Data center machine room fault diagnosis method based on knowledge graph

Abstract

The invention discloses a data center machine room fault diagnosis method based on a knowledge graph, and relates to the technical field of machine room fault diagnosis optimization, and the method comprises the steps: obtaining a historical fault case of a data center; based on the historical fault case, constructing a knowledge graph according to an entity extraction model, a relationship extraction model and an entity disambiguation model; and mapping a fault problem input by a user into the knowledge graph, and determining an optimal solution for the fault problem based on an entity linking scheme, a relationship guessing scheme and a multi-path sorting scheme. According to the invention, the accuracy of fault diagnosis optimization can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of computer room fault diagnosis optimization, and in particular to a data center computer room fault diagnosis method based on a knowledge graph. Background Art

[0002] In recent years, the fault diagnosis scheme configuration of data center computer rooms generally uses traditional TF-IDF and BM25 algorithms to perform keyword retrieval of user input questions. TF-IDF and BM25 algorithms have improved the accuracy of question retrieval to a certain extent, but this scheme has some limitations. They cannot focus on the core questions input by users, which will lead to insufficient accuracy of fault diagnosis optimization.

[0003] The above contents are only used to assist in understanding the technical solution of the present application and do not constitute an admission that the above contents are prior art. Summary of the invention

[0004] The main purpose of this application is to provide a data center computer room fault diagnosis method based on a knowledge graph, aiming to solve the technical problem that the traditional fault diagnosis optimization solution is not accurate enough due to the inability to focus on the core issues input by the user.

[0005] To achieve the above objectives, the present application proposes a data center computer room fault diagnosis method based on a knowledge graph, and the data center computer room fault diagnosis method based on a knowledge graph includes:

[0006] Obtain historical failure cases of the data center;

[0007] Based on the historical failure cases, a knowledge graph is constructed according to an entity extraction model, a relationship extraction model, and an entity disambiguation model;

[0008] The fault problem input by the user is mapped into the knowledge graph, and the optimal solution for the fault problem is determined based on the entity linking scheme, the relationship guessing scheme and the multi-path sorting scheme.

[0009] In one embodiment, the step of constructing a knowledge graph based on the historical fault cases according to an entity extraction model, a relationship extraction model, and an entity disambiguation model includes:

[0010] Extracting entity information from the historical fault cases based on the entity extraction model;

[0011] Based on the relationship extraction model, extracting target entity relationships from the historical fault cases according to the entity information;

[0012] Based on the entity disambiguation model, performing entity disambiguation processing on the entity information to obtain target entity information;

[0013] A knowledge graph is constructed based on the target entity information and the target entity relationships.

[0014] In one embodiment, the entity extraction model includes a Bert network, a Bi-LSTM network and a global pointer network, and the global pointer network introduces a rotational position encoding; wherein the step of extracting entity information from the historical fault case based on the entity extraction model includes:

[0015] Inputting the historical fault cases into the entity extraction model, and obtaining the Chinese word vectors of the historical fault cases through the Bert network;

[0016] Extracting semantic features of the historical fault cases through the Bi-LSTM network;

[0017] Entity information is extracted from the historical fault cases through the global pointer network according to the Chinese word vector and the semantic features.

[0018] In one embodiment, the step of extracting target entity relationships from the historical fault cases based on the relationship extraction model and according to the entity information includes:

[0019] Dividing the entity information into training set data and test set data;

[0020] Based on the training set data, fine-tune the Bert network of the relationship extraction model to obtain a preset relationship extraction model, and determine the training entity relationships of all entities in the training set data based on the preset relationship extraction model;

[0021] Inputting the test set data into the preset relationship extraction model to obtain a test entity relationship;

[0022] Determining a target test entity relationship from the test entity relationships according to a majority voting mechanism;

[0023] Based on the training entity relationship and the target test entity relationship, a target entity relationship is determined.

[0024] In one embodiment, the step of performing entity disambiguation processing on the entity information based on the entity disambiguation model to obtain target entity information includes:

[0025] Based on the entity disambiguation model, converting the entity information into vector information;

[0026] Determine similarity information between the entity to be disambiguated and the candidate entity in the entity information according to the vector information;

[0027] Determine the best matching entity of the entity to be disambiguated from the candidate entities according to the similarity information;

[0028] The target entity information is determined according to the best matching entity of each entity to be disambiguated in the entity information.

[0029] In one embodiment, the step of mapping the fault problem input by the user to the knowledge graph and determining the optimal solution to the fault problem based on the entity linking scheme, the relationship guessing scheme and the multi-path sorting scheme includes:

[0030] Based on the entity linking scheme, determine a mapping entity of the fault problem in the knowledge graph;

[0031] Based on the relationship guessing scheme, determine the mapping entity relationship of the fault problem in the knowledge graph;

[0032] Based on the mapping entities and the mapping entity relationships, an optimal solution to the fault problem is determined according to the multi-path sorting scheme.

[0033] In one embodiment, the step of determining the optimal solution to the fault problem based on the mapping entity and the mapping entity relationship according to the multi-path sorting scheme includes:

[0034] Determine a relationship path according to the mapping entities and the mapping entity relationships;

[0035] Determine the occurrence frequency of the relationship path in the knowledge graph according to the mapping entity;

[0036] Determining the probability of the relationship path according to the occurrence frequency;

[0037] A target relationship path is determined from the relationship paths according to the probability, and the target relationship path is determined as an optimal solution to the fault problem.

[0038] In addition, to achieve the above purpose, the present application also proposes a data center computer room fault diagnosis device based on a knowledge graph, and the data center computer room fault diagnosis device based on a knowledge graph includes:

[0039] An acquisition module is used to obtain historical failure cases of the data center;

[0040] A construction module, used to construct a knowledge graph based on the historical fault cases according to an entity extraction model, a relationship extraction model, and an entity disambiguation model;

[0041] The determination module is used to map the fault problem input by the user into the knowledge graph, and determine the optimal solution to the fault problem based on the entity linking scheme, the relationship guessing scheme and the multi-path sorting scheme.

[0042] In addition, to achieve the above-mentioned objectives, the present application also proposes a data center computer room fault diagnosis device based on a knowledge graph, the device comprising: a memory, a processor, and a computer program stored on the memory and executable on the processor, the computer program being configured to implement the steps of the data center computer room fault diagnosis method based on the knowledge graph as described above.

[0043] In addition, to achieve the above-mentioned purpose, the present application also proposes a storage medium, which is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by the processor, the steps of the data center computer room fault diagnosis method based on the knowledge graph as described above are implemented.

[0044] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, it implements the steps of the data center computer room fault diagnosis method based on the knowledge graph as described above.

[0045] One or more technical solutions proposed in this application have at least the following technical effects:

[0046] The data center computer room fault diagnosis method based on knowledge graph proposed in the present application obtains historical fault cases of the data center; based on the historical fault cases, a knowledge graph is constructed according to an entity extraction model, a relationship extraction model and an entity disambiguation model; the fault problem input by the user is mapped to the knowledge graph, and the optimal solution for the fault problem is determined based on an entity linking scheme, a relationship guessing scheme and a multi-path sorting scheme, thereby solving the technical problem that the traditional fault diagnosis optimization scheme is not accurate enough due to the inability to perform focused retrieval on the core problem input by the user. Compared with the prior art, the present application obtains each entity and each entity relationship of the data center through entity extraction, relationship extraction and entity disambiguation, and then constructs a high-reliability knowledge graph, and then uses the knowledge graph to quickly and accurately diagnose the fault problem input by the user, and generates a corresponding fault diagnosis scheme, which can not only improve the fault diagnosis optimization rate of the data center computer room, but also improve the accuracy of fault diagnosis optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0048] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0049] Figure 1 A flow chart of the first embodiment of the method for diagnosing a fault in a data center computer room based on a knowledge graph provided in this application;

[0050] Figure 2 A schematic diagram of a knowledge graph provided for Embodiment 1 of the method for diagnosing a fault in a computer room of a data center based on a knowledge graph of the present application;

[0051] Figure 3 A schematic diagram of a knowledge graph provided for Embodiment 1 of the method for diagnosing a fault in a computer room of a data center based on a knowledge graph of the present application;

[0052] Figure 4 A system structure diagram provided for the second embodiment of the method for diagnosing a fault in a data center computer room based on a knowledge graph of the present application;

[0053] Figure 5 A schematic diagram of multiple identification entities provided for the second embodiment of the method for diagnosing a fault in a data center computer room based on a knowledge graph of the present application;

[0054] Figure 6 A schematic diagram of multiple identification entities provided for the second embodiment of the method for diagnosing a fault in a data center computer room based on a knowledge graph of the present application;

[0055] Figure 7 A schematic diagram of an entity representation of a training data set provided in Embodiment 2 of the method for diagnosing a fault in a data center computer room based on a knowledge graph of the present application as data storage;

[0056] Figure 8 An overall schematic diagram of a relationship extraction model provided in Embodiment 2 of the method for diagnosing a fault in a data center computer room based on a knowledge graph of the present application;

[0057] Fig. 9 A triplet diagram of entity-entity relationships provided in Example 2 of the data center computer room fault diagnosis method based on knowledge graph of this application;

[0058] Fig.10 This is a schematic diagram of the module structure of a data center computer room fault diagnosis device based on a knowledge graph according to an embodiment of the present application;

[0059] Fig.11 This is a schematic diagram of the equipment structure of the hardware operating environment involved in the data center computer room fault diagnosis method based on the knowledge graph in the embodiment of the present application.

[0060] The purpose, features and advantages of this application will be further described in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0061] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application and are not used to limit the present application.

[0062] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0063] The main solution of the embodiment of the present application is: to obtain historical failure cases of the data center; based on the historical failure cases, to build a knowledge graph according to the entity extraction model, the relationship extraction model and the entity disambiguation model; to map the failure problem input by the user into the knowledge graph, and to determine the optimal solution to the failure problem based on the entity linking scheme, the relationship guessing scheme and the multi-path sorting scheme.

[0064] In this embodiment, for ease of description, the following is explained using the identification of data center computer room fault diagnosis equipment based on the knowledge graph as the execution subject.

[0065] Traditional fault diagnosis optimization solutions are not accurate enough because they cannot focus on searching the core issues input by users.

[0066] It can be seen from the above embodiments that the data center computer room fault diagnosis method based on the knowledge graph proposed in the present application obtains historical fault cases of the data center; based on the historical fault cases, a knowledge graph is constructed according to the entity extraction model, the relationship extraction model and the entity disambiguation model; the fault problem input by the user is mapped to the knowledge graph, and the optimal solution for the fault problem is determined based on the entity linking scheme, the relationship guessing scheme and the multi-path sorting scheme, which solves the technical problem that the traditional fault diagnosis optimization scheme is not accurate enough due to the inability to focus on the core problem input by the user. Compared with the prior art, the present application obtains each entity and each entity relationship of the data center through entity extraction, relationship extraction and entity disambiguation, and then constructs a high-reliability knowledge graph, and then uses the knowledge graph to quickly and accurately diagnose the fault problem input by the user, and generate corresponding fault diagnosis solutions, which can not only improve the fault diagnosis optimization rate of the data center computer room, but also improve the accuracy of fault diagnosis optimization.

[0067] It should be noted that the execution subject of this embodiment can be a computing service device with data processing, network communication and program running functions, such as a tablet computer, a personal computer, a mobile phone, etc., or an electronic device capable of realizing the above functions, a data center computer room fault diagnosis device based on a knowledge graph, etc. The following takes a data center computer room fault diagnosis device based on a knowledge graph as an example to illustrate this embodiment and the following embodiments.

[0068] Based on this, the present application embodiment provides a data center computer room fault diagnosis method based on knowledge graph, referring to Figure 1 , Figure 1 This is a flow chart of the first embodiment of the data center computer room fault diagnosis method based on knowledge graph of the present application.

[0069] In this embodiment, the data center computer room fault diagnosis method based on the knowledge graph includes steps S10 to S30:

[0070] Step S10, obtaining historical failure cases of the data center;

[0071] Step S20, based on the historical fault cases, construct a knowledge graph according to the entity extraction model, the relationship extraction model and the entity disambiguation model;

[0072] It should be noted that the entity extraction model can be used to extract entity information with specific meaning from the natural language text of historical fault cases, such as device name, device manufacturer, fault description, troubleshooting steps or solutions.

[0073] It should be noted that the relationship extraction model can be used to extract semantic information (ie, entity relationships) between entities from the natural language text of historical fault cases.

[0074] It should be noted that the entity disambiguation model can be used to perform entity disambiguation on the entity information extracted from the entity extraction model, so as to map different expressions of the same entity to a standard entity representation, thereby providing more accurate data for the subsequent construction of the knowledge graph.

[0075] It should be noted that the entities and entity relationships extracted from the historical failure cases of the data center can be stored in the NEO4J graph database, and then the knowledge graph of network cloud room failures can be constructed by using the NEO4J database and visualized.

[0076] It should be noted that if Figure 2As shown, the knowledge graph includes nodes representing entities and edges representing entity relationships. The nodes and edges have attributes that describe graph elements and relationships, represent semantic relationships between entities, and provide a structured representation of the knowledge base. The efficient graph data traversal and query functions provided by the graph database tool NEO4J can be used to quickly retrieve and analyze data in the knowledge graph.

[0077] It should be noted that after completing the construction of the knowledge graph, new entity attributes and other enriched knowledge graph information resources can be added to the knowledge graph through more newly collected data resources about the data room, and the knowledge graph can be iteratively optimized and expanded continuously to improve the quality and coverage of the knowledge graph.

[0078] It should be noted that in order to allow users to more intuitively understand the fault distribution and equipment association of the data center, the knowledge graph can be visualized using a network topology diagram or a relationship diagram. For example, the entities and entity relationships in the knowledge graph can be represented by nodes and edges, and different information can be marked on the nodes and edges, such as equipment model, fault type, etc. In this way, users can conduct in-depth understanding and analysis through interactive operations, and can more intuitively locate and troubleshoot faults in a visual way.

[0079] Step S30, mapping the fault problem input by the user into the knowledge graph, and determining the optimal solution to the fault problem based on the entity linking scheme, the relationship guessing scheme and the multi-path sorting scheme.

[0080] It should be noted that in the fault diagnosis and optimization part, the fault problems input by the user can be mapped to the knowledge graph through the question-answering system, and then based on functional technologies such as entity linking, relationship guessing, and multi-path sorting, the optimal solution is returned according to the sorting results.

[0081] In a specific implementation, in order to facilitate user use, a user interface may be provided, including a command line tool, a Web interface, etc., so that the user can input fault problems and view the optimal solution.

[0082] It should be noted that if Figure 3As shown in the figure, during the system development process, security and privacy protection issues also need to be fully considered. In order to ensure the security of the system, before the user enters the fault problem, it is necessary to identify the user through identity authentication, access control and encrypted transmission to improve the security construction of the system. At the same time, during the data processing process, data desensitization and permission control are used to protect the user's query records, device configuration and other information. When the user logs in, the user's identity is confirmed based on the user name and password, multi-factor authentication, and single sign-on to control their access to the system; role access control can also be assigned through system permissions, and different permissions can be assigned to each role to control the resources and operations accessed by the user. The confidentiality and integrity of data during network transmission is controlled through SSL / TLS, and the database is encrypted. Use shielding settings to replace sensitive data with symbols to protect against data leakage risks.

[0083] In a feasible implementation, the step of mapping the fault problem input by the user into the knowledge graph and determining the optimal solution to the fault problem based on the entity linking scheme, the relationship guessing scheme and the multi-path sorting scheme includes: determining the mapping entity of the fault problem in the knowledge graph based on the entity linking scheme; determining the mapping entity relationship of the fault problem in the knowledge graph based on the relationship guessing scheme; and determining the optimal solution to the fault problem based on the multi-path sorting scheme based on the mapping entity and the mapping entity relationship.

[0084] It should be noted that, for the fault problems input by users using natural language, the problem analysis module can be used to pre-process these fault problems, and invalid information such as time can be removed from the redundant input of the user.

[0085] In a specific implementation, based on the entity linking scheme, the specific steps for determining the mapping entity of the fault problem in the knowledge graph are: determining the entity text in the fault problem through an entity extraction model, and then determining the sentence vector of the entity text based on a text retrieval model, and then calculating the cosine similarity between the sentence vector and the target sentence vector matrix of the target entity in the knowledge graph to obtain a similarity list, and finally determining the mapping entity of the fault problem in the knowledge graph based on the similarity list.

[0086] In the specific implementation, for the pre-processed fault problems, the entity extraction model trained when building the network cloud computer room knowledge graph can be used to extract the entities mentioned in the fault problems, and try to use entity retrieval and linking technology to map the problem entities in the user input to the problem entities in the knowledge graph. Specifically, the mapping can be done through entity retrieval technology. The core of entity retrieval technology is cosine similarity, and the formula is as follows:

[0087]

[0088] In the formula, It is the entity text obtained after the entity extraction model processes the fault problem, and then the sentence vector is obtained by parsing and transforming the entity text based on the text retrieval model (such as the Simbert model); the matrix A is the sentence vector matrix obtained by parsing and transforming a certain type of entity in the knowledge graph through the sentence vector model. Finally, a similarity list can be obtained.

[0089] In the specific implementation, other important attributes of the entity itself, such as "manufacturer" and "device type", can also be considered when determining the mapping entity. From a business perspective, when different manufacturers' equipment face the same problem, the corresponding reasons may be different. Even for the same reason, the corresponding solutions are not exactly the same. For this hard constraint attribute, it can be used as the primary condition for filtering entities. The cosine similarity-based entity linking method with hard constraint screening has stronger business explanation and rationality in the network cloud field.

[0090] In a specific implementation, based on the relationship guessing scheme, the specific steps for determining the mapping entity relationship of the fault problem in the knowledge graph are: determining the entity relationship in the fault problem through a relationship extraction model, and then determining the sentence vector of the entity relationship based on a text retrieval model, and then calculating the cosine similarity between the sentence vector and the target sentence vector matrix of the target entity relationship in the knowledge graph to obtain a similarity list, and finally determining the mapping entity relationship of the fault problem in the knowledge graph based on the similarity list.

[0091] In one embodiment, the step of determining the optimal solution to the fault problem based on the mapping entities and the mapping entity relationships according to the multi-path sorting scheme includes: determining a relationship path according to the mapping entities and the mapping entity relationships; determining the frequency of occurrence of the relationship path in the knowledge graph according to the mapping entities; determining the probability of the relationship path according to the frequency of occurrence; determining a target relationship path from the relationship path according to the probability, and determining the target relationship path as the optimal solution to the fault problem.

[0092] It should be noted that multi-path ranking is a key step to evaluate the strength of the relationship between entities on different paths. Multi-path ranking ranks these paths by evaluating the strength of the relationship between entities on different paths to determine the most likely relationship path and return the optimal solution.

[0093] It should be noted that, for several mapping entities, some mapping entities may be filtered out by attribute information, such as manufacturer information, etc.; for the remaining mapping entities, multi-path sorting may be used for analysis.

[0094] In the specific implementation, for example, the mapping entity and the mapping entity relationship include 3 entities and 2 relationships, then the corresponding relationship path is 2 hops: "Entity 1--(Relationship 1)-->Entity 2--(Relationship 2)-->Entity 3". Specifically, we can count the frequency of occurrence of these relationship paths in the knowledge graph and convert them into probabilities. The formula is as follows:

[0095]

[0096] In the formula, Is the entity linked to 1q i The probability of By q i Departure, linked to entity 2r j The probability of It is composed of entity 2r j Departure, linked to entity 3s k The probability of; Z is the normalization factor,

[0097] It can be understood that after the probabilities of the various relationship paths are obtained, the relationship path with the maximum probability can be determined as the optimal solution.

[0098] This embodiment obtains historical fault cases of the data center; based on the historical fault cases, a knowledge graph is constructed according to an entity extraction model, a relationship extraction model, and an entity disambiguation model; the fault problem input by the user is mapped to the knowledge graph, and based on the entity linking scheme, the relationship guessing scheme, and the multi-path sorting scheme, the optimal solution to the fault problem is determined, thereby solving the technical problem that the traditional fault diagnosis optimization scheme is not accurate enough due to the inability to focus on the core problem input by the user. Compared with the prior art, the present application obtains each entity and each entity relationship of the data center through entity extraction, relationship extraction, and entity disambiguation, and then constructs a high-reliability knowledge graph, and then uses the knowledge graph to quickly and accurately diagnose the fault problem input by the user, and generates a corresponding fault diagnosis scheme, which can not only improve the fault diagnosis optimization rate of the data center computer room, but also improve the accuracy of fault diagnosis optimization.

[0099] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as those in the above-mentioned embodiment 1 can be referred to the above introduction, and will not be repeated in the following. Figure 4 , step S20 also includes steps S201 to S204:

[0100] Step S201, extracting entity information from the historical fault cases based on the entity extraction model;

[0101] In a feasible implementation manner, the entity extraction model includes a Bert network, a Bi-LSTM network and a global pointer network, and the global pointer network introduces rotational position coding; wherein, the step of extracting entity information from the historical fault cases based on the entity extraction model includes: inputting the historical fault cases into the entity extraction model, obtaining the Chinese word vectors of the historical fault cases through the Bert network; extracting the semantic features of the historical fault cases through the Bi-LSTM network; and extracting entity information from the historical fault cases according to the Chinese word vectors and the semantic features through the global pointer network.

[0102] It should be noted that the Global Pointer Network is a decoding method based on span classification. It regards the beginning and the end as a whole for judgment, so it has a more "global view" and ensures that training, prediction, and online evaluation are all evaluated at the entity level. The core idea of ​​the Global Pointer Network is the introduction of RoPE (Rotational Position Encoding). The entity extraction model also uses the Bert pre-training model (ie, Bert network) to obtain Chinese word vectors, and uses the Bi-LSTM neural network framework (ie, Bi-LSTM network) to extract semantic features.

[0103] It should be noted that the design of the global pointer network utilizes the idea of ​​global normalization to identify entities, and can identify nested entities and non-nested entities indiscriminately, achieving excellent results in both nested and non-nested entities.

[0104] In the specific implementation, assume that the length of the text sequence to be recognized is n, and assume that there is only one entity to be recognized. Each entity to be recognized is a continuous segment of the sequence with unlimited length and can be nested with each other (there is an intersection between two entities). Then the sequence has n(n+1) / 2 "candidate entities", that is, a sequence of length n has n(n+1) / 2 different continuous subsequences, which contain all possible entities. We need to find the real entity from these n(n+1) / 2 "candidate entities", which is a "n(n+1) / 2 select k" multi-label classification problem. If there are m entity types that need to be recognized, then it is made into m "n(n+1) / 2 select k" multi-label classification problems.

[0105] In theory, the design concept of the global pointer network is more reasonable than that of CRF. In practice, the global pointer network is completely parallel during training, eliminating the steps of recursively calculating the denominator and performing dynamic programming during prediction like CRF. Ideally, the time complexity is O(1).

[0106] In the specific implementation, Figure 5 and Figure 6 As shown in the GlobalPointer multi-head recognition entity graph, the two two-dimensional matrices represent entities of fault description and device name, respectively. In the upper triangular matrix shown in the figure, the horizontal coordinate position with a value of 1 represents the starting boundary of the candidate entity, and the vertical coordinate position represents the ending boundary of the candidate entity. Specifically, taking the fault description as an example, it can be determined that the coordinate position with a value of 1 in the entity graph of the fault description is (7, 16), so the fault description entity is from the 7th to the 16th character in a paragraph of text; taking the device name as an example, it can be determined that the coordinate position with a value of 1 in the entity graph of the device name is (1, 4), so the device name entity is from the 1st to the 4th character in a paragraph of text.

[0107] Step S202, based on the relationship extraction model, extracting target entity relationships from the historical fault cases according to the entity information;

[0108] In a feasible implementation, the step of extracting target entity relationships from the historical fault cases based on the entity information based on the relationship extraction model includes: dividing the entity information into training set data and test set data; based on the training set data, fine-tuning the Bert network of the relationship extraction model to obtain a preset relationship extraction model, and determining the training entity relationships of all entities in the training set data based on the preset relationship extraction model; inputting the test set data into the preset relationship extraction model to obtain test entity relationships; determining the target test entity relationships from the test entity relationships according to the majority voting mechanism; and determining the target entity relationships based on the training entity relationships and the target test entity relationships.

[0109] It should be noted that the entity information can be divided into training set data and test set data, wherein the training set data can be used to fine-tune the Bert network of the relationship extraction model, and the fine-tuned preset relationship extraction model is used to determine the training entity relationships between all entities in the training set data (that is, the entity relationships of all entities in the training set data), and then the test set data is input into the preset relationship extraction model to extract all entity relationships between entities, and then the target test entity relationships of all entity relationships are determined by majority voting technology.

[0110] In the specific implementation, the training set is first used to train and fine-tune the relation extraction model BERT to obtain a model containing entity and relation information for the field of data center computer room failures. Then the fine-tuned BERT model is used to extract the semantic representation of all entities in the training set and store it as data, such as Figure 7 shown.

[0111] In the specific implementation, when extracting entity relations, the entity information of the test set data is first sent to the fine-tuned BERT model to obtain its semantic information vector, so as to realize KNN retrieval of the test input, and obtain k examples related to the test input entity information. Then, through the majority voting mechanism, the relationship labels that account for the majority of the k examples are selected to obtain the entity relations of the test input.

[0112] It is understandable that by introducing the fine-tuned BERT model, the hidden representation semantic information of the test set entities is obtained. This approach can enhance the correlation between the test input entity and the example entity relationship, significantly improve the overall retrieval example quality, and effectively solve the problem that the model is not effective enough due to the low correlation between entities and relationships; and because the fine-tuned model can accurately identify the displayed similar NULLs, the problem of over-predicting NULLs will be greatly alleviated. The overall structure of the relationship extraction model is shown in Figure 2. Figure 8 shown.

[0113] It should be noted that if Fig. 9 As shown, after the relationship extraction model, the obtained entity relationship information is represented as a series of triples, such as <entity 1, relationship, entity 2>, <equipment 1, fault, attribute 1>, <equipment 1, fault, solution 1>, etc.

[0114] Step S203, performing entity disambiguation processing on the entity information based on the entity disambiguation model to obtain target entity information;

[0115] In a feasible implementation, the step of performing entity disambiguation processing on the entity information based on the entity disambiguation model to obtain target entity information includes: converting the entity information into vector information based on the entity disambiguation model; determining similarity information between the entity to be disambiguated and the candidate entity in the entity information based on the vector information; determining the best matching entity of the entity to be disambiguated from the candidate entities based on the similarity information; and determining the target entity information based on the best matching entity of each entity to be disambiguated in the entity information.

[0116] It should be noted that the entity disambiguation model is used to map different expressions of the same entity to a standard entity representation, and different entity names are mapped to the same entity identifier, such as pointing references to multiple servers with the same name to the same physical server. For example, the entity identifier corresponding to the entity in the knowledge graph is found in the natural language text. The goal of this task is to link the entity reference in the natural language with the entity in the knowledge graph, determine the correct entity identifier, and thus provide more accurate data for subsequent knowledge graph applications.

[0117] In the specific implementation, the entity disambiguation model adopts the vector space model method to represent entities as vectors, matches the entities to be disambiguated with possible candidate entities in the vector space and calculates the similarity information using the cosine distance method, and finally selects the best matching entity from multiple candidate entities based on the similarity information.

[0118] It is understandable that entity disambiguation and entity normalization are very important steps when performing computer room fault diagnosis and optimization. For example, when the system receives a fault report, the report may contain multiple entities, such as network equipment, services, applications, etc. However, since there may be multiple entities with the same name within the enterprise, such as multiple servers with the same name, etc., and the entity properties in the partition fault report need to be partitioned, these entities need to be disambiguated and normalized before fault diagnosis and optimization. Entity disambiguation and normalization can improve the accuracy and efficiency of tasks such as entity extraction and entity relationship mining.

[0119] Step S204: construct a knowledge graph based on the target entity information and the target entity relationship.

[0120] It should be noted that the knowledge graph includes nodes representing entities and edges representing entity relationships. The nodes and edges have attributes that describe graph elements and relationships, represent semantic relationships between entities, and provide a structured representation of the knowledge base. The efficient graph data traversal and query functions provided by the graph database tool NEO4J can be used to quickly retrieve and analyze data in the knowledge graph.

[0121] This embodiment extracts entity information from the historical fault cases based on the entity extraction model; extracts target entity relationships from the historical fault cases based on the entity information based on the relationship extraction model; performs entity disambiguation processing on the entity information based on the entity disambiguation model to obtain target entity information; and constructs a knowledge graph based on the target entity information and the target entity relationships. This embodiment automatically extracts entities and entity relationships from historical fault cases based on technologies such as entity extraction of the BERT+Bi-LSTM+GlobalPointer model, relationship extraction model based on entity-aware retrieval, and entity disambiguation, and converts them into a structured knowledge representation form, ultimately realizing the construction of a knowledge graph.

[0122] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the data center room fault diagnosis method based on the knowledge graph of the present application. More simple transformations based on this technical concept are all within the scope of protection of the present application.

[0123] This application also provides a data center computer room fault diagnosis device based on knowledge graph, please refer to Fig.10 , the data center computer room fault diagnosis device based on knowledge graph includes:

[0124] An acquisition module 10 is used to acquire historical failure cases of a data center;

[0125] A construction module 20 is used to construct a knowledge graph based on the historical fault cases according to an entity extraction model, a relationship extraction model, and an entity disambiguation model;

[0126] The determination module 30 is used to map the fault problem input by the user into the knowledge graph, and determine the optimal solution to the fault problem based on the entity linking scheme, the relationship guessing scheme and the multi-path sorting scheme.

[0127] The data center computer room fault diagnosis device based on knowledge graph provided by the present application adopts the data center computer room fault diagnosis method based on knowledge graph in the above embodiment, which can solve the technical problem that the traditional fault diagnosis optimization scheme is not accurate enough due to the inability to perform focused retrieval on the core problems input by the user. Compared with the prior art, the beneficial effects of the data center computer room fault diagnosis device based on knowledge graph provided by the present application are the same as the beneficial effects of the data center computer room fault diagnosis method based on knowledge graph provided by the above embodiment, and the other technical features of the data center computer room fault diagnosis device based on knowledge graph are the same as the features disclosed in the above embodiment method, which will not be repeated here.

[0128] The present application provides a data center computer room fault diagnosis device based on a knowledge graph, and the data center computer room fault diagnosis device based on a knowledge graph includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the data center computer room fault diagnosis method based on the knowledge graph in the above-mentioned embodiment one.

[0129] Reference below Fig.11 , which shows a schematic diagram of the structure of a data center computer room fault diagnosis device based on a knowledge graph suitable for implementing the embodiment of the present application. The data center computer room fault diagnosis device based on a knowledge graph in the embodiment of the present application may include but is not limited to mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Fig.11 The knowledge graph-based data center room fault diagnosis device shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0130] like Fig.11As shown, the data center computer room fault diagnosis equipment based on the knowledge graph may include a processing device 1001 (such as a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM: Read Only Memory) 1002 or the program loaded from the storage device 1003 to the random access memory (RAM: Random Access Memory) 1004. Various programs and data required for the operation of the data center computer room fault diagnosis equipment based on the knowledge graph are also stored in RAM1004. The processing device 1001, ROM1002 and RAM1004 are connected to each other via a bus 1005. The input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to the I / O interface 1006: input devices 1007 including, for example, a touch screen, a touchpad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; storage devices 1003 including, for example, a magnetic tape, a hard disk, etc.; and communication devices 1009. The communication device 1009 can allow the data center computer room fault diagnosis device based on the knowledge graph to communicate wirelessly or wired with other devices to exchange data. Although the figure shows a data center computer room fault diagnosis device based on the knowledge graph with various systems, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems may be implemented or have alternatively.

[0131] In particular, according to the embodiments disclosed in the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0132] The data center computer room fault diagnosis device based on knowledge graph provided by the present application adopts the data center computer room fault diagnosis method based on knowledge graph in the above embodiment, which can solve the technical problem that the traditional fault diagnosis optimization scheme cannot perform focused retrieval on the core problems input by the user, resulting in insufficient accuracy of fault diagnosis optimization. Compared with the prior art, the beneficial effects of the data center computer room fault diagnosis device based on knowledge graph provided by the present application are the same as the beneficial effects of the data center computer room fault diagnosis method based on knowledge graph provided by the above embodiment, and the other technical features of the data center computer room fault diagnosis device based on knowledge graph are the same as the features disclosed in the method of the previous embodiment, which will not be repeated here.

[0133] It should be understood that the various parts disclosed in this application can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0134] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0135] The present application provides a computer-readable storage medium having computer-readable program instructions (i.e., computer programs) stored thereon, and the computer-readable program instructions are used to execute the data center computer room fault diagnosis method based on the knowledge graph in the above-mentioned embodiment.

[0136] The computer-readable storage medium provided in the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems or devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination of the above.

[0137] The above-mentioned computer-readable storage medium may be included in the data center computer room fault diagnosis device based on the knowledge graph; or it may exist independently without being assembled into the data center computer room fault diagnosis device based on the knowledge graph.

[0138] The above-mentioned computer-readable storage medium carries one or more programs. When the above-mentioned one or more programs are executed by the data center computer room fault diagnosis device based on the knowledge graph, the data center computer room fault diagnosis device based on the knowledge graph: obtains historical fault cases of the data center; based on the historical fault cases, builds a knowledge graph according to the entity extraction model, the relationship extraction model and the entity disambiguation model; maps the fault problem input by the user to the knowledge graph, and determines the optimal solution to the fault problem based on the entity linking scheme, the relationship guessing scheme and the multi-path sorting scheme.

[0139] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0140] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0141] The modules involved in the embodiments described in this application may be implemented by software or hardware, wherein the name of the module does not constitute a limitation on the unit itself in some cases.

[0142] The readable storage medium provided by the present application is a computer-readable storage medium, which stores computer-readable program instructions (i.e., computer programs) for executing the above-mentioned data center computer room fault diagnosis method based on knowledge graph, and can solve the technical problem that the traditional fault diagnosis optimization scheme cannot perform focused retrieval on the core problems input by the user, resulting in insufficient accuracy of fault diagnosis optimization. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided by the present application are the same as the beneficial effects of the data center computer room fault diagnosis method based on knowledge graph provided by the above-mentioned embodiment, and will not be elaborated here.

[0143] The present application also provides a computer program product, including a computer program, which, when executed by a processor, implements the steps of the above-mentioned knowledge graph-based data center computer room fault diagnosis method.

[0144] The computer program product provided by the present application can solve the technical problem that the traditional fault diagnosis optimization solution is not accurate enough because it cannot focus on the core problems input by the user. Compared with the prior art, the beneficial effects of the computer program product provided by the present application are the same as the beneficial effects of the data center computer room fault diagnosis method based on the knowledge graph provided in the above embodiment, which will not be repeated here.

[0145] The above descriptions are only some embodiments of the present application, and are not intended to limit the patent scope of the present application. All equivalent structural changes made using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect applications in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A data center computer room fault diagnosis method based on knowledge graph, characterized in that: The method comprises: Obtain historical failure cases of the data center; Based on the historical failure cases, a knowledge graph is constructed according to an entity extraction model, a relationship extraction model, and an entity disambiguation model; The fault problem input by the user is mapped into the knowledge graph, and the optimal solution for the fault problem is determined based on the entity linking scheme, the relationship guessing scheme and the multi-path sorting scheme.

2. The method according to claim 1, characterized in that The step of constructing a knowledge graph based on the historical fault cases according to the entity extraction model, the relationship extraction model and the entity disambiguation model includes: Extracting entity information from the historical fault cases based on the entity extraction model; Based on the relationship extraction model, extracting target entity relationships from the historical fault cases according to the entity information; Based on the entity disambiguation model, performing entity disambiguation processing on the entity information to obtain target entity information; A knowledge graph is constructed based on the target entity information and the target entity relationships.

3. The method according to claim 2, characterized in that The entity extraction model includes a Bert network, a Bi-LSTM network and a global pointer network, and the global pointer network introduces a rotational position encoding; wherein the step of extracting entity information from the historical fault case based on the entity extraction model includes: Inputting the historical fault cases into the entity extraction model, and obtaining the Chinese word vectors of the historical fault cases through the Bert network; Extracting semantic features of the historical fault cases through the Bi-LSTM network; Entity information is extracted from the historical fault cases through the global pointer network according to the Chinese word vector and the semantic features.

4. The method according to claim 2, characterized in that The step of extracting target entity relationships from the historical fault cases based on the relationship extraction model according to the entity information includes: Dividing the entity information into training set data and test set data; Based on the training set data, fine-tune the Bert network of the relationship extraction model to obtain a preset relationship extraction model, and determine the training entity relationships of all entities in the training set data based on the preset relationship extraction model; Inputting the test set data into the preset relationship extraction model to obtain a test entity relationship; Determining a target test entity relationship from the test entity relationships according to a majority voting mechanism; Based on the training entity relationship and the target test entity relationship, a target entity relationship is determined.

5. The method according to claim 2, characterized in that The step of performing entity disambiguation processing on the entity information based on the entity disambiguation model to obtain target entity information comprises: Based on the entity disambiguation model, converting the entity information into vector information; Determine similarity information between the entity to be disambiguated and the candidate entity in the entity information according to the vector information; Determine the best matching entity of the entity to be disambiguated from the candidate entities according to the similarity information; The target entity information is determined according to the best matching entity of each entity to be disambiguated in the entity information.

6. The method according to claim 1, characterized in that The step of mapping the fault problem input by the user to the knowledge graph and determining the optimal solution to the fault problem based on the entity linking scheme, the relationship guessing scheme and the multi-path sorting scheme includes: Based on the entity linking scheme, determine a mapping entity of the fault problem in the knowledge graph; Based on the relationship guessing scheme, determine the mapping entity relationship of the fault problem in the knowledge graph; Based on the mapping entities and the mapping entity relationships, an optimal solution to the fault problem is determined according to the multi-path sorting scheme.

7. The method according to claim 6, characterized in that The step of determining the optimal solution to the fault problem based on the mapping entity and the mapping entity relationship according to the multipath sorting scheme comprises: Determine a relationship path according to the mapping entities and the mapping entity relationships; Determine the occurrence frequency of the relationship path in the knowledge graph according to the mapping entity; Determining the probability of the relationship path according to the occurrence frequency; A target relationship path is determined from the relationship paths according to the probability, and the target relationship path is determined as an optimal solution to the fault problem.

8. A data center computer room fault diagnosis device based on knowledge graph, characterized in that: The device comprises: An acquisition module is used to obtain historical failure cases of the data center; A construction module, used to construct a knowledge graph based on the historical fault cases according to an entity extraction model, a relationship extraction model, and an entity disambiguation model; The determination module is used to map the fault problem input by the user into the knowledge graph, and determine the optimal solution to the fault problem based on the entity linking scheme, the relationship guessing scheme and the multi-path sorting scheme.

9. A data center computer room fault diagnosis device based on knowledge graph, characterized in that: The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the data center computer room fault diagnosis method based on the knowledge graph as described in any one of claims 1 to 7.

10. A storage medium, characterized in that: The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the data center computer room fault diagnosis method based on the knowledge graph are implemented as described in any one of claims 1 to 7.

Images