Method, device and storage medium for locating modeling errors

Abstract

The present disclosure provides a method, device and storage medium for locating modeling errors, wherein the method includes: setting a model operating environment, adding components required to establish a target model in the model operating environment; establishing a model flow chart corresponding to the target model based on the components, configuring parameters of the components to establish the target model; running the target model, obtaining first log information of the target model, and extracting error query information from the first log information; inputting the error query information into a preset knowledge graph, using the knowledge graph to determine whether an error occurs in the target model and obtaining corresponding error location information. The method, device and storage medium disclosed herein generate a knowledge graph based on log information, can infer the source of the error, narrow the scope of error location, find the root cause node that causes the error problem, improve the efficiency of problem discovery and diagnosis, and can significantly accelerate model development efficiency.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a method, device and storage medium for locating modeling errors. Background Art

[0002] When creating a model project on a machine learning platform, you first need to create an experiment. Then, drag components from the component palette to the experiment based on the experiment's requirements. After connecting the components and configuring their parameters, you can run the experiment. Currently, after each experiment is created, there's no way to quickly locate error nodes and provide correction suggestions. If an error occurs, it's difficult to troubleshoot, quickly identify the problem, and resolve it. Therefore, a technical solution for locating modeling errors is needed. Summary of the Invention

[0003] In view of this, a technical problem to be solved by the present invention is to provide a method, device and storage medium for locating modeling errors.

[0004] According to one aspect of the present disclosure, a method for locating modeling errors is provided, comprising: setting a model operating environment, and adding components required for establishing a target model in the model operating environment; establishing a model flow chart corresponding to the target model based on the components, and performing parameter configuration on the components to establish the target model; running the target model, obtaining first log information of the target model, and extracting error query information from the first log information; inputting the error query information into a preset knowledge graph, and using the knowledge graph to determine whether an error occurs in the target model and obtain corresponding error location information.

[0005] Optionally, before extracting error query information from the first log information, a step of constructing the knowledge graph is also included, the step including: obtaining second log information corresponding to the target model, and extracting error information from the second log information; wherein, the error information includes: exception information, alarm information; obtaining logical relationships and causal relationships corresponding to the error information; obtaining knowledge graph triples based on the logical relationships and causal relationships and the error information; wherein, the knowledge graph triples include: a first entity, a second entity, and relationship information between the first entity and the second entity; and constructing the knowledge graph based on the knowledge graph triples.

[0006] Optionally, the first log information and the second log information include: log files corresponding to the nodes of the model flow chart.

[0007] Optionally, according to the logical relationship and causal relationship and the error information, obtaining the knowledge graph triple includes: extracting the first entity and the relationship information from the error information; wherein, the first entity includes: the first node in the model flowchart, error entity information; based on the logical relationship and causal relationship, obtaining the error source information corresponding to the error information, and generating the second entity according to the error source information; wherein, the second entity includes: error location information; the error location information includes: the second node in the model flowchart, error cause information.

[0008] Optionally, the first entity, the relationship information and the second entity are normalized; and the knowledge graph triple is generated based on the normalized first entity, the relationship information and the second entity.

[0009] Optionally, the error query information is input into a preset knowledge graph, and the knowledge graph is used to determine whether an error occurs in the operation of the target model and obtain corresponding error location information, including: normalizing the error query information; generating a query statement with a triple structure based on the error query information after the normalization; inputting the query statement into the knowledge graph for query to obtain the second entity.

[0010] Optionally, the second entity is input into a preset expert system to obtain an error elimination solution corresponding to the error query information output by the expert system.

[0011] Optionally, the configuration information of the component is modified based on the error elimination solution, and a new target model is generated and run.

[0012] According to another aspect of the present disclosure, there is provided an apparatus for locating modeling errors, comprising: a component processing module for setting a model operating environment and adding components required for establishing a target model in the model operating environment; a model establishment module for establishing a model flow chart corresponding to the target model based on the components, and performing parameter configuration on the components to establish the target model; an information acquisition module for running the target model, obtaining first log information of the target model, and extracting error query information from the first log information; and an error location module for inputting the error query information into a preset knowledge graph, using the knowledge graph to determine whether an error occurs in the target model and obtaining corresponding error location information.

[0013] Optionally, a graph generation module is used to construct the knowledge graph before extracting error query information from the first log information, including: an error processing unit, used to obtain second log information corresponding to the target model, and extract error information from the second log information; wherein, the error information includes: exception information, alarm information; a graph processing unit, used to obtain logical relationships and causal relationships corresponding to the error information, and obtain a knowledge graph triple based on the logical relationships and causal relationships and the error information; wherein, the knowledge graph triple includes: a first entity, a second entity, and relationship information between the first entity and the second entity; and the knowledge graph is constructed based on the knowledge graph triple.

[0014] Optionally, the first log information and the second log information include: log files corresponding to the nodes of the model flow chart.

[0015] Optionally, the graph processing unit is specifically used to extract the first entity and the relationship information from the error information; wherein, the first entity includes: the first node and error entity information in the model flowchart; based on the logical relationship and causal relationship, the error source information corresponding to the error information is obtained, and the second entity is generated according to the error source information; wherein, the second entity includes: error location information; the error location information includes: the second node and error cause information in the model flowchart.

[0016] Optionally, the graph processing unit is further used to normalize the first entity, the relationship information and the second entity; and generate the knowledge graph triple based on the normalized first entity, the relationship information and the second entity.

[0017] Optionally, the error location module is used to normalize the error query information; generate a query statement with a triple structure based on the error query information after the normalization; input the query statement into the knowledge graph for query to obtain the second entity.

[0018] Optionally, the elimination solution acquisition module is configured to input the second entity into a preset expert system, and obtain an error elimination solution output by the expert system that corresponds to the error query information.

[0019] Optionally, a model updating module is used to modify the configuration information of the component based on the error elimination solution, and generate and run a new target model.

[0020] According to yet another aspect of the present disclosure, an apparatus for locating modeling errors is provided, comprising: a memory; and a processor coupled to the memory, wherein the processor is configured to execute the method described above based on instructions stored in the memory.

[0021] According to yet another aspect of the present disclosure, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the instructions are executed by a processor to perform the above method.

[0022] The method, device, and storage medium disclosed herein for locating modeling errors generate a knowledge graph based on log information, and send alarm information, operation logs, etc. of model operation to the log graph in real time. This can infer the source of the error, narrow the scope of error location, find the root cause node that causes the error problem, improve the efficiency of problem discovery and diagnosis, and significantly accelerate model development efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the embodiments of the present disclosure or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0024] Figure 1 1 is a flow chart of an embodiment of a method for locating modeling errors according to the present disclosure;

[0025] Figure 2 A schematic diagram of a process for constructing a knowledge graph in one embodiment of a method for locating modeling errors according to the present disclosure;

[0026] Figure 3 Schematic diagram of a process for obtaining knowledge graph triples in one embodiment of a method for locating modeling errors according to the present disclosure;

[0027] Figure 4A 1 is a flow chart of obtaining error location information in one embodiment of the method for locating modeling errors according to the present disclosure; Figure 4B A schematic diagram of an example of the model's operating environment;

[0028] Figure 5 Schematic diagram of modules of an embodiment of an apparatus for locating modeling errors according to the present disclosure;

[0029] Figure 6 1 is a schematic diagram of modules of another embodiment of an apparatus for locating modeling errors according to the present disclosure;

[0030] Figure 7 This is a module schematic diagram of a graph generation module in another embodiment of the apparatus for locating modeling errors according to the present disclosure;

[0031] Figure 8 This is a schematic diagram of modules of another embodiment of an apparatus for locating modeling errors according to the present disclosure. DETAILED DESCRIPTION

[0032] The present disclosure is described more fully below with reference to the accompanying drawings, which illustrate exemplary embodiments of the present disclosure. The technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only part of the embodiments of the present disclosure, not all of the embodiments. Based on the embodiments of the present disclosure, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present disclosure.

[0033] The terms "first", "second", etc. in the following text are only used to distinguish between the two terms and have no other special meanings.

[0034] In the traditional machine learning field, a large number of tools are programmatic machine learning libraries, such as R, Python, and deep learning libraries such as Caffe and TensorFlow. These machine learning tools have certain barriers to entry and a high learning curve, making them unsuitable for general business and data analysts.

[0035] Currently, when creating a model project on a common machine learning platform, you first need to create an experiment, then drag components from the component bar to the experiment based on the experiment's requirements. After connecting the components and configuring their parameters, you can run the experiment. After each experiment is created, there's no way to quickly locate erroneous nodes and provide correction suggestions. Therefore, if an error occurs during the visual AI modeling process, it's difficult to troubleshoot, quickly identify the problem, and resolve it. Rebuilding the entire model is time-consuming and inconvenient, and this also makes the visual machine learning platform less usable.

[0036] Figure 1 FIG. 1 is a flow chart of an embodiment of a method for locating modeling errors according to the present disclosure, as shown in FIG. Figure 1 As shown:

[0037] Step 101: Set up the model running environment and add components required to build the target model in the model running environment.

[0038] Step 102: Establish a model flow chart corresponding to the target model based on the components, and configure parameters of the components to establish the target model.

[0039] In one embodiment, a predetermined graphical algorithm component can be selected and dragged to the design area for visual modeling, a visual model flow chart can be built, and the parameters of the algorithm nodes can be adjusted.

[0040] The model runtime environment can be a machine learning model project runtime environment, and the target model is a model project. A model project refers to a component-based data workflow or data application, or an experimental model project, built by a machine learning platform user on the platform. The machine learning platform allows users to obtain reasonable model projects based on existing data.

[0041] To create a model project on the machine learning platform, you first need to create an experiment. Then, based on the experimental requirements, you need to drag components from the component bar. Within this experiment, you need to build a model flow chart corresponding to the target model, configure the component parameters, and then run the experiment. The machine learning runtime environment (model runtime environment) is a virtual environment within the machine learning platform used to create and run model projects. Within this environment, you can create and run model projects.

[0042] The model flowchart can be a visual model flowchart. The various components in the model project constitute the nodes of the model project. For example, a node in the visual model flowchart is an algorithm training node, and its parameters such as the number of trees, tree depth, training columns, and target columns are configured.

[0043] Components are units of operations that can be called and executed on a machine learning platform, representing various algorithms or data sources. Examples include data import and export, data processing, data analysis, model project training, or prediction. Add components to the machine learning model project runtime environment, link the required components to inputs and outputs in a specified order, and create a model project (target model).

[0044] Step 103: Run the target model, obtain first log information of the target model, and extract error query information from the first log information.

[0045] The first log information may be log information of a node in the visual model flow chart. The first log information includes information such as exceptions and warnings. The information such as exceptions and warnings is extracted from the first log information to generate error query information. Various existing methods can be used to extract the error query information from the first log information.

[0046] Step 104: Input the error query information into a preset knowledge graph, use the knowledge graph to determine whether an error occurs in the target model, and obtain corresponding error location information.

[0047] The knowledge graph is a semantic network composed of interconnected knowledge points, used for knowledge reasoning and automated question-answering. Information such as exceptions and alarms is input into a pre-set knowledge graph, which is then used to determine whether the target model has errors and obtain the corresponding error location information. A visual interface is provided to display error location information, allowing code adjustments based on the error location information.

[0048] Based on the method for locating modeling errors in the above embodiment, high-quality machine learning modeling can be achieved, allowing developers to develop and train machine learning without coding, which can significantly speed up model development efficiency.

[0049] In one embodiment, before extracting the error query information from the first log information, a knowledge graph is pre-built. The knowledge graph can be pre-built or constructed during the training and operation of the target model. Figure 2 FIG. 1 is a flow chart of constructing a knowledge graph in one embodiment of a method for locating modeling errors according to the present disclosure, as shown in FIG. Figure 2 As shown:

[0050] Step 201: Acquire second log information corresponding to the target model, and extract error information from the second log information; the error information includes: exception information, warning information, etc.

[0051] The first and second log information include log files corresponding to nodes in the model flow chart. For example, during model training, each node generates second log information. The logs of each node can be integrated into the log graph according to the workflow.

[0052] Step 202: Obtain the logical relationship and causal relationship corresponding to the error information.

[0053] You can use algorithms such as Apriori or FP-growth to analyze the associated error information in the logs and obtain relevant logical and causal relationships. The Apriori algorithm is an association rule mining algorithm, and the FP-growth algorithm is an association analysis algorithm.

[0054] Step 203: Obtain a knowledge graph triple based on the logical relationship, causal relationship, and error information; wherein the knowledge graph triple includes: a first entity, a second entity, and relationship information between the first entity and the second entity.

[0055] In one embodiment, the smallest unit for constructing a knowledge graph is a knowledge graph triple. A knowledge graph triple consists of two knowledge graph entities and the attribute relationship between them. The basic form of a knowledge graph triple is entity-relationship-entity. Based on logical and causal relationships, error information can be processed in real time to extract knowledge, obtain entity and relationship information, and send it to the knowledge graph database for use in constructing the knowledge graph.

[0056] Step 204: construct a knowledge graph based on the knowledge graph triples.

[0057] Extract knowledge graph triples from error messages and construct a knowledge graph. The knowledge graph constructed using triple data has the logical structure capability of knowledge reasoning and can be used to locate errors.

[0058] In one embodiment, a knowledge graph can be dynamically generated based on the second log information, and the cause of the error information can be located through the knowledge graph. During the training process, each node will generate log information, and the logs of each node will be integrated together according to the workflow and integrated into the log graph. According to the correlation between the nodes in the visual model flow chart, the source of the error information can be inferred, and the logical relationship and causal relationship between the nodes can be explored to narrow the scope of the problem and find the root cause node of the problem. In the visual machine learning development model, visual error troubleshooting and interactive debugging can be provided to achieve high-quality machine learning modeling, allowing developers to develop and train machine learning without coding, thereby speeding up model development efficiency.

[0059] In one embodiment, multiple methods may be used to obtain knowledge graph triples. Figure 3 FIG. 1 is a flow chart of obtaining knowledge graph triples in one embodiment of the method for locating modeling errors according to the present disclosure, as shown in FIG. Figure 3 As shown:

[0060] Step 301: Extract the first entity and relationship information from the error information. The first entity includes the first node in the model flow chart and error entity information. For example, the error entity information includes classification, number of trees, tree depth, and number of training attempts. The relationship information includes error occurrence, less than a preset value, and less than a preset number of attempts.

[0061] Step 302: Based on the logical and causal relationships, error source information corresponding to the error information is obtained, and a second entity is generated based on the error source information. The second entity includes error location information, which includes the second node in the model flow chart and error cause information. The first node and the second node can be the same or different. For example, the error cause information includes cause information corresponding to the first node, such as "classification error," "number of trees less than a preset value," or "number of training times less than a preset number."

[0062] Step 303: normalize the first entity, the relationship information, and the second entity.

[0063] By normalizing the first entity, relationship information and second entity, different descriptions with the same meaning in the first entity, relationship information and second entity can be integrated, so that redundant descriptions with the same meaning can be eliminated, making the first entity, relationship information and second entity become normalized standard entities and standard relationship texts.

[0064] Step 304: Generate a knowledge graph triple based on the normalized first entity, relationship information, and second entity.

[0065] Figure 4A FIG4 is a flow chart of obtaining error location information in one embodiment of the method for locating modeling errors according to the present disclosure, as shown in FIG4 :

[0066] Step 401: normalize the error query information.

[0067] By normalizing the error query information, different descriptions with the same meaning in the error query information can be integrated, so that redundant descriptions with the same meaning can be eliminated to form normalized standard entities and standard relationship texts.

[0068] like Figure 4B As shown, a model runtime environment is set up, components required to build the target model are added to the model runtime environment, and a visual model flowchart corresponding to the target model is created based on the components. Nodes in the visual model flowchart include: dataset, data preprocessing, algorithm training, model evaluation, etc. Parameters are configured for the components to build the target model; first log information of the algorithm training node is obtained, and error query information is extracted from the first log information. Error query information may include: "The number of trees of the first node is less than a preset value" and so on.

[0069] Step 402: Generate a query statement with a triple structure based on the normalized error query information.

[0070] Step 403: Input the query statement into the knowledge graph to perform a query and obtain the second entity.

[0071] In one embodiment, filtering is performed based on entity terms and relational terms to extract answers from the knowledge graph. The query statement can be SPARQL. The normalized error query information is "The number of trees at the first node is less than a preset value." Using various existing methods, a query statement with a triple structure, "SELECT X WHERE (The number of trees at the first node is less than a preset value, x)," is generated and input into the knowledge graph for querying. The second entity at the other end of the "less than a preset value" relationship with the entity "The number of trees at the first node" is found. The second entity includes error location information. The error location information includes: the second node in the model flow chart and error cause information. For example, the error cause information includes: the cause information corresponding to "The number of trees at the first node is less than a preset value." The error location information can be the second node plus the cause information corresponding to "The number of trees at the first node is less than a preset value."

[0072] In step 404, the second entity is input into a preset expert system to obtain an error correction solution corresponding to the error query information output by the expert system. To obtain the error correction solution, various existing methods can be used to automatically modify the configuration information of the component based on the error correction solution to generate and run a new target model.

[0073] An expert system is an intelligent computer program that incorporates a wealth of expert-level knowledge and experience in a specific field. It leverages the knowledge and problem-solving methods of human experts to solve problems in that field. Expert systems reason and make judgments based on the knowledge and experience of one or more experts in a field, simulating the decision-making process of human experts to solve complex problems that require human expertise.

[0074] For example, an expert system can be pre-configured to input information such as the second node in the model flow chart and the cause information corresponding to "the number of trees for the first node is less than a preset value" into the expert system, which then outputs a corresponding error correction solution. A visual interface can be provided to display error location information and error correction solutions, providing suggestions to the user.

[0075] Model developers can be frontline operations engineers or those working on specific industry solutions. They only need to understand the business and don't need IDE tools for code-level debugging of models. Based on the aforementioned method for locating modeling errors, a knowledge graph is generated based on log information. Alarm information and operation logs of the model are sent to the log graph in real time. Based on the associations between various nodes, the logical and causal relationships between nodes are mined. This allows inferring the source of the error, narrowing the scope of the problem, and finding the root cause, thereby improving the efficiency of problem discovery and diagnosis.

[0076] In one embodiment, Figure 5 As shown, the present disclosure provides an apparatus 50 for locating modeling errors, comprising: a component processing module 51 , a model building module 52 , an information acquisition module 53 and an error locating module 54 .

[0077] The component processing module 51 sets up the model execution environment and adds the components required to build the target model. The model building module 52 creates a model flow chart corresponding to the target model based on the components and configures the component parameters to build the target model. The information acquisition module 52 runs the target model, obtains the first log information of the target model, and extracts error query information from the first log information. The error location module 54 inputs the error query information into a preset knowledge graph, uses the knowledge graph to determine whether the target model has errors, and obtains the corresponding error location information.

[0078] In one embodiment, Figure 6 As shown, the apparatus 50 for locating modeling errors further includes: a graph generation module 55, an elimination solution acquisition module 56, and a model updating module 57. The graph generation module 55 constructs a knowledge graph before extracting error query information from the first log information.

[0079] like Figure 7 As shown, the graph generation module 55 includes an error processing unit 551 and a graph processing unit 552. The error processing unit 551 obtains the second log information corresponding to the target model and extracts error information from the second log information; wherein the error information includes abnormal information, warning information, etc.

[0080] The graph processing unit 552 obtains the logical relationship and causal relationship corresponding to the error information, and obtains a knowledge graph triple based on the logical relationship, causal relationship and the error information; wherein the knowledge graph triple includes: the first entity, the second entity and the relationship information between the first entity and the second entity; the graph processing unit 552 constructs a knowledge graph based on the knowledge graph triple.

[0081] The first log information and the second log information include: log files corresponding to nodes of the model flow chart. The graph processing unit 552 extracts the first entity and relationship information from the error information; wherein the first entity includes: the first node in the model flow chart, error entity information, etc.

[0082] The graph processing unit 552 obtains the error source information corresponding to the error information based on the logical relationship and causal relationship, and generates a second entity according to the error source information; wherein the second entity includes: error location information; the error location information includes: the second node in the model flow chart, error cause information, etc.

[0083] The graph processing unit 552 normalizes the first entity, the relationship information, and the second entity; the graph processing unit 552 generates a knowledge graph triple based on the normalized first entity, the relationship information, and the second entity.

[0084] The error location module 54 normalizes the error query information; the error location module 54 generates a query statement with a triple structure based on the normalized error query information; the error location module 54 inputs the query statement into the knowledge graph for query to obtain the second entity.

[0085] The troubleshooting solution acquisition module 56 inputs the second entity into a preset expert system and obtains the error troubleshooting solution output by the expert system corresponding to the error query information. The model updating module 57 modifies the configuration information of the component based on the error troubleshooting solution and generates and runs a new target model.

[0086] Figure 8 FIG. 1 is a schematic diagram of a module of another embodiment of an apparatus for locating modeling errors according to the present disclosure. Figure 8 As shown, the apparatus may include a memory 81, a processor 82, a communication interface 83, and a bus 84. The memory 81 is used to store instructions, the processor 82 is coupled to the memory 81, and the processor 82 is configured to execute the above-mentioned method for locating modeling errors based on the instructions stored in the memory 81.

[0087] Memory 81 can be high-speed RAM, non-volatile memory, or a memory array. Memory 81 can also be divided into blocks, and the blocks can be combined into virtual volumes according to certain rules. Processor 82 can be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the method for locating modeling errors disclosed herein.

[0088] In one embodiment, the present disclosure provides a computer-readable storage medium storing computer instructions, and the instructions are executed by a processor to execute the method in any of the above embodiments.

[0089] The method, device and storage medium for locating modeling errors provided in the above embodiments generate a knowledge graph based on log information, and send alarm information, operation logs, etc. of model operation to the log graph in real time. According to the correlation between each node, the logical relationship and causal relationship between the nodes are mined, which can infer the source of the error, narrow the scope of error location, find the root cause node of the error problem, improve the efficiency of problem discovery and diagnosis, and achieve high-quality machine learning modeling. It allows developers to develop and train machine learning without coding, which can significantly speed up model development efficiency and improve user experience.

[0090] The methods and systems of the present disclosure may be implemented in many ways. For example, the methods and systems of the present disclosure may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above order of steps for the method is for illustration only, and the steps of the method of the present disclosure are not limited to the order specifically described above unless otherwise specified. In addition, in some embodiments, the present disclosure may also be implemented as programs recorded in a recording medium, which include machine-readable instructions for implementing the methods according to the present disclosure. Therefore, the present disclosure also covers recording media that store programs for executing the methods according to the present disclosure.

[0091] The description of the present disclosure is provided for purposes of illustration and description and is not intended to be exhaustive or to limit the disclosure to the disclosed form. Many modifications and variations will be apparent to those skilled in the art. The embodiments are selected and described in order to better illustrate the principles and practical applications of the present disclosure and to enable those skilled in the art to understand the present disclosure and design various embodiments with various modifications suitable for specific applications.

Claims

1. A method for locating modeling errors, comprising: Setting up a model operating environment, and adding components required to build a target model in the model operating environment; Establishing a model flow chart corresponding to the target model based on the components, and configuring parameters of the components to establish the target model; Running the target model, obtaining first log information of the target model, and extracting error query information from the first log information; Input the error query information into a preset knowledge graph, use the knowledge graph to determine whether an error occurs in the target model and obtain corresponding error location information; Before extracting the error query information from the first log information, the step of constructing the knowledge graph is further included, including: Acquire second log information corresponding to the target model, and extract error information from the second log information; the first log information and the second log information include: log files corresponding to nodes of the model flow chart; Obtaining logical relationships and causal relationships corresponding to the error information; According to the logical relationship and causal relationship and the error information, a knowledge graph triple is obtained, where the knowledge graph triple includes: a first entity, a second entity, and relationship information between the first entity and the second entity; wherein, according to the logical relationship and causal relationship and the error information, obtaining the knowledge graph triple includes: Extracting the first entity and the relationship information from the error information, wherein the first entity includes: a first node in the model flow chart and error entity information; Acquire error source information corresponding to the error information based on the logical relationship and causal relationship, and generate the second entity according to the error source information, wherein the second entity includes error location information, and the error location information includes the second node in the model flow chart and error cause information; The knowledge graph is constructed based on the knowledge graph triples.

2. The method according to claim 1, wherein The error information includes: abnormal information and warning information.

3. The method of claim 2, further comprising: performing normalization processing on the first entity, the relationship information, and the second entity; The knowledge graph triple is generated based on the normalized first entity, the relationship information and the second entity.

4. The method of claim 3, wherein inputting the error query information into a preset knowledge graph, and using the knowledge graph to determine whether an error occurs in the operation of the target model and obtaining corresponding error location information comprises: Normalizing the erroneous query information; generating a query statement having a triple structure according to the normalized error query information; Input the query statement into the knowledge graph to perform a query and obtain the second entity.

5. The method of claim 4, further comprising: The second entity is input into a preset expert system to obtain an error elimination solution corresponding to the error query information output by the expert system.

6. The method of claim 5, further comprising: The configuration information of the component is modified based on the error elimination solution, and a new target model is generated and run.

7. A device for locating modeling errors, comprising: A component processing module is used to set up a model operating environment and add components required to build a target model in the model operating environment; A model building module, configured to build a model flow chart corresponding to the target model based on the components, and configure parameters of the components to build the target model; an information acquisition module, configured to run the target model, acquire first log information of the target model, and extract error query information from the first log information; An error location module is used to input the error query information into a preset knowledge graph, use the knowledge graph to determine whether an error occurs in the target model and obtain corresponding error location information; A graph generation module, configured to construct the knowledge graph before extracting error query information from the first log information, comprising: an error processing unit, configured to obtain second log information corresponding to the target model and extract error information from the second log information; the first log information and the second log information include: log files corresponding to nodes of the model flow chart; A graph processing unit, configured to obtain a logical relationship and a causal relationship corresponding to the error information, and obtain a knowledge graph triple based on the logical relationship, the causal relationship, and the error information; the knowledge graph triple comprising: a first entity, a second entity, and relationship information between the first entity and the second entity; and construct the knowledge graph based on the knowledge graph triple; The graph processing unit is specifically used to extract the first entity and the relationship information from the error information; wherein, the first entity includes: the first node and error entity information in the model flow chart; based on the logical relationship and causal relationship, the error source information corresponding to the error information is obtained, and the second entity is generated according to the error source information; wherein, the second entity includes: error location information; the error location information includes: the second node and error cause information in the model flow chart.

8. The device according to claim 7, wherein The error information includes: abnormal information and warning information.

9. The device according to claim 8, wherein The graph processing unit is also used to normalize the first entity, the relationship information and the second entity; and generate the knowledge graph triple based on the normalized first entity, the relationship information and the second entity.

10. The device according to claim 9, wherein The error location module is used to normalize the error query information; generate a query statement with a triple structure based on the error query information after the normalization; input the query statement into the knowledge graph for query to obtain the second entity.

11. The apparatus of claim 10, further comprising: The elimination solution acquisition module is used to input the second entity into a preset expert system and obtain the error elimination solution corresponding to the error query information output by the expert system.

12. The apparatus of claim 11, further comprising: The model updating module is used to modify the configuration information of the component based on the error elimination solution, generate and run a new target model.

13. A network slice remapping device, comprising: Memory; and a processor coupled to the memory, wherein the processor is configured to execute the method according to any one of claims 1 to 6 based on instructions stored in the memory. 14 . A computer-readable storage medium storing computer instructions, wherein the instructions are executed by a processor to execute the method according to claim 1 .

Images