Fault work order merging method and device and nonvolatile storage medium

Abstract

The invention discloses a fault work order merging method and device and a nonvolatile storage medium. The method comprises the following steps: receiving a fault work order to be merged, and vectorizing the fault work order to obtain a fault work order vector corresponding to the fault work order; determining similar historical fault single combination records corresponding to the fault work single vectors from a fault vector library; determining a target cue word based on the fault work order single vector, the similar historical fault order merging record and the fault work order knowledge graph; and analyzing the target cue word through a large language model to obtain a fault work order merging decision corresponding to the to-be-merged fault work order. The technical problem that fault work order merging efficiency is low due to the fact that fault work order merging depends on manual operation and maintenance personnel in the prior art is solved.

Description

Technical Field

[0001] This application relates to the field of AI technology. Specifically, it relates to a method, device, and non-volatile storage medium for merging fault work orders. Background Art

[0002] In network service operation and maintenance work, due to the complexity of the network, different network types are managed separately. For example, in the case of a power outage in a certain computer room, an interruption of an external optical cable, or a drop of a certain base station, which involves complex network management situations such as multiple network types, multi-layer interconnection, and multiple network management systems, a large number of alarms will occur simultaneously, generating a huge number of fault work orders. Due to the imperfect data and diagrams, a large number of potentially duplicate fault work orders will be dispatched, resulting in a large number of maintenance tasks and affecting the timely repair rate of faults. For the names of computer rooms, equipment names, etc. involved in faults, due to different system categories, the description specifications may be different, such as capitalization, abbreviation, Chinese and English, etc., resulting in the inability to locate the root alarm position. Most of the resource systems rely on manual input to maintain data, which may cause the alarms of different network types to not be associated and converged due to different naming specifications, input errors, long and short descriptions, abbreviation methods, etc. Merging fault work orders can greatly reduce the number of operation and maintenance work orders, more accurately locate the fault position, and significantly improve the timely repair rate; in the related art, after experienced operation and maintenance monitoring personnel manually judge, the fault work orders are merged manually, resulting in low efficiency in processing fault work orders.

[0003] In view of the above problems, no effective solution has been proposed yet. Summary of the Invention

[0004] Embodiments of this application provide a method, device, and non-volatile storage medium for merging fault work orders to at least solve the technical problem in the related art that the efficiency of merging fault work orders is low due to relying on operation and maintenance personnel to manually merge fault work orders.

[0005] According to one aspect of the embodiments of this application, a method for merging fault work orders is provided, including: receiving the fault work orders to be merged, and vectorizing the fault work orders to obtain fault work order vectors corresponding to the fault work orders; determining similar historical fault order merge records corresponding to the fault work order vectors from the fault vector library; determining target prompt words based on the fault work order vectors, similar historical fault order merge records, and the fault work order knowledge graph; analyzing the target prompt words through a large language model to obtain a fault work order merge decision corresponding to the fault work orders to be merged.

[0006] In some embodiments of the present application, before receiving the trouble tickets to be merged, the method further includes: obtaining the historical trouble ticket merging records within a preset time period; performing vectorization processing on the historical trouble ticket merging records to obtain historical trouble ticket merging vectors; storing the historical trouble ticket merging vectors in a trouble ticket vector library, where the trouble ticket vector library is used to store the historical trouble ticket merging vectors corresponding to all historical trouble ticket merging records within the preset time period.

[0007] In some embodiments of the present application, the trouble ticket knowledge graph is a cross-professional knowledge graph. The trouble ticket knowledge graph includes entities and relationships. Each profession corresponds to a network type. Each node in the trouble ticket knowledge graph corresponds to an entity. The edges between entities are represented as relationships, and the relationships are used to associate all entities with the same root cause of the failure. Each entity represents a network device and includes an attribute value, and the attribute value at least includes the entity unique identifier, entity type, and entity failure location.

[0008] In some embodiments of the present application, the trouble ticket knowledge graph is constructed in the following manner: obtaining all historical trouble ticket merging vectors in the trouble ticket vector library; identifying the entity information and relationship information corresponding to each historical trouble ticket merging vector, where the entity information is used to represent the network entity corresponding to the historical trouble ticket merging vector and the attribute information corresponding to the network entity, and the relationship information is used to record all network entities that have a failure association with the network entity; determining the entity information as the entity in the trouble ticket knowledge graph, and associating the entities in the trouble ticket knowledge graph based on the relationship information to obtain the trouble ticket knowledge graph.

[0009] In some embodiments of the present application, vectorizing the trouble ticket to obtain the trouble ticket vector corresponding to the trouble ticket includes: obtaining the multi-dimensional information of the trouble ticket, where the multi-dimensional information at least includes text description, time information, location information, and device information; performing vectorization on the information of each dimension in the multi-dimensional information respectively to obtain the information vector corresponding to the information of each dimension; and merging the information vectors corresponding to the information of each dimension to obtain the trouble ticket vector.

[0010] In some embodiments of the present application, determining the similar historical trouble ticket merging record corresponding to the trouble ticket vector from the trouble ticket vector library includes: determining the similarity score between the trouble ticket vector and each historical trouble ticket merging vector in the trouble ticket vector library; according to the similarity score, determining the historical trouble ticket merging vectors ranked in the top preset number as the reference vector set; determining the historical location information corresponding to each historical trouble ticket merging vector in the reference vector set; determining the historical trouble ticket merging vectors with the matching degree between the historical location information and the location information of the trouble ticket greater than the preset location threshold as the target vectors; and determining the historical trouble ticket merging records corresponding to the target vectors as the similar historical trouble ticket merging records.

[0011] In some embodiments of the present application, determining a target prompt word based on a trouble ticket vector, a merged record of similar historical trouble tickets, and a trouble ticket knowledge graph includes: retrieving entities and relationships corresponding to the trouble ticket vector from the trouble ticket knowledge graph; obtaining the relevance between the trouble ticket and the merged record of similar historical trouble tickets to form a logical chain; integrating the merged record of similar historical trouble tickets and the entities and relationships corresponding to the trouble ticket vector into an initial text based on the logical chain; adjusting the text structure of the initial text, and generating a target prompt word through a predefined text template, where the target prompt word at least includes a trouble description, a trouble occurrence time, a trouble occurrence location, and a historical merging example, and the historical merging example is used to indicate the trouble ticket information before merging, the historical merging result, and the historical merging basis recorded in the merged record of similar historical trouble tickets.

[0012] In some embodiments of the present application, the trouble ticket merging decision includes a trouble ticket merging result, a trouble ticket merging basis, and a confidence score, where the confidence score is used to represent the accuracy of the trouble ticket merging result.

[0013] According to another aspect of the embodiments of the present application, there is also provided a trouble ticket merging device, including: a receiving module, configured to receive trouble tickets to be merged, and vectorize the trouble tickets to obtain a trouble ticket vector corresponding to the trouble tickets; a first determination module, configured to determine a merged record of similar historical trouble tickets corresponding to the trouble ticket vector from a trouble vector library; a second determination module, configured to determine a target prompt word based on the trouble ticket vector, the merged record of similar historical trouble tickets, and the trouble ticket knowledge graph; and an analysis module, configured to analyze the target prompt word through a large language model to obtain a trouble ticket merging decision corresponding to the trouble tickets to be merged.

[0014] According to another aspect of the embodiments of the present application, there is also provided a non-volatile storage medium, in which a program is stored, and when the program runs, it controls a device where the non-volatile storage medium is located to execute the trouble ticket merging method of any one of the above.

[0015] According to another aspect of the embodiments of the present application, there is also provided an electronic device, including: a memory and a processor, where the processor is configured to run a program stored in the memory, and when the program runs, it executes the trouble ticket merging method of any one of the above.

[0016] According to another aspect of the embodiments of the present application, there is also provided a computer program product, including computer instructions, and when the computer instructions are executed by a processor, they implement the trouble ticket merging method of any one of the above.

[0017] In the embodiments of the present application, a fault work order to be merged is received, and the fault work order is vectorized to obtain a fault work order vector corresponding to the fault work order; a similar historical fault order merging record corresponding to the fault work order vector is determined from a fault vector library; a target prompt word is determined based on the fault work order vector, the similar historical fault order merging record, and a fault work order knowledge graph; and a large language model is used to analyze the target prompt word to obtain a fault work order merging decision corresponding to the fault work order to be merged. By vectorizing the fault work order to be merged, then determining a similar historical fault order merging record corresponding to the fault work order vector from the fault vector library, determining a target prompt word based on the fault work order vector, the similar historical fault order merging record, and the fault work order knowledge graph, and finally generating a fault work order merging decision through the large language model, the purpose of automatically generating a fault work order merging decision through the large language model is achieved, thereby solving the technical problem in the related art that the efficiency of fault work order merging is low due to relying on operation and maintenance personnel to manually merge fault work orders. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The drawings described herein are used to provide a further understanding of the present application and form a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation to the present application. In the drawings:

[0019] Figure 1 is a hardware structure block diagram of a computer terminal for implementing a method for merging fault work orders according to an embodiment of the present application;

[0020] Figure 2 is a flowchart of a method for merging fault work orders according to an embodiment of the present application;

[0021] Figure 3 is a flowchart of another method for merging fault work orders according to an embodiment of the present application;

[0022] Figure 4 is a structural schematic diagram of a device for merging fault work orders according to an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0023] In order to enable those skilled in the art to better understand the solution of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0024] The information collected in the embodiments of this application is information and data authorized by the user or fully authorized by all parties. For the processing of relevant data, such as collection, storage, use, processing, transmission, provision, disclosure, and application, all comply with the relevant laws, regulations, and standards of the relevant regions, adopt necessary confidentiality measures, do not violate public order and good customs, and provide corresponding operation entrances for users to choose to authorize or reject the automated decision-making results; if the user chooses to reject, the expert decision-making process will be entered.

[0025] It should be noted that the terms "first", "second", etc. in the specification, claims, and above-mentioned drawings of this application are used to distinguish similar objects and do not necessarily describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of this application described here can be implemented in an order other than those illustrated or described here. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products, or devices.

[0026] To better understand the embodiments of this application, the following technical terms involved in the embodiments of this application are explained as follows:

[0027] LLM large model (Large Language Model, abbreviated as LLM, also known as large language model): It is a dense artificial intelligence model based on transformers, capable of understanding the semantics of complex natural language processing (Natural Language Processing, abbreviated as NLP), and can complete various NLP tasks through the method of prompt engineering using a single model. The parameter scale of the LLM is huge, containing billions to hundreds of billions of parameters, and can effectively extract the features of long semantic contexts for task execution and solution.

[0028] Chain-of-Thought (abbreviated as CoT) is an improved prompt text technology aimed at enhancing the ability of large models in complex reasoning tasks. Powerful logical reasoning enables the AI model to break down complex problems into multiple simple tasks and solve them step by step, thus generally solving complex problems, such as arithmetic reasoning, commonsense reasoning, and symbolic reasoning. CoT enhances the arithmetic, commonsense, and reasoning abilities of large models by requiring the model to explicitly output the intermediate reasoning steps before outputting the final answer.

[0029] Few-shot Learning is an application of Meta Learning in the field of supervised learning. The Meta Learning algorithm aims to enable the model to learn "learning", capable of handling tasks of similar types, rather than just a single classification task.

[0030] In related technologies, due to incomplete data and graphs, a large number of possibly duplicate fault work orders are dispatched, resulting in a large number of maintenance tasks and affecting the timeliness rate of fault repair. Merging fault work orders can greatly reduce the number of operation and maintenance work orders and more accurately locate the fault location, which can significantly improve the timeliness rate of repair. In related technologies, after experienced operation and maintenance monitoring personnel manually judge, the fault work orders are merged manually, resulting in low efficiency in processing fault work orders. Therefore, there is a technical problem in related technologies that the merging of fault work orders depends on the manual operation of operation and maintenance personnel, resulting in low efficiency of merging fault work orders. To solve this problem, relevant solutions are provided in the embodiments of the present application, which are described in detail below.

[0031] According to the embodiments of the present application, an embodiment of a method for merging fault work orders is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. And, although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0032] The method embodiments provided by the embodiments of the present application can be executed in a computer terminal or a similar computing device. Figure 1 A hardware structure block diagram of a computer terminal for implementing a method for merging fault work orders is shown. As Figure 1 shown, the computer terminal 10 may include one or more (shown as 102a, 102b,..., 102n in the figure) processors 102 (the processor 102 may include, but is not limited to, a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 104 for storing data, and a transmission device 106 for communication functions. In addition, it may further include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the BUS bus), a network interface, a power supply, and / or a camera. Those of ordinary skill in the art can understand that Figure 1 the structure shown is only schematic and does not limit the structure of the above-mentioned electronic device. For example, the computer terminal 10 may further include more or fewer components than Figure 1 shown, or have a different configuration from Figure 1 shown.

[0033] It should be noted that one or more of the above-mentioned processors 102 and / or other data processing circuits can generally be referred to as "data processing circuits" herein. The data processing circuit can be embodied in software, hardware, firmware, or any combination thereof, in whole or in part. In addition, the data processing circuit can be a single independent processing module, or be incorporated in whole or in part into any one of the other components in the computer terminal 10. As involved in the embodiments of the present application, the data processing circuit is a kind of processor control (such as the selection of a variable resistance terminal path connected to an interface).

[0034] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the method for merging trouble tickets in the embodiments of the present application. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, that is, implements the above-mentioned method for merging trouble tickets. The memory 104 can include high-speed random access memory, and can also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memories. In some instances, the memory 104 can further include a memory remotely set relative to the processor 102, and these remote memories can be connected to the computer terminal 10 through a network. Examples of the above-mentioned network include but are not limited to the Internet, intranet, local area network, mobile communication network, and combinations thereof.

[0035] The transmission device 106 is used to receive or send data via a network. Specific examples of the above-mentioned network can include the wireless network provided by the communication provider of the computer terminal 10. In one instance, the transmission device 106 includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices through a base station and thus communicate with the Internet. In one instance, the transmission device 106 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0036] The display can be, for example, a touch-screen liquid crystal display (LCD), which enables the user to interact with the user interface of the computer terminal 10.

[0037] Under the above operating environment, the embodiments of the present application provide an embodiment of a method for merging trouble tickets. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0038] As Figure 2As shown in the figure, it is a flowchart of a method for merging fault work orders provided according to an embodiment of the present application, including:

[0039] Step S202: Receive the fault work orders to be merged, and vectorize the fault work orders to obtain the fault work order vectors corresponding to the fault work orders.

[0040] In the technical solution provided in step S202, before receiving the fault work orders to be merged, obtain the historical fault work order merging records within a preset time period; perform vectorization processing on the historical fault work order merging records to obtain historical fault work order merging vectors; store the historical fault work order merging vectors in a fault vector library, where the fault vector library is used to store the historical fault work order merging vectors corresponding to all historical fault work order merging records within a preset time period.

[0041] There are various implementation methods for vectorizing the fault work orders to obtain the fault work order vectors corresponding to the fault work orders in the above steps. For example: obtain the multi-dimensional information of the fault work orders, where the multi-dimensional information at least includes text description, time information, location information, and device information; perform vectorization on the information in each dimension of the multi-dimensional information respectively to obtain the information vectors corresponding to the information in each dimension; merge the information vectors corresponding to the information in each dimension to obtain the fault work order vectors.

[0042] The following are specific embodiments:

[0043] Obtain the historical merged records of fault tickets within a preset time period (e.g., the past three years). The historical merged records of fault tickets are the records of successful manual ticket merging within the preset time period. The historical merged records of fault tickets include at least the basic attributes of the merged fault tickets (e.g., number, creation time, status, etc.); the entities involved in the merged fault work orders (e.g., servers, network devices, etc.) and their attributes, as well as the historical records of similar fault work orders that have been resolved. Preprocess the collected historical merged records of fault tickets, clean up irrelevant information, correct format errors, and ensure data consistency and quality. Extract key multi-dimensional information from each historical merged record of fault tickets, including the text description, time information, geographical location information, device information, etc. of the fault work order. Segment the extracted multi-dimensional information and perform vectorization processing on each part of the information (such as fault description, device name) separately. For example, use a text embedding technology model to encode the device name, fault description, and geographical location information respectively to generate corresponding vector representations. Integrate the vectors of each part of each historical merged record of fault tickets to form a historical merged vector of fault tickets to represent the semantic features of the entire historical record. Store the historical merged vector of fault tickets in a fault vector library (or called a vector database). A fault vector library is a data structure specifically for storing and managing point sets in a high-dimensional space. It can perform approximate nearest neighbor search quickly, which is very useful for many machine learning tasks. In the intelligent judgment of fault work order merging, each historical fault work order in the historical merged record of fault tickets can be represented as a vector and stored in the fault vector library. In this way, when it is necessary to find similar fault tickets, we can use the vector database for efficient retrieval.

[0044] When a large number of faults in the same area occur simultaneously, triggering the dispatch of multiple fault work orders, first, receive the fault work orders to be merged. The number of fault work orders to be merged is at least two. For the fault work orders to be merged, obtain multi-dimensional information of the fault work orders. Among them, the multi-dimensional information at least includes text description (i.e., fault description, the specific text description of the fault of the entity involved in the fault order), time information (the occurrence time of the fault), location information (the occurrence location of the fault), and equipment information (the entity where the fault occurs). For the extraction of text description, natural language processing techniques such as word segmentation, part-of-speech tagging, and named entity recognition (NER) can be used to extract key information and entities from the fault description. For the extraction of time information, the timestamp can be converted into a standardized format, timestamp or a specific date-time format. For the extraction of location information, the geographic information system (GIS) or address resolution technology can be used to convert the geographical location in the text into longitude and latitude coordinates or geocoding. For the extraction of equipment information, the equipment name can be standardized to ensure that different representations of the same equipment can be recognized as the same entity.

[0045] Use pre-trained deep learning models such as Bidirectional Encoder Representations from Transformers (BERT), General Language Model (GLM), etc., to convert the text description into a vector representation. It is achieved through the embedding method in time series analysis. For example, use a recurrent neural network (RNN) or a long short-term memory network (LSTM) to encode the time information into a vector. The geographical location information can be converted into a vector using geographical coordinate embedding or a specific location embedding model. For the vectorization of equipment information, the word embedding method can be used.

[0046] Merge the vectors obtained from each of the above dimensions into a single vector through methods such as weighted average and concatenation as the fault work order vector. At the same time, to ensure the storage and search efficiency of the vector in the vector database, it is necessary to standardize the merged vector so that it has a fixed length and range.

[0047] Step S204, determine the similar historical fault order merge records corresponding to the fault work order vector from the fault vector library.

[0048] In the technical solution provided in step S204, there are various implementation manners for determining the similar historical fault order merge record corresponding to the fault order vector from the fault vector library. For example: determining the similarity score between the fault order vector and each historical fault order merge vector in the fault vector library; according to the similarity score, determining the historical fault order merge vectors ranked in the top preset number of digits as the reference vector set; determining the corresponding historical location information of each historical fault order merge vector in the reference vector set; determining the historical fault order merge vector whose matching degree between the historical location information and the location information of the fault order is greater than the preset location threshold as the target vector; and determining the historical fault order merge record corresponding to the target vector as the similar historical fault order merge record.

[0049] The following are specific embodiments:

[0050] First, establish a connection with the fault vector library to prepare for vector retrieval. Determine the similarity score between the fault order vector and each historical fault order merge vector in the fault vector library (for example, calculate the similarity score using the L2 distance similarity (L2 Distance Similarity, or also known as the Euclidean distance). The L2 distance is the most commonly used Euclidean distance calculation method in the vector space and can effectively measure the similarity between two vectors). Sort the historical vectors according to the calculated similarity score, and select the historical fault order merge vectors ranked in the top preset number of digits (such as 10) as the reference vector set. For the reference vector set, further determine the corresponding historical location information of each historical fault order merge vector in the reference vector set (ensure that the format of the extracted historical location information is unified, such as converting it to latitude and longitude coordinates or a specific geocoding format). Determine the historical fault order merge vector whose matching degree between the historical location information and the location information of the fault order (for example, the matching degree is the distance between the historical location information and the location information of the fault order) is greater than the preset location threshold (for example, 500 meters) as the target vector; and determine the historical fault order merge record corresponding to the target vector as the similar historical fault order merge record.

[0051] Another alternative implementation manner for determining the similar historical fault order merge record corresponding to the fault order vector from the fault vector library is that for the fault orders to be merged containing multiple NER targets, that is, the fault orders involving multiple entities (such as different devices or computer rooms), use these fault orders to be merged as alternative targets, obtain the fault order vectors of these fault orders to be merged, calculate the similarity score between the fault order vector and each historical fault order merge vector in the fault vector library respectively, sort the historical vectors according to the calculated similarity score, and select the historical fault order merge vectors ranked in the top preset number of digits (such as 3 - 5) as the final vector set. Determine the historical fault order merge record corresponding to the final vector set as the similar historical fault order merge record.

[0052] Step S206: Determine the target prompt word based on the fault work order vector, the merged record of similar historical fault work orders, and the fault work order knowledge graph.

[0053] In the technical solution provided in step S206, the fault work order knowledge graph is a cross - professional knowledge graph. The fault work order knowledge graph contains entities and relationships. Among them, each profession corresponds to a network type, such as the core network, the transmission network, the access network, the data center (responsible for the maintenance of the network data center), and so on. Each node in the fault work order knowledge graph corresponds to an entity (specifically referring to a historical fault work order). The edges between entities are represented as relationships. The relationships are used to associate all entities with the same root cause of the fault. Each entity represents a network device and contains an attribute value. The attribute value at least includes the entity unique identifier (i.e., the number), the entity type, and the entity fault location. The fault work order knowledge graph is constructed in the following way: Obtain all the merged vectors of historical fault work orders in the fault vector library; Identify the entity information and relationship information corresponding to each historical fault work order merged vector. Among them, the entity information is used to represent the network entity corresponding to the historical fault work order merged vector (i.e., the specific network device that appears in the fault work order) and the attribute information corresponding to the network entity (specifically the attribute information of the network entity in the historical fault work order). The relationship information is used to record all network entities that have a fault association with the network entity; Determine the entity information as the entity in the fault work order knowledge graph, and based on the relationship information, associate the entities in the fault work order knowledge graph to obtain the fault work order knowledge graph.

[0054] The following are specific embodiments:

[0055] Obtain all the merged vectors of historical fault work orders in the fault vector library, and identify the entity information and relationship information corresponding to each historical fault work order merged vector. The entity information may include geographical location, site (such as base station, router, etc.), computer room: the physical storage location of the network device. Network device number, signal: network signal or service. The relationship information may include interconnection relationship: the direct connection between network entities in different historical fault work orders, such as the optical fiber connection between network devices. Carrying relationship: One network entity carries the service or signal of another network entity. Fault association: Record which network entities are associated with each other in a specific fault event, for example, multiple devices trigger alarms simultaneously due to the same power outage.

[0056] All the merged vectors of historical trouble tickets in the trouble ticket vector library involve troubles in different specialties. The trouble ticket knowledge graph is a structured knowledge representation form that can be used to describe the relationships between entities. For example, if two trouble tickets (i.e., two entities) both involve the same server, an edge can be added between them. In this way, the connections between all network entities can be understood as a whole, and potential trouble ticket merging opportunities can be discovered. The knowledge graph can be intelligently generated using the graph database Neo4j (abbreviated as "Neo4j Graph Database" or simply "Neo4j"). Neo4j is a high-performance graph database system that focuses on storing and querying graph data structures, namely the relationship network between entities. For example: Entity modeling: Define the identified entity information as node types in Neo4j, such as "city", "device", "signal". Attribute addition: Add attribute information to each node, such as the model of the device, the location of the computer room, the status of the signal, etc. Relationship modeling: Define the relationship types between entities, such as "located in", "carrying", "associated with", etc., and establish these relationships in the graph.

[0057] It should be noted that in the above steps, there are multiple ways to determine the target prompt word based on the trouble ticket vector, the merged records of similar historical trouble tickets, and the trouble ticket knowledge graph. For example: Retrieve the entities and relationships corresponding to the trouble ticket vector from the trouble ticket knowledge graph; Obtain the relevance between the trouble ticket and the merged records of similar historical trouble tickets to form a logical chain; Integrate the merged records of similar historical trouble tickets and the entities and relationships corresponding to the trouble ticket vector into an initial text based on the logical chain; Adjust the text structure of the initial text and generate the target prompt word through a predefined text template, where the target prompt word at least includes the trouble description, the trouble occurrence time, and the trouble occurrence location.

[0058] The following are specific embodiments:

[0059] Retrieve the entities and relationships corresponding to the fault work order vector from the fault work order knowledge graph. Specifically: In the Neo4j graph database, use query statements to retrieve relevant entity nodes and the relationships between them based on the entity information in the vector. For example, if the fault work order represented by the fault work order vector is related to "xxx Town Computer Room 1", the query will include all entities and relationships related to this computer room, such as associated network device numbers, signal status, and the inclusion, interconnection, and bearing relationships between them. Analyze the merged records of similar historical fault work orders, and extract the fault work order information before merging (such as fault description, occurrence time, location), historical merge results, and historical merge bases as historical merge examples. Compare the current work order with the historical records to find potential correlation points, such as the same or similar fault descriptions, time windows, geographical locations, involved network entities and their relationships. Through comparison, a logical chain can be formed, for example: "fault description similarity, geographical location proximity, time continuity, device interconnection or bearing relationship". When the logical chain is satisfied, integrate the corresponding entities and relationships of the fault work order with the merged records of similar historical fault work orders in an initial text, including the fault description, occurrence time, location, and examples and bases of historical work order merging. Clearly indicate the key points of the logical chain in the historical merge record in the initial text, such as "in the same computer room Y as work order X and with similar fault types, work order X was successfully merged, based on Z".

[0060] Pre-define a structured text template to guide how to construct the target prompt word, ensuring the integrity and logical clarity of the information. Use the text template to format and adjust the initial text, reorganize the fault description, fault time, location, and historical merge examples according to the template requirements, ensuring that each piece of information has a clear identifier and location, and finally obtain the target prompt word.

[0061] Example of the target prompt word:

[0062] Fault description of the current fault work order to be merged; occurrence time; fault occurrence time, fault occurrence location;

[0063] Task: Please use the above information to determine whether the current fault work order can be merged with the work orders in the historical records. If it can, please indicate the logical basis for the merge; if not, please explain the reason. Please return the result in JSON format, including the fault work order merge result (result), confidence score (score), and fault work order merge basis (reasoning_chain).

[0064] Step S208, analyze the target prompt word through a large language model to obtain the fault work order merge decision corresponding to the fault work order to be merged.

[0065] In the technical solution provided in step S208, the fault work order merging decision includes the fault work order merging result, the basis for merging the fault work orders, and the confidence score, where the confidence score is used to indicate the accuracy of the fault work order merging result.

[0066] The following are specific embodiments:

[0067] Input the target prompt into the large language model, and the large language model analyzes the target prompt to obtain the fault work order merging decision corresponding to the fault work order to be merged. For example, the fault work order merging decision corresponding to the above target prompt example is: fault work order merging result (result): mergable, confidence score (score): 0.9 (on a 1-point scale), basis for merging the fault work orders (reasoning_chain): The work orders to be merged all belong to the fault work orders under xx network management, and the time difference between the fault times is less than 1 minute, and the geographical locations involved are all related to xx Town.

[0068] The large language model (referring to a deep neural network model with a large number of parameters) is a pre-trained large language model, and the large language model is trained in the following way:

[0069] Divide the historical fault work order merging records into a training set and a test set according to a preset ratio (such as 8:2), and determine historical merging examples based on the training set. Among them, the historical merging examples are used to indicate the fault work order information before merging, the historical merging result, and the basis for historical merging recorded in the historical fault work order merging records. One historical merging example:

[0070] Historical merging result: The following 3 historical fault work orders are judged to be mergable.

[0071] Basis for historical merging: The basis is the name logical relationship and the fault time interval within 1 minute.

[0072] Fault work order information before merging:

[0073] Emergency alarm 1 (network device number 1), alarm description 1, location 1, fault time 1;

[0074] Emergency alarm 2 (network device number 2), alarm description 2, location 2, fault time 2;

[0075] Emergency alarm 3 (network device number 3), alarm description 3, location 3, fault time 3.

[0076] Use the historical merging examples to train the initial large language model until the preset number of iterations is reached and stop training to obtain the large language model, and verify the large language model on the validation set. In addition, obtain the historical fault work order merging records of the second preset time period at intervals of the preset time and re-iterate and update the large language model to achieve the optimization of the large language model.

[0077] Through the above steps, by introducing vectorization processing and large language models, a large number of fault work orders can be efficiently analyzed and processed, similar faults can be automatically identified and merger decisions can be made, significantly improving the efficiency and accuracy of fault work order processing. By using a cross-professional knowledge graph, fault information can be associated across domains, providing a more comprehensive perspective for fault analysis, which helps to quickly locate and resolve network faults. In addition, the introduction of confidence scores makes the decision-making process more transparent and traceable, enhancing the reliability of the system and user trust. Overall, this solution can effectively improve the intelligent level of fault work order processing, reduce manual intervention, speed up the fault response speed, and improve network operation and maintenance efficiency.

[0078] As Figure 3 shown, it is a flowchart of another method for merging fault work orders provided by an embodiment of the present application. As shown in the figure, the entire process of merging fault work orders is divided into two stages, namely the offline stage and the online stage. In the offline stage, first, historical manual order merging records (i.e., the above-mentioned historical fault order merging records) are obtained, sliced and embedded vector databases are stored (i.e., the above-mentioned historical fault order merging records are vectorized to obtain historical fault order merging vectors; the historical fault order merging vectors are stored in the fault vector database), and a network knowledge graph is constructed (i.e., the fault work order knowledge graph is constructed in the following manner: all historical fault order merging vectors in the fault vector database are obtained; the entity information and relationship information corresponding to each historical fault order merging vector are identified, where the entity information is used to represent the network entity corresponding to the historical fault order merging vector and the attribute information of the network entity, and the relationship information is used to record all network entities that have a fault association with the network entity; the entity information is determined as an entity in the fault work order knowledge graph, and the entities in the fault work order knowledge graph are associated based on the relationship information to obtain the fault work order knowledge graph). Then comes the online stage. First, a list of work orders to be merged and judged (i.e., the above-mentioned fault work orders to be merged) is received, and then Prompt engineering is performed to integrate the historical merging records queried from the vector database, the network knowledge graph (i.e., the above-mentioned fault work order knowledge graph), and the existing work order records to be judged (i.e., the target prompt word is determined based on the fault work order vector, similar historical fault order merging records, and the fault work order knowledge graph). Finally, the judgment result of the large model (i.e., the above-mentioned large language model analyzes the target prompt word to obtain the fault work order merging decision corresponding to the fault work order to be merged).

[0079] An embodiment of the present application also provides a device for merging fault work orders, as Figure 4 shown, including:

[0080] A receiving module 402, configured to receive fault work orders to be merged and vectorize the fault work orders to obtain fault work order vectors corresponding to the fault work orders.

[0081] The first determination module 404 is configured to determine a similar historical fault order merging record corresponding to the fault work order vector from the fault vector library;

[0082] The first determination module 404 is further configured to determine the similarity score between the fault work order vector and each historical fault order merging vector in the fault vector library; according to the similarity score, determine the historical fault order merging vectors ranked in the top preset number as the reference vector set; determine the corresponding historical position information of each historical fault order merging vector in the reference vector set; determine the historical fault order merging vectors whose matching degree between the historical position information and the position information of the fault work order is greater than the preset position threshold as the target vectors; and determine the historical fault order merging records corresponding to the target vectors as the similar historical fault order merging records.

[0083] The second determination module 406 is configured to determine a target prompt word based on the fault work order vector, the similar historical fault order merging record, and the fault work order knowledge graph;

[0084] The second determination module 406 is further configured to retrieve the entities and relationships corresponding to the fault work order vector from the fault work order knowledge graph; obtain the relevance between the fault work order and the similar historical fault order merging record to form a logical chain; integrate the similar historical fault order merging record and the entities and relationships corresponding to the fault work order vector into an initial text based on the logical chain; adjust the text structure of the initial text, and generate a target prompt word through a predefined text template, where the target prompt word at least includes a fault description, a fault occurrence time, a fault occurrence location, and a historical merging example, and the historical merging example is used to indicate the fault work order information before merging, the historical merging result, and the historical merging basis recorded in the similar historical fault order merging record.

[0085] The analysis module 408 is configured to analyze the target prompt word through a large language model to obtain a fault work order merging decision corresponding to the fault work order to be merged.

[0086] It should be noted that Figure 4 the shown fault work order merging device is used to execute Figure 2 the shown fault work order merging method, so Figure 2 the relevant explanations in the fault work order merging method in

[0087] It should be noted that each module in the above fault work order merging device can be a program module (for example, a set of program instructions implementing a specific function), or a hardware module. For the latter, it can be presented in the following forms, but not limited to: the presentation form of each of the above modules is a processor, or the functions of each of the above modules are implemented by a processor.

[0088] The embodiments of the present application also provide a non-volatile storage medium. The non-volatile storage medium includes a stored program, wherein when the program runs, it controls the device where the non-volatile storage medium is located to execute the above-mentioned method for merging trouble tickets. For example, receiving trouble tickets to be merged, and vectorizing the trouble tickets to obtain trouble ticket vectors corresponding to the trouble tickets; determining similar historical trouble ticket merging records corresponding to the trouble ticket vectors from the trouble vector library; determining target prompt words based on the trouble ticket vectors, similar historical trouble ticket merging records, and the trouble ticket knowledge graph; analyzing the target prompt words through a large language model to obtain a trouble ticket merging decision corresponding to the trouble tickets to be merged.

[0089] The embodiments of the present application also provide an electronic device. The electronic device includes a processor, and the processor is used to run a program, wherein when the program runs, it executes the above-mentioned method for merging trouble tickets. For example, receiving trouble tickets to be merged, and vectorizing the trouble tickets to obtain trouble ticket vectors corresponding to the trouble tickets; determining similar historical trouble ticket merging records corresponding to the trouble ticket vectors from the trouble vector library; determining target prompt words based on the trouble ticket vectors, similar historical trouble ticket merging records, and the trouble ticket knowledge graph; analyzing the target prompt words through a large language model to obtain a trouble ticket merging decision corresponding to the trouble tickets to be merged.

[0090] According to another aspect of the embodiments of the present application, there is also provided a computer program product, including a computer program, which when executed by a processor, implements the above-mentioned method for merging trouble tickets. For example, receiving trouble tickets to be merged, and vectorizing the trouble tickets to obtain trouble ticket vectors corresponding to the trouble tickets; determining similar historical trouble ticket merging records corresponding to the trouble ticket vectors from the trouble vector library; determining target prompt words based on the trouble ticket vectors, similar historical trouble ticket merging records, and the trouble ticket knowledge graph; analyzing the target prompt words through a large language model to obtain a trouble ticket merging decision corresponding to the trouble tickets to be merged.

[0091] In the above embodiments of the present application, the descriptions of the respective embodiments have their own emphases. For parts not detailed in a certain embodiment, reference may be made to the relevant descriptions of other embodiments.

[0092] In several embodiments provided by this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are merely illustrative. For example, the division of the units can be a logical function division. In actual implementation, there can be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection between each other can be through some interfaces. The indirect coupling or communication connection of units or modules can be in an electrical or other form.

[0093] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0094] In addition, in each embodiment of this application, the functional units can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above-mentioned integrated units can be implemented in the form of hardware or in the form of software functional units.

[0095] If the above-mentioned integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of this application, in essence, or the part that contributes to the related technology, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to enable a computer device (which can be a personal computer, a server or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of this application. The foregoing storage medium includes: various media such as USB flash drives, read-only memories (ROMs), random access memories (RAMs), mobile hard disks, magnetic disks or optical discs that can store program codes.

[0096] The above is only the preferred embodiment of this application. It should be noted that for those of ordinary skill in the art, without departing from the principle of this application, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of this application.

Claims

1. A method for merging fault work orders, characterized in that: include: Receive the fault work order to be merged, and vectorize the fault work order to obtain a fault work order vector corresponding to the fault work order; Determine from the fault vector library the similar historical fault ticket merge record corresponding to the fault work order vector; Determine a target prompt word based on the fault ticket vector, the merged record of similar historical fault tickets, and the fault ticket knowledge graph; The target prompt word is analyzed by a large language model to obtain a fault ticket merging decision corresponding to the fault ticket to be merged.

2. The method according to claim 1, characterized in that Before receiving the fault work order to be merged, the method further includes: Get the merged record of historical fault tickets within a preset time period; Vectorizing the merged record of the historical fault tickets to obtain a merged vector of the historical fault tickets; The historical fault ticket merge vector is stored in a fault vector library, wherein the fault vector library is used to store the historical fault ticket merge vectors corresponding to all historical fault ticket merge records in the preset time period.

3. The method according to claim 2, characterized in that The fault ticket knowledge graph is a cross-professional knowledge graph, and the fault ticket knowledge graph contains entities and relationships, wherein each profession corresponds to a network type, each node in the fault ticket knowledge graph corresponds to an entity, and the edges between entities are represented as relationships, and the relationships are used to associate all entities with the same root cause of the fault. Each entity represents a network device and contains an attribute value, and the attribute value includes at least the entity unique identifier, entity type, and entity fault location.

4. The method according to claim 3, characterized in that The fault ticket knowledge graph is constructed in the following way: Obtain all historical fault ticket merge vectors in the fault vector library; Identify entity information and relationship information corresponding to each historical fault ticket merge vector, wherein the entity information is used to represent the network entity corresponding to the historical fault ticket merge vector and the attribute information corresponding to the network entity, and the relationship information is used to record all network entities associated with the network entity; The entity information is determined as an entity in the fault work order knowledge graph, and the entities in the fault work order knowledge graph are associated based on the relationship information to obtain the fault work order knowledge graph.

5. The method according to claim 1, characterized in that Vectorizing the fault work order to obtain a fault work order vector corresponding to the fault work order includes: Acquire multi-dimensional information of the fault work order, wherein the multi-dimensional information includes at least text description, time information, location information, and device information; Vectorizing the information of each dimension in the multi-dimensional information to obtain an information vector corresponding to the information of each dimension; The information vectors corresponding to the information of each dimension are merged to obtain the fault work order vector.

6. The method according to claim 5, characterized in that The step of determining, from a fault vector library, a similar historical fault ticket merge record corresponding to the fault work order vector, comprises: Determine a similarity score between the fault work order vector and each historical fault order merged vector in the fault vector library; According to the similarity score, the merged vectors of the historical fault tickets ranked at the top of the preset number of digits are determined as the reference vector set; Determine the historical location information corresponding to each historical fault ticket merge vector in the reference vector set; Determine a historical fault ticket merge vector whose matching degree between the historical location information and the location information of the fault work ticket is greater than a preset location threshold as a target vector; The historical fault ticket merged record corresponding to the target vector is determined as the similar historical fault ticket merged record.

7. The method according to claim 3, characterized in that The determining of the target prompt word based on the fault ticket vector, the merged record of similar historical fault tickets, and the fault ticket knowledge graph includes: Retrieving entities and relationships corresponding to the fault work order vector from the fault work order knowledge graph; Obtaining the correlation between the fault work order and the merged records of similar historical fault orders to form a logical chain; Based on the logic chain, the similar historical fault ticket merge records and entities and relationships corresponding to the fault ticket vector are integrated into an initial text; The text structure of the initial text is adjusted, and target prompt words are generated through a predefined text template, wherein the target prompt words at least include fault description, fault occurrence time, fault occurrence location and historical merging examples, wherein the historical merging examples are used to indicate the fault work order information before merging, the historical merging results and the historical merging basis recorded in the similar historical fault ticket merging record.

8. The method according to claim 1, characterized in that The fault ticket merging decision includes a fault ticket merging result, a fault ticket merging basis and a confidence score, wherein the confidence score is used to indicate the accuracy of the fault ticket merging result.

9. A device for merging fault work orders, characterized in that: include: A receiving module, used for receiving the fault work order to be merged, and vectorizing the fault work order to obtain a fault work order vector corresponding to the fault work order; A first determination module is used to determine a similar historical fault ticket merge record corresponding to the fault work order vector from a fault vector library; A second determination module is used to determine a target prompt word based on the fault work order vector, the merged record of similar historical fault orders, and the fault work order knowledge graph; The analysis module is used to analyze the target prompt word through a large language model to obtain a fault ticket merging decision corresponding to the fault ticket to be merged.

10. A non-volatile storage medium, characterized in that: The non-volatile storage medium stores a program, wherein when the program is running, the device where the non-volatile storage medium is located is controlled to execute the method for merging fault work orders as described in any one of claims 1 to 8.

11. An electronic device, characterized in that: include: A memory and a processor, wherein the processor is used to run a program stored in the memory, wherein the program, when running, executes the method for merging fault work orders as described in any one of claims 1 to 8.

12. A computer program product comprising computer instructions, characterized in that: When the computer instructions are executed by a processor, the method for merging fault work orders described in any one of claims 1 to 8 is implemented.

Images