example, a target association graph can be represented as ;in, Represents the nodes in the target association graph. Including table nodes and field nodes . This represents the connecting edges in the target association graph. This represents the edge weight of the connecting edges. In this embodiment, the target association graph can be updated every 24 hours based on the latest change log to ensure that the target association graph reflects business dynamics in real time.

[0083] S103, extract features from the target association graph and each source data to obtain multidimensional features of the econometric database.

[0084] In the embodiments of this application, the multidimensional features of the metering database include, but are not limited to, basic features, topological features, temporal features, and linkage features.

[0085] Optionally, based on the topological characteristics of the target graph, quantitative structural features, such as node features, edge features, and global graph features, can be extracted; based on the actual values ​​of table files and fields, basic business features can be extracted; combined with the time-series change patterns of measurement data, time-dimensional features can be extracted; and based on historical change logs and business association rules, linkage features between table files or fields can also be extracted.

[0086] S104. Based on multidimensional features and target association diagrams, predict data change information of the econometric database in future periods.

[0087] In this embodiment, the data change information includes, but is not limited to, data change type, associated change topology information, and topology synchronization priority; associated change topology information includes nodes that change in the target association graph in the future time period; topology synchronization priority includes the priority of each node in the target association graph for synchronous updates.

[0088] Optionally, multi-dimensional features can be fused first to obtain fused features. Further, based on the predicted target type in the data change information, and combining the target association graph and the fused features, an appropriate algorithm can be selected to construct a model. For example, regarding the data change type in the data change information, a time-series prediction model can be used to predict the data change type based on the fused features and the target association graph; a graph neural network can be used to predict the association change topology information and topology synchronization priority based on the topology structure of the target association graph and the fused features, ultimately obtaining the data change information.

[0089] The aforementioned data change prediction method involves acquiring source data from various source databases within the econometric database. Each source data includes at least two table files, and each table file contains at least two fields. Based on the source data, a target association graph for the econometric database is constructed. This target association graph includes nodes, connecting edges between nodes, and edge weights. Nodes represent table files or fields, and connecting edges between different nodes represent the relationships between different table files and / or fields. Feature extraction is performed on the target association graph and each source data to obtain multidimensional features of the econometric database. Based on these multidimensional features and the target association graph, data change information for the econometric database in future time periods is predicted. This approach presents the table files and fields in the source database as nodes, uses connecting edges and edge weights to represent relationships, transforming the originally scattered econometric data into a visualized association graph that clearly shows the data flow path and dependencies. By combining the relationship features of the target association graph with the multidimensional attributes of the multi-source data, data change information is predicted, uncovering hidden patterns in data changes and making the prediction results more realistic.

[0090] Optionally, in one embodiment, such as Figure 2 As shown, a method for constructing a target association graph of a measurement database is provided, which specifically includes the following steps:

[0091] S201, extract the table file and the fields contained in the table file from each source data.

[0092] Optionally, in this embodiment of the application, an appropriate extraction method is adopted for different types of source databases to obtain both table-level and field-level information.

[0093] S202, construct the nodes in the target association graph based on the extracted table files and fields.

[0094] In this embodiment of the application, the extracted table files and fields can be used as nodes in the target association graph. Table nodes can be labeled with "table-source database" attribution, and field nodes can be labeled with "field-table-source database" attribution.

[0095] S203. Based on the historical business association information between different fields and / or different table files, as well as the historical linkage information and historical change association information between different table files, construct the connection edges between different nodes in the target association graph, and determine the edge weights of the connection edges between different nodes.

[0096] Optionally, one can mine the relationships between nodes from two dimensions: business logic relationships and actual data relationships. This ensures that the relationships conform to business rules and reflect the actual data flow patterns. Specifically, one can mine the inherent logical relationships between nodes based on business processes; one can also analyze the linkage update patterns and change relationship patterns of nodes in historical data to uncover hidden dynamic relationships, which reflect the actual data flow characteristics. Alternatively, one can integrate logical and data relationships to construct connection edges in a relationship graph, classify them according to the nature of the relationships, and clarify the attributes of the edges to ensure that the relationships are clear and traceable.

[0097] Furthermore, edge weights are used to quantify the strength of the association between nodes. The higher the weight value, the closer the association between nodes and the greater their influence on each other. This step, combined with the characteristics of metering business, constructs a weight calculation model from three dimensions: business importance, data association frequency, and change impact, to ensure that the weight values ​​objectively reflect the inherent association of metering data.

[0098] In this embodiment, the source data is decomposed into two levels of nodes: table files and fields. Relationship edges are constructed in conjunction with measurement business rules, transforming the data that was originally scattered across different source databases into a visualized relationship graph. The edge weights are calculated based on three dimensions: business importance, data association frequency, and impact of changes. This approach considers both the hard constraints of measurement regulations and business processes and incorporates the actual flow patterns of historical data. Compared to a single qualitative judgment of whether a relationship exists, the quantified weight values ​​can accurately reflect the degree of connection between nodes.

[0099] Optionally, in an exemplary embodiment, the connection edge includes a foreign key connection edge, a business logic connection edge, and a change-related connection edge; based on this, such as Figure 3 As shown, a method for determining the edge weight of a connection between different nodes is provided, which specifically includes the following steps:

[0100] S301, Based on the historical linkage information between different table files, construct foreign key connection edges between nodes in the target association graph corresponding to different table files, and determine the edge weights of the foreign key connection edges.

[0101] Historical linkage information includes the number of linkage changes.

[0102] Optionally, in this embodiment, for any two table files, after obtaining the historical linkage information between the two table files, a foreign key connection edge can be constructed between nodes in the target association graph corresponding to the two table files; the linkage frequency between the two table files is determined based on the ratio between the number of linkage changes between the two table files and the total number of data changes in the metering database within the historical period; and the edge weight of the foreign key connection edge is determined based on the linkage frequency. Specifically, the edge weight of the foreign key connection edge can be expressed as:

[0103]

[0104]

[0105] in, The edge weight of the foreign key connection edge; For two table files and The frequency of interaction between them; For two table files and Number of times the changes are linked between them; This represents the total number of data changes.

[0106] S302, Based on the historical business association information between different fields and / or different table files, construct business logic connection edges between nodes in the target association graph corresponding to different fields and / or different table files, and determine the edge weights of the business logic connection edges.

[0107] Optionally, business logic connection edges can be constructed between nodes in the target association graph corresponding to different fields based on historical business association information between different fields, and business logic connection edges can also be constructed between nodes in the target association graph corresponding to different table files based on historical business association information between different table files.

[0108] Specifically, in this embodiment, the importance score of business associations between different table files or fields can be determined based on historical business association information. Furthermore, the edge weight of the business logic connection edge can be expressed as:

[0109]

[0110] in, The edge weights for connecting edges in the business logic; Score the importance of business relevance.

[0111] S303, Based on the historical change association information between different table files, construct change association connection edges between nodes in the target association graph corresponding to different table files, and determine the edge weights of the change association connection edges.

[0112] Optionally, for any two table files, the probability that a change in one table file will trigger a change in the other table file can be determined based on historical change association information. Furthermore, the edge weight of the change association connection edge can be expressed as:

[0113]

[0114] in, To change the edge weights of associated connections; For table files Changes triggered in the form file The probability of change.

[0115] In this embodiment, connection edges are constructed according to different data association characteristics. Foreign key connection edges lock hard data associations between tables, business logic connection edges map soft rule associations, and change association connection edges capture dynamic causal associations. The three types of edges together form a complete association network covering static structure and dynamic flow, making the association relationships that were originally hidden in multi-source data explicit, and solving the problem of mixed association types in measurement data and difficulty in accurate identification.

[0116] Optionally, in one embodiment, the multidimensional features include basic features, topological features, temporal features, and linkage features; based on this, such as Figure 4 As shown, a method for determining multidimensional features of a metrological database is provided, specifically including the following steps:

[0117] S401, encodes each source data to obtain basic features.

[0118] Optionally, encoding and statistical methods can be used to process the source data. For example, One-Hot encoding (INSERT=[1,0,0], UPDATE=[0,1,0], DELETE=[0,0,1]) can be used to encode each source data point, obtaining basic characteristics such as field change frequency and data volume features. Specifically, the number of field changes within the last 30 minutes can be counted to determine the field change frequency; the number of rows involved in a single change can be counted to determine the data volume features.

[0119] S402 uses a graph embedding algorithm to extract features from the target association graph and obtain topological features.

[0120] Optionally, a pre-defined graph embedding algorithm can be used to extract features from the target association graph, obtaining topological features such as node importance, association path features, and neighbor node features. For example, the PageRank algorithm can be used to calculate the importance score of table file nodes or field nodes in the target association graph, analyzing the node importance of these nodes. Furthermore, the shortest path length from the currently modified node to the core table node can be calculated to determine association path features; the number of first-order neighbor nodes of the current node can be extracted to extract neighbor node features.

[0121] S403 performs time-series feature analysis on each source data to obtain time-series features and linkage features.

[0122] Optionally, time-series analysis can be performed on each source data to extract time-series features such as change time intervals, time-series trends, and peak correlations. For example, the time difference between the current change and the last change in the same field can be analyzed to extract the change time interval; the variance of changes in the same field within the past hour can be calculated (variance > 0.8 indicates frequent changes, < 0.3 indicates stability) to extract time-series trend features; and it can be determined whether the current change occurred during peak electricity consumption hours, marking it as 1 if yes and 0 otherwise to extract peak correlation features.

[0123] In addition, historical logs from various source data can be statistically analyzed to extract linkage features such as linkage probability, linkage delay, and linkage range. For example, the probability of a current change triggering changes in related tables can be analyzed (calculated based on linkage records from the past 6 months in historical logs) to extract linkage probability features; the average time difference between the current change and the first change in the related tables can be analyzed to extract linkage delay features; and the average number of tables linked by the current change can be analyzed to extract linkage range features.

[0124] In this embodiment, the encoding processing of each source data essentially transforms fragmented and heterogeneous measurement data into basic features in a unified format; the graph embedding algorithm is used to extract topological features from the target association graph, the core of which is to transform the association structure of table files or field nodes into quantitative features; the temporal feature analysis of the source data simultaneously produces temporal features reflecting data changes over time and linkage features reflecting data interaction patterns, upgrading the analysis of measurement data from static description to dynamic prediction, providing support for the forward-looking management of measurement business.

[0125] Optionally, in one embodiment, such as Figure 5 As shown, a method for predicting data changes in a metering database over future periods is provided, specifically including the following steps:

[0126] S501 performs feature fusion on multidimensional features to obtain fused features.

[0127] Optionally, the topological features can be transformed into 64-dimensional node embedding vectors through a graph convolutional network; the basic features, temporal features, and linkage features can be mapped into 32-dimensional vectors through a fully connected layer; further, the node embedding vectors and the three 32-dimensional vectors can be concatenated into a 160-dimensional initial feature vector, and the initial feature vector can be normalized to first identify the differences in feature dimensions and obtain the fused features.

[0128] S502, based on graph attention network, determines attention features according to fusion features and target association graph.

[0129] S503, based on graph convolutional networks, determines convolutional features based on attention features and target association graphs.

[0130] S504 performs a nonlinear transformation on the convolutional features to obtain global features.

[0131] S505 analyzes global features to obtain data change information for the metrology database in future time periods.

[0132] Optionally, a multi-task graph neural network architecture with a graph feature sharing layer and a task-specific output layer can be pre-trained to simultaneously perform the tasks of feature extraction and data change information prediction, thereby improving the accuracy and real-time performance of association modeling.

[0133] In this embodiment, the graph feature sharing layer can be implemented using a stacked structure of a graph attention network, a graph convolutional network, and a fully connected layer. For example, the first layer is a graph attention network. In this layer, the fused features and the target association graph are input, and the attention weights between nodes in the target association graph are calculated through an attention mechanism to highlight the feature contributions of key nodes, outputting 64-dimensional attention features. The second layer is a graph convolutional layer. In this layer, convolution operations are performed based on the attention features and the target association graph to capture the global topological information of the association graph, outputting 32-dimensional convolutional features. For example, the formula for graph convolution operations can be expressed as:

[0134]

[0135] in, The normalized adjacency matrix of the target association graph; and These are learnable weights; This is an attention feature.

[0136] The third layer is a fully connected layer. In this network layer, the convolutional features are subjected to a non-linear transformation (ReLU activation function) to output 24-dimensional global features.

[0137] The task-specific output layer implements task-based mapping of features and outputs data change information of the econometric database for future periods. In this embodiment, the data change information can include multi-dimensional information; therefore, for different dimensions of information, dedicated output layers for each dimension can be designed in the task-specific output layer.

[0138] In this embodiment, graph attention networks and graph convolutional networks are used to fully explore the topological structure and fusion features of the target association graph, which is more in line with the association characteristics of the measurement data compared with traditional machine learning models. The attention mechanism of graph attention networks can automatically identify the node associations that play a key role in data changes, avoiding the subjectivity of manually setting weights. By aggregating local features and integrating global features through nonlinear transformations, graph convolutional networks can accurately capture the overall change patterns of the measurement database.

[0139] It should be noted that the multi-task graph neural network architecture provided in the above embodiments can be jointly trained and optimized. In this embodiment, 10 linear historical change logs can be collected from the econometric database to construct a dataset containing 2 million labeled samples as the training dataset; the AdamW optimizer (initial learning rate 0.0005, weight decay 0.001) is used, and an early stopping mechanism is employed (training stops if the validation set loss does not decrease for 5 consecutive rounds), iterating for 80 rounds until the loss converges (training set loss < 0.15, validation set loss < 0.18).

[0140] Optionally, in one embodiment, the data change information includes the data change type, associated change topology information, and topology synchronization priority; based on this, such as Figure 6 As shown, a method for determining data change information is provided, which specifically includes the following steps:

[0141] S601, based on a preset classifier, maps global features to category probabilities that conform to a preset probability distribution, and determines the data change type based on the category probabilities.

[0142] The preset classifier can be a Softmax classifier, and the loss function can be cross-entropy loss.

[0143] Optionally, global features can be input into a preset classifier to output the probability distribution of data change types (INSERT / UPDATE / DELETE). (like =[0.98,0.01,0.01] represents an INSERT operation.

[0144] S602 performs joint analysis on global features and historical data change information to obtain associated change topology information.

[0145] Among them, the association change topology information includes nodes that will change in the target association graph in the future time period.

[0146] Optionally, global features and historical data change information can be input into the graph convolutional-long short-term memory (LSTM) temporal network to output a node-field topology matrix that may undergo association changes in future time periods. The LSTM network has a hidden layer dimension of 24, and the loss function can be binary cross-entropy loss.

[0147] S603 determines the topology synchronization priority based on global features and the importance of each node in the target association graph.

[0148] Among them, the topology synchronization priority includes the priority of each node in the target association graph for synchronization updates.

[0149] Optionally, the importance of each node in the global feature and target association graph can be input into the Sigmoid regressor to output the topology synchronization priority, and the loss function can be the mean squared error loss.

[0150] In this embodiment, a classifier is used to quantitatively determine the type of change, and association mining is used to locate change nodes, avoiding blind investigation and improving the targeting of change management. Synchronization priority is set based on the importance of nodes to ensure that core metering business nodes are updated first, reducing the interference of topology synchronization on business. High-probability change nodes and propagation paths are identified in advance, and contingency plans can be formulated in a targeted manner to avoid problems such as metering data disconnection and business interruption caused by changes.

[0151] In this embodiment, the data capture strategy can be dynamically adjusted based on data change information and the topological features of the target association graph, reducing source database dependence and resource consumption. Specifically, the intelligent log parsing strategy can be dynamically adjusted, topological path association pre-capture can be performed, and topological weight-driven priority scheduling can be implemented. For example, regarding the intelligent log parsing strategy, based on the data change type, for UPDATE operations, only key fields with a weight greater than or equal to 0.8 in the global features can be parsed, skipping non-key fields; for DELETE operations, only table name-primary key information can be recorded, without storing field change values. Based on the node importance score, a real-time incremental parsing strategy is adopted for core nodes in the target association graph with an importance score greater than or equal to 0.8; for non-core nodes with a score less than 0.3, a batch parsing scheme (e.g., merging and parsing once every 60 seconds) is adopted to balance real-time performance and resource consumption.

[0152] For topology path association pre-capture, the topology path from the current changed node to a high-probability associated node can be extracted based on the association change topology information; for first-order associated nodes on the path, a field snapshot of the last 10 minutes is cached locally; for second- and third-order associated nodes, the mapping relationship between their primary key and key fields is cached; when an association change occurs, the basic data is read directly from the cache without querying the source database; the cached node range is updated every 30 seconds based on the latest association change topology information to avoid invalid cache occupying storage space.

[0153] For topology weight-driven priority scheduling, priority stratification can be performed in advance, and changes can be divided into three priorities based on topology synchronization priority and node importance score:

[0154] High priority (topology synchronization priority ≥ 0.8 and node importance ≥ 0.7): Critical changes are handled using a "real-time push + redundant backup" mode.

[0155] Medium priority (0.5 ≤ topology synchronization priority < 0.8): Regular changes are handled using a "near real-time push" mode.

[0156] Low priority (topology synchronization priority < 0.5): For statistical / log changes, use the "batch merge push" mode.

[0157] When high-priority changes compete with low-priority changes for computing power, a topology weight preemption mechanism is triggered (the weight of a high-priority change is the topology synchronization priority × the importance of the node, giving it priority in occupying computing resources), ensuring that critical changes are not delayed.

[0158] In one embodiment, the performance of the data change prediction method provided in this application is verified through experiments. For example, four typical source databases from a bibliometric database are selected as experimental objects. The performance metrics of the traditional FlinkCDC tool, the conventional DNN-DCC (Deep Neural Network - Dynamic Conditional Correlation) scheme, and the scheme of this application are compared. Basic information and results of the experimental datasets are shown in Table 1, and performance metric comparisons are shown in Table 2.

[0159] Table 1 Basic Information of Experimental Dataset

[0160] Source Database Number of table files (pages) Average number of changes per day (in ten thousand) Number of related tables (pages / table) Average number of fields per change (number of fields) DB1 28 12.5 3-5 8-12 DB2 45 28.3 5-8 10-15 DB3 62 45.7 2-6 6-10 DB4 18 8.2 3-4 7-9

[0161] Table 2 Performance Comparison Results

[0162] Performance indicators Traditional FlinkCDC Conventional DNN-DCC This application proposal Increase Average synchronization delay 185 92 48 74.6% Related change identification rate (%) 68.3 91.2 96.5 41.3% Source database CPU utilization (%) 22.5 15.8 8.3 63.1% Change priority determination accuracy (%) 72.8 85.5 94.2 29.4%

[0163] As shown in Table 2, the proposed solution is significantly superior to traditional and conventional solutions in terms of synchronization delay, correlation identification accuracy, and resource consumption control, and fully meets the real-time and accuracy requirements of the metering system database synchronization.

[0164] Figure 7 This is a flowchart illustrating a data change prediction method in another embodiment. Based on the above embodiments, this embodiment provides an optional example of a data change prediction method. (Combined with...) Figure 7 The specific implementation process is as follows:

[0165] S701, Obtain source data from each source database contained in the measurement database.

[0166] The source data includes at least two table files, and each table file includes at least two fields.

[0167] S702, extract table files and the fields contained in the table files from each source data.

[0168] S703, construct the nodes in the target association graph based on the extracted table files and fields.

[0169] S704. Based on the historical business association information between different fields and / or different table files, as well as the historical linkage information and historical change association information between different table files, construct the connection edges between different nodes in the target association graph, and determine the edge weights of the connection edges between different nodes.

[0170] The connection edges include foreign key connection edges, business logic connection edges, and change-related connection edges.

[0171] Optionally, based on historical linkage information between different table files, foreign key connection edges are constructed between nodes in the target association graph corresponding to different table files, and the edge weights of the foreign key connection edges are determined; based on historical business association information between different fields and / or different table files, business logic connection edges are constructed between nodes in the target association graph corresponding to different fields and / or different table files, and the edge weights of the business logic connection edges are determined; based on historical change association information between different table files, change association connection edges are constructed between nodes in the target association graph corresponding to different table files, and the edge weights of the change association connection edges are determined.

[0172] Historical linkage information includes the number of linkage changes.

[0173] Optionally, for any two table files, after obtaining the historical linkage information between the two table files, construct foreign key connection edges between nodes in the target association graph corresponding to the two table files; determine the linkage frequency between the two table files based on the ratio between the number of linkage changes between the two table files and the total number of data changes in the metering database during the historical period; and determine the edge weight of the foreign key connection edge based on the linkage frequency.

[0174] S705 encodes each source data to obtain the basic features in the multidimensional features.

[0175] S706 employs a graph embedding algorithm to extract features from the target association graph, obtaining topological features from the multidimensional features.

[0176] S707 performs time-series feature analysis on each source data to obtain time-series features and linkage features in the multi-dimensional features.

[0177] S708 performs feature fusion on multidimensional features to obtain fused features.

[0178] S709, based on graph attention network, determines attention features according to fusion features and target association graph.

[0179] S710, based on graph convolutional networks, determines convolutional features based on attention features and target association graphs.

[0180] S711 performs a nonlinear transformation on the convolutional features to obtain global features.

[0181] S712 analyzes global features to obtain data change information of the metrology database in future periods.

[0182] The data change information includes the data change type, associated change topology information, and topology synchronization priority.

[0183] Optionally, based on a preset classifier, global features are mapped to category probabilities that conform to a preset probability distribution, and the data change type is determined according to the category probabilities; joint analysis is performed on global features and historical data change information to obtain associated change topology information; wherein, associated change topology information includes nodes that change in the target association graph in the future time period; the topology synchronization priority is determined according to the importance of each node in the global features and the target association graph; wherein, the topology synchronization priority includes the priority of each node in the target association graph to be synchronized and updated.

[0184] The specific processes of S701-S712 described above can be found in the description of the above method embodiments. Their implementation principles and technical effects are similar and will not be repeated here.

[0185] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0186] Based on the same inventive concept, this application also provides a data change prediction apparatus for implementing the data change prediction method described above. The solution provided by this apparatus is similar to the implementation scheme described in the above method; therefore, the specific limitations in one or more data change prediction apparatus embodiments provided below can be found in the limitations of the data change prediction method described above, and will not be repeated here.

[0187] In one exemplary embodiment, such as Figure 8 As shown, a data change prediction device 800 is provided, including: a data acquisition module 810, a graph construction module 820, a feature extraction module 830, and an information prediction module 840, wherein:

[0188] The data acquisition module 810 is used to acquire source data from each source database contained in the measurement database; wherein the source data includes at least two table files, and each table file includes at least two fields.

[0189] The graph construction module 820 is used to construct a target association graph of the econometric database based on the source data. The target association graph includes nodes, connecting edges between nodes, and edge weights of the connecting edges. Nodes represent table files or fields, and connecting edges between different nodes represent the association relationships between different table files and / or fields.

[0190] The feature extraction module 830 is used to extract features from the target association graph and each source data to obtain multidimensional features of the econometric database.

[0191] The information prediction module 840 is used to predict data change information of the econometric database in future periods based on multidimensional features and target association graphs.

[0192] The aforementioned data change prediction device acquires source data from various source databases included in the econometric database. Each source data includes at least two table files, and each table file includes at least two fields. Based on the source data, a target association graph of the econometric database is constructed. This target association graph includes nodes, connecting edges between nodes, and edge weights. Nodes represent table files or fields, and connecting edges between different nodes represent the relationships between different table files and / or fields. Feature extraction is performed on the target association graph and each source data to obtain multidimensional features of the econometric database. Based on these multidimensional features and the target association graph, data change information of the econometric database in future time periods is predicted. This scheme presents the table files and fields in the source database as nodes, uses connecting edges and edge weights to represent relationships, transforming the originally scattered econometric data into a visualized association graph that clearly shows the data flow path and dependencies. By combining the relationship features of the target association graph with the multidimensional attributes of the multi-source data, data change information is predicted, uncovering hidden patterns in data changes and making the prediction results more consistent with reality.

[0193] In one embodiment, the graph construction module 820 includes:

[0194] The information extraction unit is used to extract table files and the fields contained in the table files from various source data.

[0195] The node building unit is used to construct nodes in the target association graph based on the extracted table files and fields.

[0196] The edge construction unit is used to construct the connection edges between different nodes in the target association graph based on the historical business association information between different fields and / or different table files, as well as the historical linkage information and historical change association information between different table files, and to determine the edge weights of the connection edges between different nodes.

[0197] In one embodiment, the connection edge includes a foreign key connection edge, a business logic connection edge, and a change-related connection edge; the edge construction unit includes:

[0198] The first construction unit is used to construct foreign key connection edges between nodes in the target association graph corresponding to different table files based on the historical linkage information between different table files, and to determine the edge weight of the foreign key connection edges.

[0199] The second construction unit is used to construct business logic connection edges between nodes in the target association graph corresponding to different fields and / or different table files based on historical business association information between different fields and / or different table files, and to determine the edge weights of the business logic connection edges.

[0200] The third construction unit is used to construct change association connection edges between nodes in the target association graph corresponding to different table files based on the historical change association information between different table files, and to determine the edge weight of the change association connection edges.

[0201] In one embodiment, historical linkage information includes the number of linkage changes; the first construction unit is specifically used for:

[0202] For any two table files, after obtaining the historical linkage information between the two table files, construct foreign key connection edges between nodes in the target association graph corresponding to the two table files; determine the linkage frequency between the two table files based on the ratio between the number of linkage changes between the two table files and the total number of data changes in the metering database during the historical period; determine the edge weight of the foreign key connection edge based on the linkage frequency.

[0203] In one embodiment, the multidimensional features include basic features, topological features, temporal features, and linkage features; the feature extraction module 830 is specifically used for:

[0204] The source data are encoded to obtain basic features; a graph embedding algorithm is used to extract features from the target association graph to obtain topological features; and temporal feature analysis is performed on the source data to obtain temporal features and linkage features.

[0205] In one embodiment, the information prediction module 840 includes:

[0206] The feature fusion unit is used to fuse multidimensional features to obtain fused features.

[0207] The first extraction unit is used to determine attention features based on the graph attention network, according to the fusion features and the target association graph.

[0208] The second extraction unit is used to determine convolutional features based on the graph convolutional network, according to the attention features and the target association graph.

[0209] The feature extraction unit is used to perform nonlinear transformations on the convolutional features to obtain global features.

[0210] The information prediction unit is used to analyze global features and obtain data change information of the metrology database in future periods.

[0211] In one embodiment, the data change information includes the data change type, associated change topology information, and topology synchronization priority; the information prediction unit is specifically used for:

[0212] Based on a preset classifier, global features are mapped to category probabilities that conform to a preset probability distribution, and the data change type is determined according to the category probabilities. The global features and historical data change information are jointly analyzed to obtain the associated change topology information. The associated change topology information includes the nodes that will change in the target association graph in the future time period. The topology synchronization priority is determined according to the importance of each node in the global features and the target association graph. The topology synchronization priority includes the priority of each node in the target association graph to be synchronized and updated.

[0213] Each module in the aforementioned data change prediction device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0214] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 9As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs stored in the non-volatile storage media to run. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it implements a data change prediction method.

[0215] Those skilled in the art will understand that Figure 9 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0216] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0217] Obtain the source data from each source database contained in the measurement database; wherein the source data includes at least two table files, and each table file includes at least two fields;

[0218] Based on the source data, construct the target association graph of the econometric database; the target association graph includes nodes, connecting edges between nodes, and edge weights of connecting edges; nodes represent table files or fields, and connecting edges between different nodes represent the association relationships between different table files and / or fields;

[0219] Feature extraction is performed on the target association graph and each source data to obtain multidimensional features of the econometric database;

[0220] Based on multidimensional features and target association diagrams, we predict data change information in the econometric database for future periods.

[0221] In one embodiment, when the processor executes a computer program to construct a target association graph of the metering database based on the source data, it also performs the following steps:

[0222] Extract table files and their fields from the source data; construct nodes in the target association graph based on the extracted table files and fields; construct connection edges between different nodes in the target association graph based on historical business association information between different fields and / or different table files, as well as historical linkage information and historical change association information between different table files, and determine the edge weights of the connection edges between different nodes.

[0223] In one embodiment, the connection edges include foreign key connection edges, business logic connection edges, and change-related connection edges; when the processor executes the computer program to construct the connection edges between different nodes in the target association graph based on historical business association information between different fields and / or different table files, as well as historical linkage information and historical change-related information between different table files, and determines the edge weights of the connection edges between different nodes, it also implements the following steps:

[0224] Based on historical linkage information between different table files, construct foreign key connection edges between nodes in the target association graph corresponding to different table files, and determine the edge weights of the foreign key connection edges; based on historical business association information between different fields and / or different table files, construct business logic connection edges between nodes in the target association graph corresponding to different fields and / or different table files, and determine the edge weights of the business logic connection edges; based on historical change association information between different table files, construct change association connection edges between nodes in the target association graph corresponding to different table files, and determine the edge weights of the change association connection edges.

[0225] In one embodiment, historical linkage information includes the number of linkage changes; when the processor executes the computer program to construct foreign key connection edges between nodes in the target association graph corresponding to different table files based on the historical linkage information between different table files, and determines the edge weights of the foreign key connection edges, it also implements the following steps:

[0226] For any two table files, after obtaining the historical linkage information between the two table files, construct foreign key connection edges between nodes in the target association graph corresponding to the two table files; determine the linkage frequency between the two table files based on the ratio between the number of linkage changes between the two table files and the total number of data changes in the metering database during the historical period; determine the edge weight of the foreign key connection edge based on the linkage frequency.

[0227] In one embodiment, the multidimensional features include basic features, topological features, temporal features, and linkage features; when the processor executes a computer program to extract features from the target association graph and each source data to obtain the multidimensional features of the metrology database, it also performs the following steps:

[0228] The source data are encoded to obtain basic features; a graph embedding algorithm is used to extract features from the target association graph to obtain topological features; and temporal feature analysis is performed on the source data to obtain temporal features and linkage features.

[0229] In one embodiment, when the processor executes a computer program to predict data change information of the metering database for future periods based on multidimensional features and target association graphs, it also performs the following steps:

[0230] Multidimensional features are fused to obtain fused features; attention features are determined based on the fused features and the target association graph using a graph attention network; convolution features are determined based on the attention features and the target association graph using a graph convolutional network; nonlinear transformations are applied to the convolutional features to obtain global features; and the global features are analyzed to obtain data change information of the econometric database in future time periods.

[0231] In one embodiment, the data change information includes the data change type, associated change topology information, and topology synchronization priority; when the processor executes a computer program to analyze global features and obtains the data change information of the metering database for a future period, it also performs the following steps:

[0232] Based on a preset classifier, global features are mapped to category probabilities that conform to a preset probability distribution, and the data change type is determined according to the category probabilities. The global features and historical data change information are jointly analyzed to obtain the associated change topology information. The associated change topology information includes the nodes that will change in the target association graph in the future time period. The topology synchronization priority is determined according to the importance of each node in the global features and the target association graph. The topology synchronization priority includes the priority of each node in the target association graph to be synchronized and updated.

[0233] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:

[0234] Obtain the source data from each source database contained in the measurement database; wherein the source data includes at least two table files, and each table file includes at least two fields;

[0235] Based on the source data, construct the target association graph of the econometric database; the target association graph includes nodes, connecting edges between nodes, and edge weights of connecting edges; nodes represent table files or fields, and connecting edges between different nodes represent the association relationships between different table files and / or fields;

[0236] Feature extraction is performed on the target association graph and each source data to obtain multidimensional features of the econometric database;

[0237] Based on multidimensional features and target association diagrams, we predict data change information in the econometric database for future periods.

[0238] In one embodiment, when the processor executes a computer program to construct a target association graph of the metering database based on the source data, it also performs the following steps:

[0239] Extract table files and their fields from the source data; construct nodes in the target association graph based on the extracted table files and fields; construct connection edges between different nodes in the target association graph based on historical business association information between different fields and / or different table files, as well as historical linkage information and historical change association information between different table files, and determine the edge weights of the connection edges between different nodes.

[0240] In one embodiment, the connection edges include foreign key connection edges, business logic connection edges, and change-related connection edges; when the processor executes the computer program to construct the connection edges between different nodes in the target association graph based on historical business association information between different fields and / or different table files, as well as historical linkage information and historical change-related information between different table files, and determines the edge weights of the connection edges between different nodes, it also implements the following steps:

[0241] Based on historical linkage information between different table files, construct foreign key connection edges between nodes in the target association graph corresponding to different table files, and determine the edge weights of the foreign key connection edges; based on historical business association information between different fields and / or different table files, construct business logic connection edges between nodes in the target association graph corresponding to different fields and / or different table files, and determine the edge weights of the business logic connection edges; based on historical change association information between different table files, construct change association connection edges between nodes in the target association graph corresponding to different table files, and determine the edge weights of the change association connection edges.

[0242] In one embodiment, historical linkage information includes the number of linkage changes; when the processor executes the computer program to construct foreign key connection edges between nodes in the target association graph corresponding to different table files based on the historical linkage information between different table files, and determines the edge weights of the foreign key connection edges, it also implements the following steps:

[0243] For any two table files, after obtaining the historical linkage information between the two table files, construct foreign key connection edges between nodes in the target association graph corresponding to the two table files; determine the linkage frequency between the two table files based on the ratio between the number of linkage changes between the two table files and the total number of data changes in the metering database during the historical period; determine the edge weight of the foreign key connection edge based on the linkage frequency.

[0244] In one embodiment, the multidimensional features include basic features, topological features, temporal features, and linkage features; when the processor executes a computer program to extract features from the target association graph and each source data to obtain the multidimensional features of the metrology database, it also performs the following steps:

[0245] The source data are encoded to obtain basic features; a graph embedding algorithm is used to extract features from the target association graph to obtain topological features; and temporal feature analysis is performed on the source data to obtain temporal features and linkage features.

[0246] In one embodiment, when the processor executes a computer program to predict data change information of the metering database for future periods based on multidimensional features and target association graphs, it also performs the following steps:

[0247] Multidimensional features are fused to obtain fused features; attention features are determined based on the fused features and the target association graph using a graph attention network; convolution features are determined based on the attention features and the target association graph using a graph convolutional network; nonlinear transformations are applied to the convolutional features to obtain global features; and the global features are analyzed to obtain data change information of the econometric database in future time periods.

[0248] In one embodiment, the data change information includes the data change type, associated change topology information, and topology synchronization priority; when the processor executes a computer program to analyze global features and obtains the data change information of the metering database for a future period, it also performs the following steps:

[0249] Based on a preset classifier, global features are mapped to category probabilities that conform to a preset probability distribution, and the data change type is determined according to the category probabilities. The global features and historical data change information are jointly analyzed to obtain the associated change topology information. The associated change topology information includes the nodes that will change in the target association graph in the future time period. The topology synchronization priority is determined according to the importance of each node in the global features and the target association graph. The topology synchronization priority includes the priority of each node in the target association graph to be synchronized and updated.

[0250] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, performs the following steps:

[0251] Obtain the source data from each source database contained in the measurement database; wherein the source data includes at least two table files, and each table file includes at least two fields;

[0252] Based on the source data, construct the target association graph of the econometric database; the target association graph includes nodes, connecting edges between nodes, and edge weights of connecting edges; nodes represent table files or fields, and connecting edges between different nodes represent the association relationships between different table files and / or fields;

[0253] Feature extraction is performed on the target association graph and each source data to obtain multidimensional features of the econometric database;

[0254] Based on multidimensional features and target association diagrams, we predict data change information in the econometric database for future periods.

[0255] In one embodiment, when the processor executes a computer program to construct a target association graph of the metering database based on the source data, it also performs the following steps:

[0256] Extract table files and their fields from the source data; construct nodes in the target association graph based on the extracted table files and fields; construct connection edges between different nodes in the target association graph based on historical business association information between different fields and / or different table files, as well as historical linkage information and historical change association information between different table files, and determine the edge weights of the connection edges between different nodes.

[0257] In one embodiment, the connection edges include foreign key connection edges, business logic connection edges, and change-related connection edges; when the processor executes the computer program to construct the connection edges between different nodes in the target association graph based on historical business association information between different fields and / or different table files, as well as historical linkage information and historical change-related information between different table files, and determines the edge weights of the connection edges between different nodes, it also implements the following steps:

[0258] Based on historical linkage information between different table files, construct foreign key connection edges between nodes in the target association graph corresponding to different table files, and determine the edge weights of the foreign key connection edges; based on historical business association information between different fields and / or different table files, construct business logic connection edges between nodes in the target association graph corresponding to different fields and / or different table files, and determine the edge weights of the business logic connection edges; based on historical change association information between different table files, construct change association connection edges between nodes in the target association graph corresponding to different table files, and determine the edge weights of the change association connection edges.

[0259] In one embodiment, historical linkage information includes the number of linkage changes; when the processor executes the computer program to construct foreign key connection edges between nodes in the target association graph corresponding to different table files based on the historical linkage information between different table files, and determines the edge weights of the foreign key connection edges, it also implements the following steps:

[0260] For any two table files, after obtaining the historical linkage information between the two table files, construct foreign key connection edges between nodes in the target association graph corresponding to the two table files; determine the linkage frequency between the two table files based on the ratio between the number of linkage changes between the two table files and the total number of data changes in the metering database during the historical period; determine the edge weight of the foreign key connection edge based on the linkage frequency.

[0261] In one embodiment, the multidimensional features include basic features, topological features, temporal features, and linkage features; when the processor executes a computer program to extract features from the target association graph and each source data to obtain the multidimensional features of the metrology database, it also performs the following steps:

[0262] The source data are encoded to obtain basic features; a graph embedding algorithm is used to extract features from the target association graph to obtain topological features; and temporal feature analysis is performed on the source data to obtain temporal features and linkage features.

[0263] In one embodiment, when the processor executes a computer program to predict data change information of the metering database for future periods based on multidimensional features and target association graphs, it also performs the following steps:

[0264] Multidimensional features are fused to obtain fused features; attention features are determined based on the fused features and the target association graph using a graph attention network; convolution features are determined based on the attention features and the target association graph using a graph convolutional network; nonlinear transformations are applied to the convolutional features to obtain global features; and the global features are analyzed to obtain data change information of the econometric database in future time periods.

[0265] In one embodiment, the data change information includes the data change type, associated change topology information, and topology synchronization priority; when the processor executes a computer program to analyze global features and obtains the data change information of the metering database for a future period, it also performs the following steps:

[0266] Based on a preset classifier, global features are mapped to category probabilities that conform to a preset probability distribution, and the data change type is determined according to the category probabilities. The global features and historical data change information are jointly analyzed to obtain the associated change topology information. The associated change topology information includes the nodes that will change in the target association graph in the future time period. The topology synchronization priority is determined according to the importance of each node in the global features and the target association graph. The topology synchronization priority includes the priority of each node in the target association graph to be synchronized and updated.

[0267] It should be noted that the data involved in this application (including but not limited to data used for analysis, data stored, data displayed, etc.) are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0268] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0269] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0270] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A data change prediction method, characterized in that, The method includes: Obtain source data from each source database contained in the measurement database; wherein, the source data includes at least two table files, and each table file includes at least two fields; Based on the source data, a target association graph of the measurement database is constructed; wherein, the target association graph includes nodes, connecting edges between nodes, and edge weights of the connecting edges; nodes represent table files or fields, and connecting edges between different nodes represent the association relationships between different table files and / or fields; Feature extraction is performed on the target association graph and each of the source data to obtain the multidimensional features of the metrology database; Based on the multidimensional features and the target association graph, the data change information of the measurement database in future time periods is predicted.

2. The method according to claim 1, characterized in that, The step of constructing the target association graph of the measurement database based on each of the source data includes: Extract the table file and the fields contained in the table file from each of the source data; Based on the extracted table files and fields, construct the nodes in the target association graph; Based on the historical business association information between different fields and / or different table files, as well as the historical linkage information and historical change association information between different table files, the connection edges between different nodes in the target association graph are constructed, and the edge weights of the connection edges between different nodes are determined.

3. The method according to claim 2, characterized in that, The connection edges include foreign key connection edges, business logic connection edges, and change-related connection edges; The step of constructing connection edges between different nodes in the target association graph based on historical business association information between different fields and / or different table files, as well as historical linkage information and historical change association information between different table files, and determining the edge weights of the connection edges between different nodes includes: Based on the historical linkage information between different table files, foreign key connection edges are constructed between nodes in the target association graph corresponding to different table files, and the edge weights of the foreign key connection edges are determined. Based on the historical business association information between different fields and / or different table files, construct business logic connection edges between nodes in the target association graph corresponding to different fields and / or different table files, and determine the edge weights of the business logic connection edges. Based on the historical change association information between different table files, construct change association connection edges between nodes in the target association graph corresponding to different table files, and determine the edge weight of the change association connection edges.

4. The method according to claim 3, characterized in that, The historical linkage information includes the number of linkage changes; The step of constructing foreign key connection edges between nodes in the target association graph corresponding to different table files based on historical linkage information between different table files, and determining the edge weights of the foreign key connection edges, includes: For any two table files, after obtaining the historical linkage information between the two table files, construct the foreign key connection edges between the nodes in the target association graph corresponding to the two table files; The linkage frequency between the two table files is determined based on the ratio between the number of linkage changes between the two table files and the total number of data changes in the metering database during the historical period. The edge weight of the foreign key connection edge is determined based on the linkage frequency.

5. The method according to claim 1, characterized in that, The multidimensional features include basic features, topological features, temporal features, and linkage features; The step of extracting features from the target association graph and each of the source data to obtain the multidimensional features of the metrology database includes: The source data are encoded to obtain the basic features; The topological features are obtained by using a graph embedding algorithm to extract features from the target association graph. The time-series features and linkage features are obtained by performing time-series feature analysis on each of the source data.

6. The method according to claim 5, characterized in that, The step of predicting data change information of the measurement database in future time periods based on the multidimensional features and the target association graph includes: Multidimensional features are fused to obtain fused features; Based on a graph attention network, attention features are determined according to the fusion features and the target association graph; Based on the graph convolutional network, convolutional features are determined according to the attention features and the target association graph; The convolutional features are subjected to a nonlinear transformation to obtain global features; By analyzing the global features, the data change information of the metering database in future time periods can be obtained.

7. The method according to claim 6, characterized in that, The data change information includes the data change type, associated change topology information, and topology synchronization priority; The analysis of the global features to obtain the data change information of the metering database in future time periods includes: Based on a preset classifier, the global features are mapped to category probabilities that conform to a preset probability distribution, and the data change type is determined according to the category probabilities. The global features and historical data change information are jointly analyzed to obtain the association change topology information; wherein, the association change topology information includes nodes in the target association graph that have changed in the future time period; The topology synchronization priority is determined based on the global features and the node importance of each node in the target association graph; wherein, the topology synchronization priority includes the priority of each node in the target association graph for synchronization updates.

8. A data change prediction device, characterized in that, The device includes: The data acquisition module is used to acquire source data from each source database included in the measurement database; wherein, the source data includes at least two table files, and each table file includes at least two fields; The graph construction module is used to construct a target association graph of the measurement database based on the source data. The target association graph includes nodes, connecting edges between nodes, and edge weights of the connecting edges. Nodes represent table files or fields, and connecting edges between different nodes represent the association relationships between different table files and / or fields. The feature extraction module is used to extract features from the target association graph and each of the source data to obtain the multidimensional features of the metrology database; The information prediction module is used to predict data change information of the measurement database in future time periods based on the multidimensional features and the target association graph.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

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

Images