Network fault solution generation method and device, computer equipment, storage medium and program product

Abstract

The invention relates to a network fault solution generation method and device, computer equipment, a storage medium and a computer program product. The method comprises the following steps: acquiring alarm information of a network system, inputting the alarm information into a first large language model, and determining a fault network segment in network segments through the first large language model; performing fault analysis on the alarm information by adopting each fault analysis model corresponding to the fault network segment to obtain network fault analysis information output by each fault analysis model; inputting the network fault analysis information output by each fault analysis model into a second large language model, and generating a network fault analysis result through the second large language model; and under the condition that the network fault analysis result represents that the alarm information is real alarm information, generating a network fault solution based on the network fault analysis result through a second large language model. By adopting the method, the fault processing efficiency of the network system can be improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a method, apparatus, computer equipment, storage medium and computer program product for generating a network fault solution. Background Art

[0002] In intelligent network operation and maintenance scenarios, traditional AI models are often used to build network optimization systems for identifying network faults and providing solutions, and to provide network optimization solutions for network systems. However, since this network optimization system only uses traditional AI models, it can only provide a general network optimization direction based on the network detection results provided by the AI ​​model, and cannot provide detailed solutions, resulting in the inability to efficiently respond to faults that are about to occur or have already occurred in the network system, and thus the inability to maintain the network system in a timely manner.

[0003] Therefore, there is a problem in the traditional technology that the fault handling efficiency of the network system is low. Summary of the invention

[0004] Based on this, it is necessary to provide a method, device, computer equipment, computer-readable storage medium and computer program product for generating a network fault solution that can improve the efficiency of fault handling of the network system in response to the above technical problems.

[0005] A method for generating a network fault solution is applied to a network system, the network system includes various network segments, and the method includes:

[0006] Acquire alarm information of the network system, and input the alarm information into the first large language model, and determine the faulty network segment in each network segment through the first large language model; the first large language model is obtained by pre-training a general large language model using professional domain knowledge corpus data of each network segment;

[0007] Perform fault analysis on the alarm information using each fault analysis model corresponding to the fault network segment, and obtain network fault analysis information output by each fault analysis model;

[0008] The network fault analysis information output by each fault analysis model is input into the second largest language model, and the network fault analysis result is generated by the second largest language model; the second largest language model is obtained by fine-tuning the first largest language model in advance using the business scenario knowledge corpus data of each network segment;

[0009] When the network fault analysis result indicates that the alarm information is real alarm information, a network fault solution is generated based on the network fault analysis result through the second language model; the network fault solution is used to assist in generating a fault handling work order for the alarm information.

[0010] In one embodiment, obtaining alarm information of a network system includes:

[0011] Receive alarm indicator description information sent by each network segment in the network system;

[0012] The alarm indicator description information sent by each network segment is spliced ​​and integrated to obtain the alarm information.

[0013] In one embodiment, a network fault solution is generated based on the network fault analysis result by using the second language model, including:

[0014] The knowledge graph and knowledge base of the network system are called through the second language model to query the structured data that matches the network fault analysis results;

[0015] The second largest language model is used to perform semantic understanding of structured data and generate solutions to network failures.

[0016] In one embodiment, the method further comprises:

[0017] Obtain the full corpus data and build a knowledge graph of the network system based on the full corpus data;

[0018] Based on the knowledge graph of the network system, the professional domain knowledge corpus data and business scenario knowledge corpus data of each network segment are extracted;

[0019] The professional domain knowledge corpus data of each network segment is used to pre-train the general large language model, and the first large language model with the function of locating the faulty network segment is obtained;

[0020] The business scenario knowledge corpus data of each network segment is used to fine-tune the first language model to obtain the second language model with network fault analysis function.

[0021] In one embodiment, the first language model is fine-tuned using the business scenario knowledge corpus data of each network segment to obtain a second language model with network fault analysis function, including:

[0022] The business scenario knowledge corpus data of each network segment is used to fine-tune the first large language model to obtain a fine-tuned large language model.

[0023] Use the fine-tuned large language model to perform entity recognition on the entire corpus data to obtain the entity recognition results;

[0024] Obtaining expert review results for each entity recognition result, and separating positive samples and negative samples from each entity recognition result based on the expert review results;

[0025] Based on the positive samples and the negative samples, the fine-tuned large language model is iteratively fine-tuned to obtain the second large language model with network fault analysis function.

[0026] In one embodiment, based on positive samples and negative samples, the fine-tuned large language model is iteratively fine-tuned to obtain a second large language model with network fault analysis function, including:

[0027] Based on the entity recognition results corresponding to the positive samples, the knowledge graph is updated to obtain an updated knowledge graph;

[0028] The business scenario knowledge corpus data of each network segment is re-extracted from the updated knowledge graph, and the fine-tuned large language model is fine-tuned using the re-extracted business scenario knowledge corpus data of each network segment.

[0029] In one embodiment, based on positive samples and negative samples, the fine-tuned large language model is iteratively fine-tuned to obtain a second large language model with network fault analysis function, including:

[0030] Based on the entity recognition results corresponding to the negative samples, construct prompt word samples;

[0031] Use the prompt word samples to fine-tune the fine-tuned large language model.

[0032] In one embodiment, after the step of generating a network fault solution based on the network fault analysis result by using the second largest language model, the method further includes:

[0033] Get the fault handling result after executing the fault handling work order;

[0034] When the fault handling result indicates that the network fault triggering the alarm information has not been resolved, feedback information is generated;

[0035] The feedback information is input into the first large language model and the second large language model to update the first large language model and the second large language model.

[0036] A device for generating a network fault solution is applied to a network system, the network system includes various network segments, and the device includes:

[0037] An alarm information acquisition module is used to acquire the alarm information of the network system and input the alarm information into the first large language model, and determine the faulty network segment in each network segment through the first large language model; the first large language model is obtained by pre-training a general large language model using the professional domain knowledge corpus data of each network segment;

[0038] An analysis information acquisition module is used to perform fault analysis on the alarm information using each fault analysis model corresponding to the fault network segment, and obtain network fault analysis information output by each fault analysis model;

[0039] An analysis result generation module is used to input the network fault analysis information output by each fault analysis model into the second largest language model, and generate a network fault analysis result through the second largest language model; the second largest language model is obtained by fine-tuning the first largest language model in advance using the business scenario knowledge corpus data of each network segment;

[0040] A solution generation module is used to generate a network fault solution based on the network fault analysis result through the second language model when the network fault analysis result indicates that the alarm information is real alarm information; the network fault solution is used to assist in generating a fault handling work order for the alarm information.

[0041] A computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above method when executing the computer program.

[0042] A computer-readable storage medium stores a computer program, which implements the steps of the above method when executed by a processor.

[0043] A computer program product comprises a computer program, and when the computer program is executed by a processor, the steps of the above method are implemented.

[0044] The method, device, computer equipment, storage medium and computer program product for generating a network fault solution obtain alarm information of a network system, input the alarm information into a first large language model, and determine the faulty network segment in each network segment through the first large language model; the first large language model is obtained by pre-training a general large language model using professional domain knowledge corpus data of each network segment; the alarm information is analyzed using each fault analysis model corresponding to the faulty network segment to obtain network fault analysis information output by each fault analysis model; the network fault analysis information output by each fault analysis model is input into a second large language model, and the faulty network segment is determined through the first large language model; the faulty network segment is determined by pre-training a general large language model using professional domain knowledge corpus data of each network segment ... The network fault analysis result is generated through the second language model; the second language model is obtained by fine-tuning the first language model with the business scenario knowledge corpus data of each network segment in advance; when the network fault analysis result indicates that the alarm information is real alarm information, the network fault solution is generated based on the network fault analysis result through the second language model; the network fault solution is used to assist in generating a fault handling work order for the alarm information; in this way, the first language model obtained by pre-training the general large language model with the professional field knowledge corpus data of each network segment can understand the semantics of the alarm information and accurately determine the fault in each network segment. The fault network segment is identified, and the fault analysis models corresponding to the fault network segment are used to output the network fault analysis information for the alarm information. The business scenario knowledge corpus data of each network segment is then used to fine-tune the first language model to obtain the second language model to perform semantic understanding on each network fault analysis information and generate accurate network fault analysis results. When the network fault analysis result indicates that the alarm information is real alarm information, the second language model is used to generate a network fault solution based on the network fault analysis result, realizing the use of knowledge graph to enhance the large language model's ability to understand the interrelationships between network segments in the network system, while taking into account the knowledge graph. High accuracy, strong interpretability and strong semantic understanding ability of large language models are conducive to the overall detection and tracing of the entire network segment of the network system. At the same time, it realizes the fusion of large language models with different functions and traditional fault analysis models with different functions, taking into account the semantic understanding characteristics of large language models with different functions and the professionalism and lightweight characteristics of traditional fault analysis models with different functions, improving the timeliness and accuracy of network system anomalies. Overall, it integrates the advantages of knowledge graphs, large language models and traditional fault analysis models, and can improve the efficiency and accuracy of handling network system faults. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related technologies, the drawings required for use in the embodiments or the related technical descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0046] Figure 1 An application environment diagram of a method for generating a network fault solution in an embodiment;

[0047] Figure 2 A schematic diagram of a flow chart of a method for generating a network fault solution in an embodiment;

[0048] Figure 3 A schematic diagram of a method for generating a network fault solution in an embodiment;

[0049] Figure 4 A schematic diagram of the relationship between a general large language model, a first large language model, and a second large language model in one embodiment;

[0050] Figure 5 A schematic diagram of generating alarm information for a server of a network optimization system in one embodiment;

[0051] Figure 6 is a schematic diagram of a mutual enhancement process between a knowledge graph and a large language model in one embodiment;

[0052] Figure 7 A schematic diagram of an application mode when a large language model is used as a system entry of a network optimization system in an embodiment;

[0053] Figure 8 A schematic diagram of an application architecture of a network optimization system in one embodiment;

[0054] Fig. 9 A schematic flow chart of a method for generating a network fault solution in another embodiment;

[0055] Fig.10 A structural block diagram of a device for generating a network fault solution in an embodiment;

[0056] Fig.11 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0058] Intelligent operation and maintenance uses artificial intelligence, automation, big data analysis and other technologies to monitor and analyze network system operation data in real time, which can detect and predict network system failures in advance, optimize network system performance, and improve network system stability and reliability. In the current network optimization field, traditional artificial intelligence algorithms and knowledge graphs are often used to improve the intelligence of network systems.

[0059] In recent years, large language models have emerged rapidly. At present, there is no mature technology to realize the professional transformation of general large language models, nor is there a mature technology to integrate large language models with traditional AI small models and knowledge graphs to integrate multi-source information in network systems, so as to achieve the high structure and interpretability of knowledge graphs, the professionalism and lightweight of traditional AI small models, and the versatility and language understanding ability of large language models.

[0060] The method for generating a network fault solution provided in the embodiment of the present application can integrate multi-source information in the network system, achieve the high structure and interpretability of the knowledge graph, the professionalism and lightness of the traditional AI small model, and the versatility and language understanding ability of the large language model.

[0061] The method for generating a network fault solution provided in the embodiment of the present application can be applied to Figure 1In the application environment shown, the terminal 102 communicates with the server 104 of the network optimization system through the network. The data storage system can store data that the server 104 of the network optimization system needs to process. The data storage system can be integrated on the server 104 of the network optimization system, or it can be placed on the cloud or other network servers. The server 104 of the network optimization system obtains the alarm information of the network system, and inputs the alarm information into the first language model, and determines the faulty network segment in each network segment through the first language model; the first language model is obtained by pre-training the general language model with the professional domain knowledge corpus data of each network segment; the server 104 of the network optimization system uses each fault analysis model corresponding to the faulty network segment to perform fault analysis on the alarm information, and obtains the network fault analysis information output by each fault analysis model; the server 104 of the network optimization system inputs the network fault analysis information output by each fault analysis model into the second language model, and generates the network fault analysis result through the second language model; the second language model is obtained by fine-tuning the first language model with the business scenario knowledge corpus data of each network segment in advance; the server 104 of the network optimization system generates a network fault solution based on the network fault analysis result through the second language model when the network fault analysis result indicates that the alarm information is real alarm information; the fault solution is used to assist in generating a fault handling work order for the alarm information. The server 104 of the network optimization system can send the fault handling work order to the terminal 102. The terminal 102 may be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, IoT devices, and portable wearable devices. The IoT devices may be smart speakers, smart TVs, smart air conditioners, smart car-mounted devices, etc. The portable wearable devices may be smart watches, smart bracelets, head-mounted devices, etc. The server 104 may be implemented as an independent server or a server cluster consisting of multiple servers.

[0062] In an exemplary embodiment, Figure 2 As shown, a method for generating a network fault solution is provided, and the method is applied to Figure 1 The server 104 of the network optimization system in the embodiment is used as an example to illustrate the method, which includes the following steps S202 to S208. Among them:

[0063] Step S202, obtain the alarm information of the network system, and input the alarm information into the first large language model, and determine the faulty network segment in each network segment through the first large language model; the first large language model is obtained by pre-training the general large language model with the professional domain knowledge corpus data of each network segment.

[0064] Among them, the alarm information can come from customer reports, alarm system acquisition and control alarms and alarm system logs. The alarm system is a system used to conduct real-time monitoring of access networks, backbone networks, core networks, wireless networks and other network segments.

[0065] Among them, the first large language model is obtained by pre-training the general large language model using the professional domain knowledge corpus data of each network segment.

[0066] Among them, the professional domain knowledge corpus data of each network segment is generated based on the graph data corresponding to the professional domain knowledge graph of each network segment. The professional domain knowledge graph of any network segment is constructed based on the professional domain knowledge of the network segment. The graph data corresponding to the professional domain knowledge graph of each network segment is used to train the general large language model, so that the large language model can be familiar with the professional domain knowledge of each network segment, so that when an alarm information is received, it can quickly determine which network segment the alarm information comes from.

[0067] In actual applications, the professional field knowledge graphs of each network segment are dynamically updated and continuously improved.

[0068] Among them, the network system includes access network, backbone network, core network, wireless network and other network segments. In the traditional network optimization system, each network segment is maintained independently, but in fact, the network segments are interconnected. The alarm generated by the wireless network may be a problem with the core network. It is very necessary to conduct overall detection and tracing of the entire network in a timely manner when the alarm information is received. The method provided in this application utilizes the knowledge graph to solve the problem that the traditional network optimization system cannot conduct overall detection and tracing of the entire network.

[0069] Optionally, refer to Figure 3 , the alarm information of the network system can be obtained through customer declarations, the collection and control alarms of the alarm system and the alarm system logs, and the alarm information is input into the first language model as a prompt word. The alarm information is semantically understood by the first language model to determine the network segment that issues the alarm information and identify the network segment as a faulty network segment.

[0070] In fact, step S202 is to classify the alarm information and determine which network segment the alarm information belongs to.

[0071] Step S204: perform fault analysis on the alarm information using the fault analysis models corresponding to the faulty network segment to obtain network fault analysis information output by the fault analysis models.

[0072] Among them, the fault analysis model corresponding to the faulty network segment may include various fault analysis models built based on association rule mining algorithms, heuristic search algorithms, Prophet (time series prediction algorithm), LSTM (Long Short-Term Memory, long short-term memory network), etc.

[0073] Among them, network fault analysis information may include fault classification, fault probability, fault location, fault trend judgment, traffic warning and other information.

[0074] Optionally, refer to Figure 3 , use the fault analysis models corresponding to the faulty network segment to perform fault analysis on the alarm information, and obtain the fault classification, fault probability, fault location, fault trend judgment, traffic warning and other information output by each fault analysis model.

[0075] Step S206, input the network fault analysis information output by each fault analysis model into the second largest language model, and generate the network fault analysis result through the second largest language model; the second largest language model is obtained by fine-tuning the first largest language model in advance using the business scenario knowledge corpus data of each network segment.

[0076] Among them, the second largest language model is obtained by fine-tuning the first largest language model in advance using the business scenario knowledge corpus data of each network segment.

[0077] In order to facilitate the understanding of those skilled in the art, Figure 4 A diagram of the relationship between the general large language model, the first large language model, and the second large language model is provided. Figure 4The L0 basic large language model corresponds to the general large language model, the L1 wireless network domain large language model and the L1 core network domain large language model correspond to the first large language model, and the L2 wireless network scenario large language model and the L2 core network scenario large language model correspond to the second large language model. The L0 basic large model can include the CV large model (a large deep learning model for computer vision tasks), the NLP large model (Natural Language Processing, a large natural language processing model), the multimodal large model, the scientific computing large model, the graph network large model, ChatGPT (Chat Generative Pre-trained Transformer, built on the GPT system large model), ChatGGLM (a generative pre-trained model based on an autoregressive language model), Baichuan (a large language model with over 100 billion parameters), InternLM (an advanced natural language processing model with powerful semantic understanding capabilities. By using InternLM, automatic text classification, summary extraction, entity recognition and other functions can be achieved), Vicuna (an open source chatbot), LLaMA (Large Language Model Meta AI, large language model Meta AI), Alpaca (a fine-tuned instruction-following language model based on Meta's LLaMA7B model), TeleChat (Xingchen semantic large model), Ziya (a Chinese-English bilingual pre-trained language model with 13 billion parameters based on LLaMa); the first large language model, including the L1 wireless network field large language model and the L1 core network field large language model, has the functions of knowledge understanding, analysis and prediction, automatic optimization, and solution generation; the L2 wireless network scenario large language model has the function of generating solutions for abnormal OMC gateway load, network element security, automatic site commissioning, automatic acceptance, and cell decommissioning. The L2 core network scenario large language model has the function of generating solutions for network element failures, capacity overruns, link failures, address pool utilization overruns, software versions, etc.

[0078] Among them, the business scenario knowledge corpus data of each network segment is generated based on the graph data corresponding to the business scenario knowledge graph of each network segment. The business scenario knowledge graph of any network segment is constructed based on the business scenario knowledge of the network segment. The business scenario knowledge graph of each network segment is used to train the first largest language model with the function of locating the faulty network segment. The trained second largest language model is familiar with the business scenario knowledge of each network segment. Therefore, when the second largest language model is familiar with the business scenario knowledge of the faulty network segment, the second largest language model can identify whether the alarm information is a false alarm based on the network fault analysis information output by each fault analysis model corresponding to the faulty network segment. When the alarm information is not a false alarm, the fault can be located.

[0079] In actual applications, the business scenario knowledge graphs of each network segment are dynamically updated and continuously improved.

[0080] Among them, the network fault analysis result can be the network fault analysis conclusion obtained after the second largest language model combines the network fault analysis information output by each fault analysis model for semantic understanding. The conclusion indicates whether the alarm information is a false alarm, and the corresponding fault location result when the alarm information is not a false alarm.

[0081] Optionally, refer to Figure 3 , the network fault analysis information output by each fault analysis model is input into the second largest language model, the network fault analysis information output by each fault analysis model is semantically understood by the second largest language model, and the network fault analysis result is output.

[0082] Step S208, when the network fault analysis result indicates that the alarm information is real alarm information, a network fault solution is generated based on the network fault analysis result by using the second language model; the network fault solution is used to assist in generating a fault handling work order for the alarm information.

[0083] Among them, the network fault solution can be used as an auxiliary solution to assist in generating fault handling work orders for alarm information.

[0084] Optionally, refer to Figure 3 ,When the network fault analysis result indicates that the alarm information is not a false alarm, the specific fault location is determined in the fault network segment according to the fault location provided by the network fault analysis result, and the corresponding network fault solution is generated through the second language model.

[0085] In actual applications, when the network fault analysis results indicate that the alarm information is a false alarm, the expert will further review the alarm information. If the expert determines that the alarm information is indeed a false alarm, the process ends.

[0086] In the above-mentioned method for generating a network fault solution, alarm information of the network system is obtained and input into the first language model, and the faulty network segment is determined in each network segment through the first language model; the first language model is obtained by pre-training a general language model with professional domain knowledge corpus data of each network segment; each fault analysis model corresponding to the faulty network segment is used to perform fault analysis on the alarm information to obtain network fault analysis information output by each fault analysis model; the network fault analysis information output by each fault analysis model is input into the second language model, and the network fault analysis is generated through the second language model Results: The second language model is obtained by fine-tuning the first language model using the business scenario knowledge corpus data of each network segment in advance; when the network fault analysis result indicates that the alarm information is real alarm information, the second language model is used to generate a network fault solution based on the network fault analysis result; the network fault solution is used to assist in generating a fault handling work order for the alarm information; in this way, the first language model obtained by pre-training the general large language model using the professional field knowledge corpus data of each network segment can understand the semantics of the alarm information, accurately determine the faulty network segment in each network segment, and then use the faulty network segment to identify the faulty network segment. Each fault analysis model corresponding to the segment outputs network fault analysis information for the alarm information, and then the second language model obtained by fine-tuning the first language model using the business scenario knowledge corpus data of each network segment performs semantic understanding on each network fault analysis information to generate accurate network fault analysis results. When the network fault analysis result indicates that the alarm information is real alarm information, the second language model is used to generate a network fault solution based on the network fault analysis result, realizing the use of knowledge graph to enhance the understanding ability of the large language model of the mutual correlation relationship between each network segment in the network system, taking into account the high accuracy and strong interpretability of the knowledge graph and the strong semantic understanding ability of the large language model, which is conducive to the overall detection and tracing of the entire network segment of the network system. At the same time, the large language model with different functions is integrated with the traditional fault analysis model with different functions, taking into account the semantic understanding characteristics of the large language model with different functions and the professionalism and lightweight characteristics of the traditional fault analysis model with different functions, improving the timeliness and accuracy of the network system in discovering anomalies. Overall, the advantages of the knowledge graph, the large language model and the traditional fault analysis model are integrated to improve the processing efficiency and accuracy of network system faults.

[0087] In an exemplary embodiment, obtaining alarm information of a network system includes: receiving alarm indicator description information sent by each network segment in the network system; and splicing and fusing the alarm indicator description information sent by each network segment to obtain alarm information.

[0088] Among them, the alarm indicator description information can be the information obtained by describing the abnormal device status indicator in natural language when the device status is abnormal. The alarm indicator description information can also be the description information of indicators such as KPI key performance indicators and KQI key quality indicators.

[0089] Optionally, the server of the network optimization system can receive alarm indicator description information sent by each network segment in the network system, and splice and fuse the alarm indicator description information sent by each network segment to obtain alarm information. Subsequently, the alarm information can be input as a prompt word into the first largest language model so that the first largest language model can perform semantic understanding of the alarm information to determine the network segment where the fault occurs.

[0090] refer to Figure 5 , each network segment will send alarm indicator description information to the alarm system when an abnormality occurs, and the alarm system will send the alarm indicator description information to the network optimization system, or the user will submit complaint information to the network optimization when an abnormality is detected. The server of the network optimization system will splice and fuse the alarm indicator description information sent by the alarm system and the user complaint information to obtain the alarm information, and then execute steps S202 to S208 of the previous embodiment to obtain a network fault solution. Among them, since the second largest language model is fine-tuned based on the graph data corresponding to the professional field knowledge graph of each network segment, the second largest language model can output the poor quality problem location, fault cause analysis, solution process and recommended solution for the alarm information based on the learned knowledge to assist decision-making.

[0091] In this embodiment, alarm indicator description information sent by each network segment in the network system is received; the alarm indicator description information sent by each network segment is spliced ​​and fused to obtain alarm information; in this way, the alarm data from different network segments can be spliced ​​and used as the input of the first language model, thereby realizing the fusion of multi-source information across network segments, which is conducive to improving the timeliness and accuracy of problem discovery.

[0092] In an exemplary embodiment, a network fault solution is generated based on the network fault analysis result through the second largest language model, including: calling the knowledge graph and knowledge base of the network system through the second largest language model to query structured data that matches the network fault analysis result; and performing semantic understanding of the structured data through the second largest language model to generate a network fault solution.

[0093] Among them, the knowledge graph of the network system can be a pre-built professional field knowledge graph corresponding to the network system, which can represent the association relationship between the various network segments of the network system, and the knowledge graph serves as stock data.

[0094] The knowledge base of the network system may be a text vector library storing professional knowledge of the network system. In practical applications, the knowledge base may be mounted for invocation by the second largest language model.

[0095] In practical applications, the data of the knowledge graph is used as stock data, and the data of the knowledge base is used as real-time data. The knowledge graph and the knowledge base can be understood as an information retrieval system for the second largest language model to call. Therefore, the network optimization system can be compatible with stock data processing and implementation data processing. The knowledge graph and the knowledge base can be used as verification tools in the scenario of accurate problem solving to reduce model illusions and improve the accuracy and security of LLM input.

[0096] Among them, structured data can be the query results output by the knowledge graph. Since the knowledge graph is a structured semantic knowledge base, the search results output by the knowledge graph are also structured data.

[0097] Optionally, refer to Figure 3 When the network fault analysis result indicates that the alarm information is not a false alarm, the knowledge graph and knowledge base of the network system are called through the second largest language model, and the network fault analysis result is input into the knowledge graph and the mounted knowledge base, and the structured data matching the network fault analysis result is queried and obtained. Then, the structured data is semantically understood through the second largest language model to generate a network fault solution.

[0098] In actual applications, after a network fault solution is generated, the pre-dispatching process will begin. Experts will review the solution for the pre-dispatching process. If the experts determine that the solution is accurate, the dispatching step will be executed.

[0099] In this embodiment, the knowledge graph and knowledge base of the network system are called through the second largest language model to query the structured data matching the network fault analysis results; the structured data is semantically understood through the second largest language model to generate a network fault solution; in this way, when the network fault analysis results are obtained, the existing knowledge graph and the real-time knowledge base can be entered to query the structured data matching the network fault analysis results, and the second largest language model is used to semantically understand the structured data to generate a targeted network fault solution, thereby improving the accuracy of solution generation.

[0100] In an exemplary embodiment, the method also includes: obtaining full corpus data, and constructing a knowledge graph of the network system based on the full corpus data; extracting professional domain knowledge corpus data of each network segment and business scenario knowledge corpus data of each network segment based on the knowledge graph of the network system; using the professional domain knowledge corpus data of each network segment to pre-train a general large language model to obtain a first large language model with a faulty network segment location function; using the business scenario knowledge corpus data of each network segment to fine-tune the first large language model to obtain a second large language model with a network fault analysis function.

[0101] The full corpus data can include coarse-grained corpus data such as wiki, white papers, dictionaries, etc., and fine-grained corpus data such as cases, alarms, and work orders. Figure 6 As shown in the figure, the full corpus data will be updated in real time based on the collected data, and the knowledge graph of the network system is also in the process of continuous improvement and updating. Therefore, the professional domain knowledge corpus data of each network segment and the business scenario knowledge corpus data of each network segment will also be updated, and the first and second language models will be continuously updated. In the process of pre-training the first language model, the input text sequence is encoded as , the entity encoding stored in the knowledge graph is , it is necessary to linearly combine the input text sequence encoding with the entity encoding and modify it with the weight matrix W of expert knowledge, expressed as .

[0102] Among them, the faulty network segment location function may refer to the function of determining the faulty network segment among the network segments in the network system. In practical applications, the faulty network segment location function means that the first language model can semantically understand the received alarm information based on the learned professional domain knowledge of each network segment, thereby determining the faulty network.

[0103] Among them, the network fault analysis function may refer to: the second largest language model performs semantic understanding of the network fault analysis information output by each fault analysis model of the faulty network segment based on the learned professional domain knowledge of each network segment and the learned business scenario knowledge of each network segment, so as to determine whether the alarm information is a false alarm, and locate the fault if the alarm information is not a false alarm.

[0104] Optionally, refer to Figure 6The server obtains the full corpus data and builds the knowledge graph of the network system based on the full corpus data. Based on the knowledge graph of the network system, the server extracts the professional domain knowledge corpus data and the business scenario knowledge corpus data of each network segment. The server uses the professional domain knowledge corpus data of each network segment (corresponding to coarse-grained corpus data such as wiki, white paper, dictionary, etc.) to pre-train the general large language model (corresponding to L0 general LLM). The pre-training process includes sampling, preprocessing and the use of RoPE (Rotary Position Embedding) method to obtain the first large language model (corresponding to L1 domain LLM) with the function of locating the faulty network segment. The business scenario knowledge corpus data of each network segment (corresponding to fine-grained corpus data such as cases, alarms, and work orders) is used to fine-tune the first large language model to obtain the second large language model (corresponding to L2 scenario LLM) with the function of network fault analysis.

[0105] In this embodiment, a knowledge graph of the network system is constructed by acquiring full corpus data and based on the full corpus data; based on the knowledge graph of the network system, professional domain knowledge corpus data of each network segment and business scenario knowledge corpus data of each network segment are extracted; the professional domain knowledge corpus data of each network segment is used to pre-train a general large language model to obtain a first large language model with a faulty network segment locating function; the business scenario knowledge corpus data of each network segment is used to fine-tune the first large language model to obtain a second large language model with a network fault analysis function; in this way, the knowledge graph can be used to perform intensive training on the large language model to obtain the first large language model with a faulty network segment locating function and the second large language model with a network fault analysis function, which is beneficial for accurately locating the faulty network segment and accurately judging the authenticity of the alarm information.

[0106] In an exemplary embodiment, the business scenario knowledge corpus data of each network segment is used to fine-tune the first large language model to obtain a second large language model with a network fault analysis function, including: using the business scenario knowledge corpus data of each network segment to fine-tune the first large language model to obtain a fine-tuned large language model; using the fine-tuned large language model to perform entity recognition on the full corpus data to obtain entity recognition results; obtaining expert review results for each entity recognition result, and separating positive samples and negative samples from each entity recognition result based on the expert review results; based on the positive samples and negative samples, iteratively fine-tuning the fine-tuned large language model to obtain a second large language model with a network fault analysis function.

[0107] Among them, the entity recognition result can correspond to specific things in the full corpus data. For example, performing entity recognition on a certain corpus data can obtain specific things such as "equipment", "indicator", "fault" and so on involved in the corpus data.

[0108] Among them, a positive sample may refer to an entity recognition result that an expert determines the output of the large language model to be a correct one, and a negative sample may refer to an entity recognition result that an expert determines the output of the large language model to be an incorrect one.

[0109] Optionally, refer to Figure 6 , the business scenario knowledge corpus data of each network segment (corresponding to fine-grained corpus data such as cases, alarms, and work orders) is used to fine-tune the first large language model (corresponding to the L1 scenario LLM). The fine-tuning can adopt the LoRA method (a method for fine-tuning large pre-trained language models. The core idea is to add small, low-rank matrices in the key layers of the model to adjust the behavior of the model, rather than directly changing the structure of the entire model) to obtain a fine-tuned large language model (corresponding to the L2 scenario LLM). Then, the fine-tuned large language model (corresponding to the L2 scenario LLM) is used to perform entity recognition on the full corpus data to obtain the entity recognition results. The RLHF (Reinforcement Learning from Human Feedback) method is used for reinforcement learning. The entity recognition results are reviewed by experts. Based on the expert review results, positive samples and negative samples are separated from the entity recognition results. The fine-tuned large language model (corresponding to the L2 scenario LLM) is iteratively fine-tuned using the positive samples and negative samples to obtain the second large language model (corresponding to the L2 scenario LLM) with network fault analysis function.

[0110] It should be noted that the "fine-tuned large language model" and the "second largest language model with network fault analysis function" in this embodiment are both the second largest language model with network fault analysis function. The two correspond to the states of the second largest language model at different times, and the second largest language model is always in the process of iterative fine-tuning.

[0111] In this embodiment, the first large language model is fine-tuned by using the business scenario knowledge corpus data of each network segment to obtain a fine-tuned large language model; the fine-tuned large language model is used to perform entity recognition on the full corpus data to obtain each entity recognition result; an expert review result for each entity recognition result is obtained, and positive samples and negative samples are separated from each entity recognition result based on the expert review result; based on the positive samples and negative samples, the fine-tuned large language model is iteratively fine-tuned to obtain a second large language model with network fault analysis function; in this way, the accuracy of the fine-tuned large language model can be verified by performing entity recognition on the full corpus data. In the process of verifying the accuracy of the fine-tuned large language model, a reinforcement learning technology based on human feedback is used, that is, expert review feedback is used to capture subtle differences and subjectivity, which is conducive to more effective self-learning of the large language model, and can accelerate the fine-tuning process of the large language model and improve the performance of the large language model.

[0112] In an exemplary embodiment, the fine-tuned large language model is iteratively fine-tuned based on positive samples and negative samples to obtain a second large language model with network fault analysis function, including: updating the knowledge graph based on the entity recognition results corresponding to the positive samples to obtain an updated knowledge graph; re-extracting the business scenario knowledge corpus data of each network segment from the updated knowledge graph, and using the re-extracted business scenario knowledge corpus data of each network segment to fine-tune the fine-tuned large language model.

[0113] Alternatively, see Figure 6 Based on the entity recognition results corresponding to the positive samples, the knowledge graph of the network system is updated, and the business scenario knowledge corpus data of each network segment is re-extracted from the updated knowledge graph. The re-extracted business scenario knowledge corpus data of each network segment is used to fine-tune the fine-tuned large language model.

[0114] In this embodiment, the knowledge graph is updated based on the entity recognition results corresponding to the positive samples to obtain an updated knowledge graph; the business scenario knowledge corpus data of each network segment is re-extracted from the updated knowledge graph, and the fine-tuned large language model is fine-tuned using the re-extracted business scenario knowledge corpus data of each network segment; in this way, the knowledge graph of the network system can be complemented and updated using the positive samples, and on this basis, the business scenario knowledge corpus data of each network segment can be re-extracted from the complemented and updated knowledge graph, and the fine-tuned large language model can be fine-tuned, thereby achieving iterative fine-tuning of the second largest language model.

[0115] In an exemplary embodiment, based on positive samples and negative samples, the fine-tuned large language model is iteratively fine-tuned to obtain a second large language model with network fault analysis function, including: constructing a prompt word sample based on the entity recognition result corresponding to the negative sample; using the prompt word sample to fine-tune the fine-tuned large language model.

[0116] Alternatively, see Figure 6 ,Based on the entity recognition results corresponding to the negative samples, a prompt word sample is constructed, and the prompt word sample is used to fine-tune the fine-tuned large language model.

[0117] In this embodiment, a prompt word sample is constructed based on the entity recognition result corresponding to the negative sample; the prompt word sample is used to fine-tune the fine-tuned large language model; in this way, the prompt word sample can be constructed using the negative sample and fed back to the fine-tuned large language model to achieve iterative fine-tuning of the second largest language model.

[0118] In an exemplary embodiment, after the step of generating a network fault solution based on the network fault analysis result through the second largest language model, the method also includes: obtaining the fault handling result after executing the fault handling work order; generating feedback information when the fault handling result indicates that the network fault that triggered the alarm information has not been resolved; and inputting the feedback information into the first largest language model and the second largest language model to update the first largest language model and the second largest language model.

[0119] The fault handling result may be that the network fault that triggered the alarm information has been resolved and the alarm information no longer appears, or that the network fault that triggered the alarm information has not been resolved and the same alarm information still appears.

[0120] Optionally, refer to Figure 3 After the second language model generates a network fault solution based on the network fault analysis result, it is processed based on the fault handling work order to obtain the fault handling result. When the fault handling result is that the network fault that triggered the alarm information has not been resolved and the same alarm information still appears, feedback information is generated. Since the first language model and the second language model both participate in data processing from obtaining the alarm information to generating the network fault solution, inputting the feedback information into the first language model and the second language model can make the first language model and the second language model more accurate in the next data processing process.

[0121] In this embodiment, the fault handling result after executing the fault handling work order is obtained; when the fault handling result indicates that the network fault that triggered the alarm information has not been resolved, feedback information is generated; the feedback information is input into the first largest language model and the second largest language model to update the first largest language model and the second largest language model; in this way, the semantic understanding accuracy of the first largest language model and the second largest language model can be verified according to the fault handling result after executing the fault handling work order, and when the fault handling result indicates that the network fault that triggered the alarm information has not been resolved, the fault handling result can be fed back to the first largest language model and the second largest language model to improve the semantic understanding ability of the first largest language model and the second largest language model in the next semantic understanding process.

[0122] In practical applications, such as Figure 7 As shown, the large language model can also be used as the system entrance of the network optimization system to understand the user's intention and generate answers for the user's intention. When the network optimization system receives the user's intention, it understands the user's intention and determines the query operation that needs to be performed. For example, it executes the cloud network operation status indicator query, enters the big data lake indicator system for query, and obtains the first query result, and executes the disaster impact query, enters the knowledge base for query, and obtains the second query result, and executes the solution query, enters the knowledge base for query and then calls the GIS interface for query, and obtains the third query result. Based on the first query result, the second query result and the third query result, the answer is generated. At the same time, when the network optimization system receives the user's intention, it understands the user's intention, logs in to the 5GR system, queries the poor quality cells, distributed files and system data, and generates the corresponding query results, which are displayed in the 5GR system display.

[0123] In order to facilitate the understanding of those skilled in the art, Figure 8An application architecture diagram of the network optimization system is provided. The application architecture of the network optimization system includes a data collection and processing module, a knowledge management module, an AI capability training model, a large language model training and fine-tuning module, and an application module. Among them, the data collection and processing module is used to collect data from the network optimization system, data in the data lake, wikis, web pages, white papers, open data sets, and other data, and perform preprocessing to obtain processed data; the knowledge management module is used to build a knowledge graph for the network system based on the processed data to obtain a knowledge graph that can be used for knowledge retrieval, intelligent recommendation, intelligent question and answer, knowledge extraction, and intelligent evaluation. The knowledge graph can be used to generate corpus data to pre-train and fine-tune the large language model; the AI ​​capability training model is used to train traditional AI models using processed data to obtain data that can be used for quality difference identification, The fault analysis models for fault location, intent understanding, text summarization, and trend prediction can be used to collaborate with the large language model in the predictive reasoning stage; the large language model training and fine-tuning module is used to pre-train and fine-tune the large language model using the corpus data generated by the knowledge graph, processed data, and each trained fault analysis model, and to complete the knowledge graph of the network system with the new knowledge generated by the large language model during pre-training and fine-tuning; the application module is used to use the pre-trained first large language model and the fine-tuned second large language model for network optimization, quality difference identification, intelligent question and answer, and intelligent recommendation.

[0124] The network optimization system of the present application, by combining a large language model with traditional knowledge graphs, knowledge bases and AI models, not only improves the utilization rate and application flexibility of various resources, but also improves the accuracy of fault location in existing network optimization scenarios, enhances the early warning capability of network faults, strengthens the inheritance of professional knowledge in the network optimization process, and realizes intelligent operation and maintenance.

[0125] In another embodiment, Fig. 9 As shown, a method for generating a network fault solution is provided, and the method is applied to Figure 1 The server 104 of the network optimization system in the embodiment is used as an example to illustrate, and the following steps are included:

[0126] Step S902: Receive alarm indicator description information sent by each network segment in the network system.

[0127] Step S904, concatenate and merge the alarm indicator description information sent by each network segment to obtain alarm information.

[0128] Step S906, input the alarm information into the first large language model, and determine the faulty network segment in each network segment through the first large language model; the first large language model is obtained by pre-training a general large language model using professional domain knowledge corpus data of each network segment.

[0129] Step S908: Perform fault analysis on the alarm information using the fault analysis models corresponding to the faulty network segment to obtain network fault analysis information output by the fault analysis models.

[0130] Step S910, input the network fault analysis information output by each fault analysis model into the second largest language model, and generate the network fault analysis result through the second largest language model; the second largest language model is obtained by fine-tuning the first largest language model in advance using the business scenario knowledge corpus data of each network segment.

[0131] Step S912, when the network fault analysis result indicates that the alarm information is real alarm information, the knowledge graph and knowledge base of the network system are called through the second language model to query the structured data matching the network fault analysis result;

[0132] Step S914, semantic understanding of the structured data is performed through the second largest language model to generate a network fault solution; the fault solution is used to assist in generating a fault handling work order for the alarm information.

[0133] It should be noted that the specific limitations of the above steps can refer to the specific limitations of the method for generating a network fault solution mentioned above.

[0134] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0135] Based on the same inventive concept, the embodiment of the present application also provides a device for generating a network fault solution for implementing the method for generating a network fault solution involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in the embodiments of the device for generating one or more network fault solutions provided below can refer to the limitations of the method for generating a network fault solution above, and will not be repeated here.

[0136] In an exemplary embodiment, Fig.10As shown, a device for generating a network fault solution is provided, which is applied to a network system, the network system includes various network segments, and the device for generating a network fault solution includes: an alarm information acquisition module 1002, an analysis information acquisition module 1004, an analysis result generation module 1006 and a solution generation module 1008, wherein:

[0137] The alarm information acquisition module 1002 is used to obtain the alarm information of the network system and input the alarm information into the first large language model, and determine the faulty network segment in each network segment through the first large language model; the first large language model is obtained by pre-training the general large language model with the professional domain knowledge corpus data of each network segment;

[0138] The analysis information acquisition module 1004 is used to perform fault analysis on the alarm information using each fault analysis model corresponding to the fault network segment, and obtain network fault analysis information output by each fault analysis model;

[0139] The analysis result generating module 1006 is used to input the network fault analysis information output by each fault analysis model into the second largest language model, and generate the network fault analysis result through the second largest language model; the second largest language model is obtained by fine-tuning the first largest language model in advance using the business scenario knowledge corpus data of each network segment;

[0140] The solution generation module 1008 is used to generate a network fault solution based on the network fault analysis result through the second language model when the network fault analysis result indicates that the alarm information is real alarm information; the network fault solution is used to assist in generating a fault handling work order for the alarm information.

[0141] In one of the embodiments, the alarm information acquisition module 1002 is specifically used to receive alarm indicator description information sent by each network segment in the network system; and to splice and merge the alarm indicator description information sent by each network segment to obtain alarm information.

[0142] In one of the embodiments, the solution generation module 1008 is specifically used to call the knowledge graph and knowledge base of the network system through the second largest language model to query the structured data that matches the network fault analysis results; perform semantic understanding of the structured data through the second largest language model to generate a network fault solution.

[0143] In one of the embodiments, the device also includes: a training module, used to obtain full corpus data, and build a knowledge graph of the network system based on the full corpus data; based on the knowledge graph of the network system, extract the professional domain knowledge corpus data of each network segment and the business scenario knowledge corpus data of each network segment; use the professional domain knowledge corpus data of each network segment to pre-train the general large language model to obtain a first large language model with a faulty network segment location function; use the business scenario knowledge corpus data of each network segment to fine-tune the first large language model to obtain a second large language model with a network fault analysis function.

[0144] In one of the embodiments, the training module is specifically used to use the business scenario knowledge corpus data of each network segment to fine-tune the first large language model to obtain a fine-tuned large language model; use the fine-tuned large language model to perform entity recognition on the full corpus data to obtain entity recognition results; obtain expert review results for each entity recognition result, and separate positive samples and negative samples from each entity recognition result based on the expert review results; based on the positive samples and negative samples, iteratively fine-tune the fine-tuned large language model to obtain a second large language model with network fault analysis function.

[0145] In one of the embodiments, the training module is specifically used to update the knowledge graph based on the entity recognition results corresponding to the positive samples to obtain an updated knowledge graph; re-extract the business scenario knowledge corpus data of each network segment from the updated knowledge graph, and use the re-extracted business scenario knowledge corpus data of each network segment to fine-tune the fine-tuned large language model.

[0146] In one of the embodiments, the training module is further used to construct a prompt word sample based on the entity recognition result corresponding to the negative sample; and use the prompt word sample to fine-tune the fine-tuned large language model.

[0147] In one of the embodiments, the device also includes: a feedback module, which is used to obtain the fault handling result after executing the fault handling work order; generate feedback information when the fault handling result indicates that the network fault that triggered the alarm information has not been resolved; and input the feedback information into the first largest language model and the second largest language model to update the first largest language model and the second largest language model.

[0148] Each module in the above-mentioned network fault solution generation device can be implemented in whole or in part by software, hardware and a combination thereof. Each of the above-mentioned modules can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each of the above modules.

[0149] In an exemplary embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as shown in FIG. Fig.11 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the generation data of the network fault solution. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for generating a network fault solution is implemented.

[0150] Those skilled in the art will understand that Fig.11 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0151] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the method for generating a network fault solution. The steps of the method for generating a network fault solution here may be the steps in the method for generating a network fault solution in each of the above embodiments.

[0152] In one embodiment, a computer-readable storage medium is provided, which stores a computer program. When the computer program is executed by a processor, the processor executes the steps of the method for generating a network fault solution. The steps of the method for generating a network fault solution here may be the steps in the method for generating a network fault solution in each of the above embodiments.

[0153] In one embodiment, a computer program product is provided, including a computer program. When the computer program is executed by a processor, the processor executes the steps of the method for generating a network fault solution. The steps of the method for generating a network fault solution here may be the steps in the method for generating a network fault solution in each of the above embodiments.

[0154] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0155] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0156] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for generating a network fault solution, characterized in that: Applied to a network system, the network system includes various network segments, and the method includes: Acquire alarm information of the network system, and input the alarm information into a first large language model, and determine the faulty network segment in each of the network segments through the first large language model; the first large language model is obtained by pre-training a general large language model using professional domain knowledge corpus data of each of the network segments; Performing fault analysis on the alarm information using each fault analysis model corresponding to the fault network segment to obtain network fault analysis information output by each fault analysis model; The network fault analysis information output by each of the fault analysis models is input into a second language model, and a network fault analysis result is generated by the second language model; the second language model is obtained by fine-tuning the first language model in advance using the business scenario knowledge corpus data of each of the network segments; When the network fault analysis result indicates that the alarm information is real alarm information, a network fault solution is generated based on the network fault analysis result through the second largest language model; the network fault solution is used to assist in generating a fault handling work order for the alarm information.

2. The method according to claim 1, characterized in that The obtaining of the alarm information of the network system includes: Receiving alarm indicator description information sent by each network segment in the network system; The alarm indicator description information sent by each of the network segments is spliced ​​and merged to obtain the alarm information.

3. The method according to claim 1, characterized in that Generating a network fault solution based on the network fault analysis result by using the second largest language model includes: The knowledge graph and knowledge base of the network system are called by the second largest language model to query structured data matching the network fault analysis result; The structured data is semantically understood by using the second largest language model to generate the network fault solution.

4. The method according to claim 1, characterized in that: The method further comprises: Acquire the full amount of corpus data, and build a knowledge graph of the network system based on the full amount of corpus data; Based on the knowledge graph of the network system, extract the professional domain knowledge corpus data of each network segment and the business scenario knowledge corpus data of each network segment; Pre-training the universal large language model using the professional domain knowledge corpus data of each of the network segments to obtain the first large language model with a faulty network segment location function; The first largest language model is fine-tuned using the business scenario knowledge corpus data of each of the network segments to obtain the second largest language model with network fault analysis function.

5. The method according to claim 4, characterized in that The fine-tuning of the first language model by using the business scenario knowledge corpus data of each network segment to obtain the second language model with network fault analysis function includes: Using the business scenario knowledge corpus data of each of the network segments, fine-tuning the first large language model to obtain a fine-tuned large language model; Using the fine-tuned large language model to perform entity recognition on the entire corpus data to obtain entity recognition results; Obtaining an expert review result for each of the entity recognition results, and separating positive samples and negative samples from each of the entity recognition results based on the expert review result; Based on the positive samples and the negative samples, the fine-tuned large language model is iteratively fine-tuned to obtain the second large language model with the network fault analysis function.

6. The method according to claim 5, characterized in that The iterative fine-tuning of the fine-tuned large language model based on the positive samples and the negative samples to obtain the second large language model having the network fault analysis function includes: Based on the entity recognition result corresponding to the positive sample, the knowledge graph is updated to obtain an updated knowledge graph; The business scenario knowledge corpus data of each network segment is re-extracted from the updated knowledge graph, and the fine-tuned large language model is fine-tuned using the re-extracted business scenario knowledge corpus data of each network segment.

7. The method according to claim 5, characterized in that The iterative fine-tuning of the fine-tuned large language model based on the positive samples and the negative samples to obtain the second large language model having the network fault analysis function includes: Constructing a prompt word sample based on the entity recognition result corresponding to the negative sample; The cue word sample is used to fine-tune the fine-tuned large language model.

8. The method according to claim 1, characterized in that After the step of generating a network fault solution based on the network fault analysis result by using the second largest language model, the method further includes: Obtaining the fault handling result after executing the fault handling work order; generating feedback information when the fault handling result indicates that the network fault triggering the alarm information has not been resolved; The feedback information is input into the first large language model and the second large language model to update the first large language model and the second large language model.

9. A device for generating a network fault solution, characterized in that: Applied to a network system, the network system includes various network segments, and the device includes: an alarm information acquisition module, used for acquiring the alarm information of the network system, and inputting the alarm information into a first large language model, and determining the faulty network segment in each of the network segments through the first large language model; the first large language model is obtained by pre-training a general large language model using professional domain knowledge corpus data of each of the network segments; An analysis information acquisition module, used to perform fault analysis on the alarm information using each fault analysis model corresponding to the fault network segment, and obtain network fault analysis information output by each fault analysis model; An analysis result generating module, used for inputting the network fault analysis information output by each of the fault analysis models into a second language model, and generating a network fault analysis result through the second language model; the second language model is obtained by fine-tuning the first language model in advance using the business scenario knowledge corpus data of each of the network segments; A solution generation module is used to generate a network fault solution based on the network fault analysis result through the second largest language model when the network fault analysis result indicates that the alarm information is real alarm information; the network fault solution is used to assist in generating a fault handling work order for the alarm information.

10. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 8 are implemented.

11. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

12. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

Images