Intelligent construction method of software knowledge graph based on meta-level low-code platform

Abstract

The present invention provides a method for intelligently constructing a software knowledge graph based on a meta-level low-code platform, which relates to the technical field of knowledge graph construction, including: extracting target knowledge elements from different types of data sources, integrating them into a target meta-model, and establishing a target metadata management system; based on a preset graph database, realizing the graph construction of low-code development knowledge to obtain a target knowledge graph; monitoring the changing state of the code in the code warehouse in real time, and dynamically updating the target knowledge graph according to the changing state data. By extracting target knowledge elements and integrating them into a meta-model; based on a preset graph database, realizing the graph construction of low-code development knowledge; monitoring the changing state of the code in the code warehouse in real time, and dynamically updating the knowledge graph according to the changing state data, it can support the rapid construction and updating of software knowledge graphs, and be applied to various low-code development platforms to improve development efficiency and software quality.

Description

Technical Field

[0001] The present invention relates to the technical field of knowledge graph construction, and in particular to a method for intelligently constructing a software knowledge graph based on a meta-level low-code platform. Background Art

[0002] In the current context of digital transformation, enterprises have an increasing demand for intelligence, automation and integration. As a key technology for connecting different systems and data, software knowledge graphs have broad application prospects. However, traditional software knowledge graph construction methods often have problems such as long development cycles, high technical barriers, and high maintenance costs, which make it difficult for enterprises to quickly respond to market changes. Therefore, how to quickly build and update software knowledge graphs has become one of the current research centers.

[0003] Therefore, the present invention provides a method for intelligently constructing a software knowledge graph based on a meta-level low-code platform, which can be applied to various low-code development platforms, supports rapid construction and updating of software knowledge graphs, and improves development efficiency and software quality. Summary of the invention

[0004] The present invention provides a method for intelligently constructing a software knowledge graph based on a meta-level low-code platform, which is used to extract target knowledge elements and integrate them into a meta-model; based on a preset graph database, it realizes the graph construction of low-code development knowledge; it monitors the changing status of the code in the code warehouse in real time, and dynamically updates the knowledge graph according to the changing status data, which can support the rapid construction and updating of the software knowledge graph, and can be applied to various low-code development platforms to improve development efficiency and software quality.

[0005] The present invention provides a method for intelligently constructing a software knowledge graph based on a meta-level low-code platform, comprising:

[0006] Step 1: Extract target knowledge elements from different types of data sources, integrate them into the target meta-model, and establish a target metadata management system;

[0007] Step 2: Based on the preset graph database, realize the graph construction of low-code development knowledge and obtain the target knowledge graph;

[0008] Step 3: Monitor the code change status in the code repository in real time, and dynamically update the target knowledge graph based on the change status data.

[0009] Preferably, target knowledge elements are extracted from different types of data sources, integrated into a target meta-model, and a target metadata management system is established, including;

[0010] Utilize the intelligent knowledge extraction function of the current meta-level low-code platform to extract the first knowledge element from different types of data sources;

[0011] Monitor the extraction process of the first knowledge element in real time and respond when an anomaly is detected;

[0012] After cleaning the extracted first knowledge element data, the data is converted into the format and type supported by the target meta-model to obtain the target knowledge element;

[0013] Based on the set mapping rules, the target knowledge elements are mapped to corresponding elements and relationships in the target meta-model;

[0014] Storing the mapping relationship between the target knowledge elements and the corresponding elements and relationships in the target meta-model in the target knowledge base;

[0015] Based on data requirements, establish a target metadata management system to store the metadata of the current meta-level low-code platform.

[0016] Preferably, the data source types include structured types, semi-structured types and unstructured types.

[0017] Preferably, real-time monitoring of the extraction process of the first knowledge element and responding when an abnormality is detected include:

[0018] Using a target monitoring tool, the first knowledge element extraction process is monitored in real time to obtain first performance indicator data;

[0019] By comparing and analyzing the first performance indicator data with a set performance threshold, each first performance indicator data is calibrated in state;

[0020] Establishing an indicator state set table according to the state calibration result of the first performance indicator data;

[0021] If only normal state indicators exist in the indicator state set table, it is determined that the current knowledge element extraction process is normal;

[0022] If there is an abnormal state indicator or a possible abnormal state indicator in the indicator state set table, it is determined that there is an abnormality in the current first knowledge element extraction process;

[0023] When an exception occurs in the first knowledge element extraction process, a corresponding exception handling strategy is adopted to handle the exception.

[0024] Preferably, when an exception occurs in the first knowledge element extraction process, a corresponding exception handling strategy is adopted to handle the exception, including:

[0025] If there is an abnormal state indicator in the indicator state set table, determine the first abnormal type based on the abnormal state indicator from the setting type-exception handling mapping table, and extract a first setting exception handling rule matching the current first abnormal type;

[0026] Extracting an exception handling measure from the first set exception handling rule and using it as a first measure;

[0027] If there are possible abnormal state indicators in the indicator state set table, extract the historical performance data of the possible abnormal state indicators within a preset time period, and identify potential abnormal points by using a machine learning algorithm;

[0028] Determining a first time window according to the timestamp of the potential abnormal point;

[0029] Extracting all log entries within a first time window from a log database, and performing entry screening to obtain a first log entry;

[0030] Performing correlation analysis on the first log entry and potential abnormal points to obtain an abnormal correlation coefficient;

[0031] Mark the first log entry with an abnormal correlation coefficient greater than a set correlation threshold as the first high correlation entry;

[0032] Extracting the abnormality-related features of the first highly associated item and inputting them into a pre-established type recognition model, outputting an abnormality type recognition result corresponding to the abnormality-related features and serving as a second abnormality type;

[0033] Extracting a second set exception handling rule matching the current second exception type from the set type-exception handling mapping table;

[0034] Extracting an exception handling measure from the second set exception handling rule and treating it as a second measure;

[0035] When there is an abnormal state indicator in the indicator state set table and there is no possible abnormal state indicator, the first measure is used as an exception handling strategy to immediately handle the exception;

[0036] When there are possible abnormal state indicators in the indicator state set table, but no abnormal state indicators, all the acquired second measures are prioritized according to the set priority evaluation mechanism, and then used as the exception handling strategy to perform corresponding exception handling in sequence;

[0037] When both abnormal state indicators and possible abnormal state indicators exist in the indicator state set table, the first measure obtained is immediately executed to perform corresponding exception processing; if there is a second measure that is consistent with the first measure, the current second measure is deleted, and the remaining second measures are prioritized according to the set priority evaluation mechanism, and then the corresponding exception processing is performed in sequence; if there is no second measure that is consistent with the first measure, the second measures obtained are prioritized according to the set priority evaluation mechanism, and then the corresponding exception processing is performed in sequence.

[0038] Preferably, the calculation formula of the abnormal correlation coefficient is as follows:

[0039] ; In the formula, It is expressed as the abnormal correlation coefficient between the i-th first target item and the current potential abnormal point; It is represented as the time interval between the timestamp of the i-th first target entry and the occurrence time of the current potential anomaly point; It is represented as the text similarity between the i-th first target item and the current potential outlier description; It is represented as the total number of entities of the i-th first target item; It is represented as the number of identical entities between the i-th first target item and the current potential outlier; It is represented as the total entity importance of the same entity of the i-th first target item and the current potential outlier; It is represented as the number of similar entities between the i-th first target item and the current potential outlier; It is expressed as the similarity between the i-th first target item and the j-th similar entity of the current potential outlier; It is represented as the entity importance of the i-th first target item and the j-th similar entity of the current potential outlier.

[0040] Preferably, based on a preset graph database, the low-code development knowledge is graphed to obtain a target knowledge graph, including:

[0041] Select the set import method to import the mapped target knowledge elements and their relationships into the preset graph database;

[0042] Use the target language provided by the preset graph database to create nodes and relationships to obtain the target knowledge graph;

[0043] The target knowledge graph is visualized using a preset visualization tool.

[0044] Preferably, the change status of the code in the code repository is monitored in real time, and the target knowledge graph is dynamically updated according to the change status data, including:

[0045] From the monitoring methods, select the target monitoring method based on the analysis of the screening requirement indicators to monitor the changes in the code repository and obtain the code change content;

[0046] Extracting a first file associated with the target knowledge graph from the code change content using a set matching rule;

[0047] The first file is parsed using a set text processing method to obtain target change information;

[0048] Determine a change type according to the target change information;

[0049] Based on the change type, corresponding changes are made to the nodes, relationships or attributes in the target knowledge graph to obtain a changed knowledge graph;

[0050] Indexes are created for the attributes of the frequently queried nodes and the attributes of the frequently queried relationships in the changed knowledge graph.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] By extracting target knowledge elements and integrating them into the metamodel; based on the preset graph database, the graph construction of low-code development knowledge is realized; the changing status of the code in the code repository is monitored in real time, and the knowledge graph is dynamically updated according to the changing status data. It can support the rapid construction and updating of software knowledge graphs, which can be applied to various low-code development platforms and improve development efficiency and software quality.

[0053] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description and the accompanying drawings.

[0054] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0056] Figure 1 This is a flowchart of a method for intelligently constructing a software knowledge graph based on a meta-level low-code platform in an embodiment of the present invention. DETAILED DESCRIPTION

[0057] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0058] The embodiment of the present invention provides a method for intelligently constructing a software knowledge graph based on a meta-level low-code platform, such as Figure 1 As shown, including:

[0059] Step 1: Extract target knowledge elements from different types of data sources, integrate them into the target meta-model, and establish a metadata management system;

[0060] Step 2: Based on the preset graph database, realize the graph construction of low-code development knowledge and obtain the target knowledge graph;

[0061] Step 3: Monitor the code change status in the code repository in real time, and dynamically update the target knowledge graph based on the change status data.

[0062] In this embodiment, the low-code platform refers to a development platform that allows users to quickly build applications through a graphical interface and programming code; data source types include structured, semi-structured and unstructured; target knowledge elements include entity, relationship and attribute knowledge elements, among which entity types include objects, concepts and events, etc.; relationship types include association, inheritance, dependency and aggregation, etc.; attribute types include basic attributes, associated attributes, characteristic attributes and dynamic attributes, among which basic attributes refer to the basic information of an entity, such as entity name and identifier; associated attributes are used to describe the associated attributes of an entity with other entities, such as the category to which an entity belongs; characteristic attributes are used to describe the attributes of an entity's unique characteristics, such as entity size; dynamic attributes refer to attributes of an entity that change over time, such as status.

[0063] In this embodiment, the target metamodel refers to a pre-established model that covers core elements such as scenario knowledge, requirement knowledge, and template knowledge, and clarifies the structure and relationship between the core elements. Scenario knowledge refers to a knowledge set under a specific business scenario, which is used to describe how the application should be built, configured and used in the corresponding scenario, including a description of the scenario, the business processes involved, the required data model, etc.; requirement knowledge refers to the specific requirements of users or business parties for the application, including functional requirements, performance requirements, interface requirements, etc.; template knowledge refers to the predefined reusable components, modules or frameworks in the low-code platform, including the type, structure, attributes of the template and its association with other elements, etc.; the structure between core elements refers to the hierarchical structure, combination rules and constraints of the core elements, and the relationships include association, inheritance, dependency and aggregation, etc.

[0064] In this embodiment, the preset graph database is a predetermined graph database for constructing a knowledge graph, such as Neo4J, wherein the graph database refers to a data management system that uses points and edges as basic storage units and is designed with efficient storage and query of graph data as the principle, and can be applied to the field of knowledge graphs, including Neo4J, JanusGraph, Galaxybase, TuGraph, and the like.

[0065] In this embodiment, the target metadata management system is used to store and manage various metadata in the meta-level low-code platform, and is composed of a data storage module, a data management module and a data utilization module. The data storage module is used to define a unified metadata format and standard and store metadata; the data management module is used to implement data entry, data verification, version control and permission management functions; the data utilization module is used to implement data query and analysis, data sharing and data visualization functions; the code warehouse refers to a centralized or distributed storage space used to store, manage, track and control code changes.

[0066] The beneficial effects of the above technical solution are: by extracting the target knowledge elements and integrating them into the metamodel; based on the preset graph database, realizing the graph construction of low-code development knowledge; real-time monitoring of the changing status of the code in the code repository, and dynamically updating the knowledge graph according to the changing status data, it can support the rapid construction and updating of software knowledge graphs, be applied to various low-code development platforms, and improve development efficiency and software quality.

[0067] The embodiment of the present invention provides a method for intelligently constructing a software knowledge graph based on a meta-level low-code platform, extracting target knowledge elements from different types of data sources, integrating them into a target meta-model, and establishing a target metadata management system, including:

[0068] Utilize the intelligent knowledge extraction function of the current meta-level low-code platform to extract the first knowledge element from different types of data sources;

[0069] Monitor the extraction process of the first knowledge element in real time and respond when an anomaly is detected;

[0070] After cleaning the extracted first knowledge element data, the data is converted into the format and type supported by the target meta-model to obtain the target knowledge element;

[0071] Based on the set mapping rules, the target knowledge elements are mapped to corresponding elements and relationships in the target metamodel;

[0072] Storing the mapping relationship between the target knowledge elements and the corresponding elements and relationships in the target meta-model in the target knowledge base;

[0073] Based on data requirements, establish a target metadata management system to store the metadata of the current meta-level low-code platform.

[0074] In this embodiment, the low-code platform refers to a development platform that allows users to quickly build applications through a graphical interface and programming code; the intelligent knowledge extraction function refers to extracting entity, relationship, and attribute knowledge elements from different types of data sources; data source types include structured, semi-structured, and unstructured; target knowledge elements include entity, relationship, and attribute knowledge elements; the target metamodel refers to a pre-established model that covers core elements such as scenario knowledge, requirement knowledge, and template knowledge and clarifies the structure and relationship between core elements, among which scenario knowledge refers to a knowledge set under a specific business scenario, which is used to describe how the application should be built, configured, and used in the corresponding scenario, including a description of the scenario, The business processes involved, the required data models, etc.; requirement knowledge refers to the specific requirements of users or business parties for the application, including functional requirements, performance requirements, interface requirements, etc.; template knowledge refers to the predefined reusable components, modules or frameworks in the low-code platform, including the template type, structure, attributes and association with other elements, etc.; the structure between core elements refers to the hierarchical structure, combination rules and constraints of the core elements, and the relationships include association, inheritance, dependency and aggregation, etc.; mapping the target knowledge elements to the corresponding elements and relationships in the target metamodel is achieved through the mapping tools of the low-code platform; data cleaning includes removing duplicate data, correcting erroneous data and processing missing values.

[0075] In this embodiment, the target knowledge base is pre-established and is used to store and manage the mapped target knowledge elements and their relationships, provide query, update, deletion and other operations on the target knowledge base, and support data version control and permission management; data requirements refer to operational requirements for metadata, such as storage, verification, query and visualization, etc.; the target metadata management system is used to store and manage various metadata in the meta-level low-code platform, and is composed of a data storage module, a data management module and a data utilization module, wherein the data storage module is used to define a unified metadata format and standard and store metadata; the data management module is used to implement data entry, data verification, version control and permission management functions; the data utilization module is used to implement query and analysis, data sharing and data visualization functions.

[0076] The beneficial effect of the above technical solution is: by utilizing the intelligent knowledge extraction function of the meta-level low-code platform to extract knowledge elements from different types of data sources and perform processing and mapping, as well as establishing a target metadata management system, it can significantly improve data integration efficiency and accuracy and reduce development and maintenance costs.

[0077] The embodiment of the present invention provides a method for intelligently constructing a software knowledge graph based on a meta-level low-code platform, which monitors the extraction process of the first knowledge element in real time and responds when an abnormality is detected, including:

[0078] Using a target monitoring tool, the first knowledge element extraction process is monitored in real time to obtain first performance indicator data;

[0079] By comparing and analyzing the first performance indicator data with a set performance threshold, each first performance indicator data is calibrated in state;

[0080] Establishing an indicator state set table according to the state calibration result of the first performance indicator data;

[0081] If only normal state indicators exist in the indicator state set table, it is determined that the current knowledge element extraction process is normal;

[0082] If there is an abnormal state indicator or a possible abnormal state indicator in the indicator state set table, it is determined that there is an abnormality in the current first knowledge element extraction process;

[0083] When an exception occurs in the first knowledge element extraction process, a corresponding exception handling strategy is adopted to handle the exception.

[0084] In this embodiment, the target monitoring tool refers to a monitoring tool used to monitor the first knowledge element extraction process in real time, such as Prometheus open source monitoring; the first performance indicator data includes response time, throughput, resource utilization, error rate, etc.

[0085] In this embodiment, the steps for establishing the indicator status set table are as follows:

[0086] (1) comparing the first performance indicator data with a set performance threshold;

[0087] (2) If the first performance indicator data is not greater than the set performance threshold and does not exceed the corresponding preset indicator range, the current first performance indicator data is marked as a normal state indicator;

[0088] (3) If there is a first performance indicator data that is not greater than the set performance threshold and exceeds the corresponding preset indicator range, the current first performance indicator data is marked as a possible abnormal state indicator;

[0089] (4) If there is a first performance indicator data that is greater than the set performance threshold, the current first performance indicator data is marked as an abnormal state indicator;

[0090] (5) The first performance indicator data after the collection mark state is obtained to obtain the indicator state collection table.

[0091] In this embodiment, the set performance threshold refers to the value obtained by adding the average value and the standard deviation calculated for the historical performance data within a period of time; the preset indicator range is composed of a preset indicator upper limit and a preset indicator lower limit, wherein the preset indicator upper limit refers to a value that accounts for 95% of the set performance threshold, and the preset indicator lower limit refers to a value that accounts for 15% of the set performance threshold; the state calibration results include normal, abnormal and possibly abnormal; the indicator state set table is composed of normal state indicators, abnormal state indicators or possibly abnormal state indicators.

[0092] In this embodiment, for example, there is a response time e1 of 60s, a set performance threshold value E0 of the response time of 120s, and a preset difference range of the response time of ,Right now , where s is the time unit, expressed in seconds;

[0093] If the response time e1 is less than the set performance threshold E0 and the response time e1 does not exceed the preset indicator range, the current response time e1 is marked as a normal state indicator.

[0094] The beneficial effects of the above technical solution are: by using target monitoring tools to monitor the knowledge element process in real time, and managing performance status and potential problems through threshold comparison, judgment and response mechanisms, it is possible to improve resource utilization efficiency, shorten problem response time, and improve processing efficiency.

[0095] The embodiment of the present invention provides a method for intelligently constructing a software knowledge graph based on a meta-level low-code platform. When an exception occurs in the first knowledge element extraction process, a corresponding exception handling strategy is adopted to handle the exception, including:

[0096] If there is an abnormal state indicator in the indicator state set table, determine the first abnormal type based on the abnormal state indicator from the setting type-exception handling mapping table, and extract a first setting exception handling rule matching the current first abnormal type;

[0097] Extracting an exception handling measure from the first set exception handling rule and using it as a first measure;

[0098] If there are possible abnormal state indicators in the indicator state set table, extract the historical performance data of the possible abnormal state indicators within a preset time period, and identify potential abnormal points by using a machine learning algorithm;

[0099] Determining a first time window according to the timestamp of the potential abnormal point;

[0100] Extracting all log entries within a first time window from a log database, and performing entry screening to obtain a first log entry;

[0101] Performing correlation analysis on the first log entry and potential abnormal points to obtain an abnormal correlation coefficient;

[0102] Mark the first log entry with an abnormal correlation coefficient greater than a set correlation threshold as the first high correlation entry;

[0103] Extracting the abnormality-related features of the first highly associated item and inputting them into a pre-established type recognition model, outputting an abnormality type recognition result corresponding to the abnormality-related features and serving as a second abnormality type;

[0104] Extracting a second set exception handling rule matching the current second exception type from the set type-exception handling mapping table;

[0105] Extracting an exception handling measure from the second set exception handling rule and treating it as a second measure;

[0106] When there is an abnormal state indicator and no possible abnormal state indicator in the indicator state set table, the first measure is used as an exception handling strategy to immediately handle the exception;

[0107] When there is an abnormal state indicator in the indicator state set table and there is no possible abnormal state indicator, the first measure is used as an exception handling strategy to immediately handle the exception;

[0108] When there are possible abnormal state indicators in the indicator state set table, but no abnormal state indicators, all the acquired second measures are prioritized according to the set priority evaluation mechanism, and then used as the exception handling strategy to perform corresponding exception handling in sequence;

[0109] When both abnormal state indicators and possible abnormal state indicators exist in the indicator state set table, the first measure obtained is immediately executed to perform corresponding exception processing; if there is a second measure that is consistent with the first measure, the current second measure is deleted, and the remaining second measures are prioritized according to the set priority evaluation mechanism, and then the corresponding exception processing is performed in sequence; if there is no second measure that is consistent with the first measure, the second measures obtained are prioritized according to the set priority evaluation mechanism, and then the corresponding exception processing is performed in sequence.

[0110] In this embodiment, a setting type-exception handling mapping table is pre-established based on the exception data in the historical knowledge extraction process and the corresponding successful processing strategy, and is composed of exception type, exception type description, performance indicators and set exception handling rules, wherein the exception types include data exceptions, performance exceptions, system errors, and logical errors; the setting exception handling rules are composed of exception identification types and exception handling measures, wherein the exception handling measures include data rollback, resource allocation adjustment, extraction rule optimization, mapping logic optimization, logging, personnel notification, etc.

[0111] In this embodiment, the first exception type is a type screened out from a setting type-exception handling mapping table according to an abnormal state indicator; the first setting exception handling rule refers to a setting exception handling rule that matches the first exception type; and the first measure refers to an exception handling measure extracted from the first setting exception handling rule.

[0112] In this embodiment, the preset time period is pre-set; the potential abnormal point is the recognition result output after inputting the historical performance data into the abnormal recognition model established in advance based on the machine learning algorithm, wherein the abnormal recognition model is a model obtained by training the isolation forest model with the historical data related to the performance index acquired in advance; the first time window is a time period determined based on the timestamp of the potential abnormal point, for example, there is a timestamp of the potential abnormal point x1, that is, the time of occurrence is T0, and the start time of the first time window is , and the end time is , where 0.5 It is expressed as half an hour, h is the time unit, expressed as hours.

[0113] In this embodiment, the log database is composed of log data of various components such as the system, application, and database; a log entry refers to all entries within a first time window filtered out from the log database, and an entry refers to record information automatically generated during the operation of a computer system, which usually includes the time when a specific event occurred, the type of event, related components or modules, error codes, detailed description information, etc.; a first log entry refers to an entry filtered out from the acquired target entries by using specific keywords as selection conditions, wherein the keywords include component names related to performance indicators, warning information, etc.

[0114] In this embodiment, the anomaly correlation coefficient is used to quantify the degree of correlation between log entries and potential anomaly points; the set correlation threshold refers to the average value and the average value of the correlation coefficients of a preset number of different historical log entries and known anomaly points; the first high correlation entry refers to the first log entry whose anomaly correlation coefficient is greater than the set correlation threshold; anomaly-related features include error codes, anomaly information, and changes in performance indicators related to the anomaly, such as extended response time, increased resource consumption, and specific behaviors of system components.

[0115] In this embodiment, the type recognition model refers to a model obtained by training a neural network using historical abnormal features extracted from a preset amount of normal log entries and abnormal log entries as training data, and is used to identify the abnormal type; the second abnormal type refers to the abnormal type recognition result output by inputting the abnormal related features of the first highly associated entry into the type recognition model; the second set abnormal handling rule refers to a preset set abnormal handling rule that matches the second abnormal type in the set type-abnormal handling mapping table; the second measure refers to the abnormal handling measure extracted from the second set abnormal handling rule; setting a priority evaluation mechanism refers to obtaining the priority coefficient of the second measure and prioritizing the second measure from large to small, wherein the priority coefficient refers to the value obtained by multiplying the absolute difference between the corresponding possible state indicator of the current second measure and the set performance threshold by the indicator importance weight of the corresponding possible state indicator, wherein the indicator importance weight is a weight value calculated by using the hierarchical analysis method with the knowledge element extraction speed as the target, and determining the relative importance of the indicator by pairwise comparison.

[0116] The beneficial effect of the above technical solution is: by matching the corresponding exception handling strategy to handle the exception when there is an exception in the knowledge element extraction process, the efficiency of knowledge graph construction can be improved and the accuracy of knowledge graph construction can be ensured, thereby enhancing the reliability of the knowledge graph.

[0117] The embodiment of the present invention provides a method for intelligently constructing a software knowledge graph based on a meta-level low-code platform. The calculation formula of the abnormal correlation coefficient is as follows:

[0118] ; In the formula, It is expressed as the abnormal correlation coefficient between the i-th first target item and the current potential abnormal point; It is represented as the time interval between the timestamp of the i-th first target entry and the occurrence time of the current potential anomaly point; It is represented as the text similarity between the i-th first target item and the current potential outlier description; It is represented as the total number of entities of the i-th first target item; It is represented as the number of identical entities between the i-th first target item and the current potential outlier; It is represented as the total entity importance of the same entity of the i-th first target item and the current potential outlier; It is represented as the number of similar entities between the i-th first target item and the current potential outlier; It is expressed as the similarity between the i-th first target item and the j-th similar entity of the current potential outlier; It is represented as the entity importance of the i-th first target item and the j-th similar entity of the current potential outlier.

[0119] In this embodiment, the anomaly correlation coefficient is used to quantify the degree of correlation between the log entry and the potential anomaly point; the timestamp of the first target entry refers to the time point when the first target entry is generated; the potential anomaly point description refers to keywords and error codes; the text similarity is a numerical value calculated by using the TF-IDF algorithm to quantify the text similarity between the first target entry and the current potential anomaly point description; the entity refers to the service name, file name, IP address, etc.; the entity importance refers to a numerical value that quantifies the importance of the entity, which is determined by calculating the frequency of the entity appearing in all first target entries; the similarity of similar entities is calculated by using the cosine similarity algorithm.

[0120] The beneficial effect of the above technical solution is: by calculating the anomaly correlation, log entries with high correlation with potential anomalies are selected, providing data support for identifying the current anomaly type, which in turn helps to adopt accurate exception handling strategies for exception handling and improve the efficiency of knowledge element extraction.

[0121] The embodiment of the present invention provides a method for intelligently constructing a software knowledge graph based on a meta-level low-code platform, which realizes the graph construction of low-code development knowledge based on a preset graph database to obtain a target knowledge graph, including:

[0122] Select the set import method to import the mapped target knowledge elements and their relationships into the preset graph database;

[0123] Use the target language provided by the preset graph database to create nodes and relationships to obtain the target knowledge graph;

[0124] The target knowledge graph is visualized using a preset visualization tool.

[0125] In this embodiment, an import method is set to import the mapped target knowledge elements and their relationships into a preset graph database. For example, when the preset graph database is a Neo4j graph database, the corresponding import method for importing large-scale data is the Neo4j import tool, and the corresponding import method for importing small-scale or incremental data is the Cypher language, wherein the data scale is determined by the amount of imported data. When the amount of imported data exceeds 300,000 records, or the size of the imported data file exceeds 500 megabytes, it is determined to be large-scale; when the amount of imported data does not exceed 300,000 records, or the size of the imported data file does not exceed 500 megabytes, it is determined to be small-scale.

[0126] In this embodiment, the preset graph database is a predetermined graph database for knowledge graph construction, such as Neo4J. The Neo4J graph database refers to a high-performance NoSQL graph database that stores data with the characteristics of a graph model; a graph database refers to a data management system that uses points and edges as basic storage units and is designed with efficient storage and query of graph data as the principle, and can be applied to the field of knowledge graphs, including Neo4J, JanusGraph, Galaxybase, TuGraph, etc.; the target language refers to a language used in the preset graph database to create, change, and delete nodes and relationships, such as the Cypher language in the Neo4J graph database; the target knowledge graph refers to a low-code developed knowledge graph that is gradually constructed by creating and connecting nodes in the preset graph database; the preset visualization tool is used to visualize the target knowledge graph. When the preset graph database has a built-in visualization tool, the corresponding built-in visualization tool is regarded as the preset visualization tool, such as the Neo4J Browser in the Neo4J graph database. When the preset graph database does not have a built-in visualization tool, a third-party visualization tool compatible with the current preset graph database is used as the preset visualization tool to visualize the target knowledge graph.

[0127] The beneficial effect of the above technical solution is: by importing the mapped target knowledge elements and their relationships into a preset graph database to form a knowledge graph and visualize it, it can support the rapid construction of a software knowledge graph, thereby improving development efficiency and software quality.

[0128] The embodiment of the present invention provides a method for intelligently constructing a software knowledge graph based on a meta-level low-code platform, which monitors the change status of codes in a code repository in real time and dynamically updates the target knowledge graph according to the change status data, including:

[0129] From the monitoring methods, select the target monitoring method based on the analysis of the screening requirement indicators to monitor the changes in the code repository and obtain the code change content;

[0130] Extracting a first file associated with the target knowledge graph from the code change content using a set matching rule;

[0131] The first file is parsed using a set text processing method to obtain target change information;

[0132] Determine a change type according to the target change information;

[0133] Based on the change type, corresponding changes are made to the nodes, relationships or attributes in the target knowledge graph to obtain a changed knowledge graph;

[0134] Indexes are created for the attributes of the frequently queried nodes and the attributes of the frequently queried relationships in the changed knowledge graph.

[0135] In this embodiment, the monitoring method is used to monitor changes in the code repository, including Webhooks, timed polling and other methods, wherein the code repository refers to a centralized or distributed storage space used to store, manage, track and control code changes; the target monitoring method is a monitoring method with the highest monitoring adaptation coefficient selected from the monitoring methods that support the current code repository, wherein the monitoring adaptation coefficient refers to the coefficient obtained by averaging the indicator evaluation values ​​obtained by evaluating the current code repository monitoring based on the screening requirement indicators; the screening requirement indicators include real-time requirements, resource consumption requirements and security requirements; setting matching rules refers to text path pattern matching; code change content refers to the list and content of newly added, modified or deleted files; change status data refers to the code change content.

[0136] In this embodiment, setting matching rules means taking the keywords in the file name and the file content as matching conditions and traversing the acquired code change content, wherein the keywords refer to the business logic keywords of the software, including class names, method names, variable names, etc.; the first file refers to the file associated with the target knowledge graph extracted from the code change content using the set matching rules; the set text processing method is predetermined, including regular expressions and corresponding parsing libraries in special formats, wherein the special formats include Java, JSON, XML, etc.; the target change information is the information obtained by parsing the content of the first file, including newly added functions, modified fields, class names, etc.; the change types include node addition, relationship modification, and attribute update, etc.; the changed knowledge graph refers to the information obtained after the nodes, relationships, or attributes in the target knowledge graph are changed accordingly according to the change type, wherein a node refers to an entity or a concept; a relationship refers to the relationship between different entities, such as an inheritance relationship, a dependency relationship, a causal relationship, etc.; an attribute refers to a specific feature that describes an entity, such as an entity's name, identifier, category, etc.; a frequently queried node refers to a node whose query times exceed the set query times; a frequently queried relationship refers to a relationship whose query times exceed the set query times; the set query times refer to the query peak value of node attributes or relationship attributes within a preset time period.

[0137] The beneficial effect of the above technical solution is: by real-time monitoring of the change status of the code in the code repository and analyzing the change status data, the dynamic update of the target knowledge graph is effectively realized, thereby ensuring the timeliness of the knowledge graph.

[0138] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. Intelligent construction method of software knowledge graph based on meta-level low-code platform, It is characterized in that include: Step 1: Extract target knowledge elements from different types of data sources, integrate them into the target meta-model, and establish a target metadata management system; Step 2: Based on the preset graph database, realize the graph construction of low-code development knowledge and obtain the target knowledge graph; Step 3: Monitor the change status of the code in the code repository in real time, and dynamically update the target knowledge graph based on the change status data; Among them, target knowledge elements are extracted from different types of data sources, integrated into the target meta-model, and a target metadata management system is established, including: Utilize the intelligent knowledge extraction function of the current meta-level low-code platform to extract the first knowledge element from different types of data sources; Monitor the extraction process of the first knowledge element in real time and respond when an anomaly is detected; After cleaning the extracted first knowledge element data, the data is converted into the format and type supported by the target meta-model to obtain the target knowledge element; Based on the set mapping rules, the target knowledge elements are mapped to corresponding elements and relationships in the target metamodel; Storing the mapping relationship between the target knowledge elements and the corresponding elements and relationships in the target meta-model in the target knowledge base; Based on data requirements, establish a target metadata management system to store the metadata of the current meta-level low-code platform; The extraction process of the first knowledge element is monitored in real time, and a response is made when an abnormality is detected, including: Using a target monitoring tool, the first knowledge element extraction process is monitored in real time to obtain first performance indicator data; By comparing and analyzing the first performance indicator data with a set performance threshold, each first performance indicator data is calibrated in state; Establishing an indicator state set table according to the state calibration result of the first performance indicator data; If only normal state indicators exist in the indicator state set table, it is determined that the current knowledge element extraction process is normal; If there is an abnormal state indicator or a possible abnormal state indicator in the indicator state set table, it is determined that there is an abnormality in the current first knowledge element extraction process; When there is an exception in the first knowledge element extraction process, a corresponding exception handling strategy is adopted to handle the exception; Among them, when there is an exception in the first knowledge element extraction process, a corresponding exception handling strategy is adopted to handle the exception, including: If there is an abnormal state indicator in the indicator state set table, determine the first abnormal type based on the abnormal state indicator from the setting type-exception handling mapping table, and extract a first setting exception handling rule matching the current first abnormal type; Extracting an exception handling measure from the first set exception handling rule and using it as a first measure; If there are possible abnormal state indicators in the indicator state set table, extract the historical performance data of the possible abnormal state indicators within a preset time period, and identify potential abnormal points by using a machine learning algorithm; Determining a first time window according to the timestamp of the potential abnormal point; Extracting all log entries within a first time window from a log database, and performing entry screening to obtain a first log entry; Performing correlation analysis on the first log entry and potential abnormal points to obtain an abnormal correlation coefficient; Mark the first log entry with an abnormal correlation coefficient greater than a set correlation threshold as the first high correlation entry; Extracting the abnormality-related features of the first highly associated item and inputting them into a pre-established type recognition model, outputting an abnormality type recognition result corresponding to the abnormality-related features and serving as a second abnormality type; Extracting a second set exception handling rule matching the current second exception type from the set type-exception handling mapping table; Extracting an exception handling measure from the second set exception handling rule and treating it as a second measure; When there is an abnormal state indicator in the indicator state set table and there is no possible abnormal state indicator, the first measure is used as an exception handling strategy to immediately handle the exception; When there are possible abnormal state indicators in the indicator state set table, but no abnormal state indicators, all the acquired second measures are prioritized according to the set priority evaluation mechanism, and then used as the exception handling strategy to perform corresponding exception handling in sequence; When both abnormal state indicators and possible abnormal state indicators exist in the indicator state set table, the first measure obtained is immediately executed to perform corresponding abnormal processing; if there is a second measure that is consistent with the first measure, the current second measure is deleted, and the remaining second measures are prioritized according to the set priority evaluation mechanism, and then the corresponding abnormal processing is performed in sequence; if there is no second measure that is consistent with the first measure, the second measures obtained are prioritized according to the set priority evaluation mechanism, and then the corresponding abnormal processing is performed in sequence; The calculation formula of the abnormal correlation coefficient is as follows: ; In the formula, It is expressed as the abnormal correlation coefficient between the i-th first target item and the current potential abnormal point; It is represented as the time interval between the timestamp of the i-th first target entry and the occurrence time of the current potential anomaly point; It is represented as the text similarity between the i-th first target item and the current potential outlier description; It is represented as the total number of entities of the i-th first target item; It is represented as the number of identical entities between the i-th first target item and the current potential outlier; It is represented as the total entity importance of the same entity of the i-th first target item and the current potential outlier; It is represented as the number of similar entities between the i-th first target item and the current potential outlier; It is expressed as the similarity between the i-th first target item and the j-th similar entity of the current potential outlier; It is represented as the entity importance of the i-th first target item and the j-th similar entity of the current potential outlier.

2. According to the method for intelligently constructing a software knowledge graph based on a meta-level low-code platform according to claim 1, It is characterized in that Data source types include structured type, semi-structured type and unstructured type.

3. According to the method for intelligently constructing a software knowledge graph based on a meta-level low-code platform according to claim 1, It is characterized in that Based on the preset graph database, the low-code development knowledge is graphed and constructed to obtain the target knowledge graph, including: Select the set import method to import the mapped target knowledge elements and their relationships into the preset graph database; Use the target language provided by the preset graph database to create nodes and relationships to obtain the target knowledge graph; The target knowledge graph is visualized using a preset visualization tool.

4. According to the method for intelligently constructing a software knowledge graph based on a meta-level low-code platform according to claim 1, It is characterized in that Monitor the change status of the code in the code repository in real time, and dynamically update the target knowledge graph based on the change status data, including: From the monitoring methods, select the target monitoring method based on the analysis of the screening requirement indicators to monitor the changes in the code repository and obtain the code change content; Extracting a first file associated with the target knowledge graph from the code change content using a set matching rule; The first file is parsed using a set text processing method to obtain target change information; Determine a change type according to the target change information; Based on the change type, corresponding changes are made to the nodes, relationships or attributes in the target knowledge graph to obtain a changed knowledge graph; Indexes are created for the attributes of the frequently queried nodes and the attributes of the frequently queried relationships in the changed knowledge graph.

Images