LTE log problem positioning method and device

Abstract

The embodiment of the invention provides a method and a device for positioning an LTE (Long Term Evolution) log problem. The method comprises the following steps: acquiring training set data and test set data from a knowledge base of long term evolution (LTE) log information; performing first training on the large language model according to preset hyper-parameters and the training set data, and performing second training on the large language model subjected to the first training according to the test set data; and performing problem positioning on LTE log information to be analyzed based on the large language model after the second training. Therefore, the problems of insufficient generalization ability and low processing efficiency of the LTE log processing method in the related technology can be solved, so that the LTE log processing efficiency is improved.

Description

TECHNICAL FIELD

[0001] Embodiments of the present application relate to the technical field of communication, in particular to a method and device for positioning LTE log problems. BACKGROUND

[0002] Large Language Models (LLMs) are an important advancement in the field of Natural Language Processing (NLP), which learn complex patterns and structures of language by training on large-scale text datasets. These large language models are typically based on the Transformer architecture and can handle a variety of natural language tasks.

[0003] Long-Term Evolution (LTE) is a standard for 4G mobile communication technology, and its log files usually contain a large amount of technical information for network troubleshooting, performance optimization, etc. The information in the LTE log file is usually huge and contains complex technical terms and formats.

[0004] The LTE log processing method in the related art has the following problems: 1) The LTE log file usually contains a large number of professional technical parameters and state information, which needs to be analyzed by professional engineers, and the processing efficiency is low; 2) Some automated tools can process log files, but they mostly rely on predefined rules and pattern matching, and have insufficient adaptability to unknown problems or complex scenarios, and low generalization ability; 3) Traditional log analysis tools may have low ability to understand the context information in the log: for example, they cannot accurately identify the association between network events mentioned in the log; 4) Traditional log analysis tools may only provide problem descriptions, but lack intelligent recommendations for possible solutions, and still need professional engineers to make judgments based on personal experience. SUMMARY

[0005] Embodiments of the present application provide a method and device for positioning LTE log problems to at least solve the problem of insufficient generalization ability and low processing efficiency of the LTE log processing method in the related art.

[0006] According to an embodiment of the present application, a method for positioning LTE log problems is provided, comprising: obtaining training set data and test set data from a knowledge base of Long-Term Evolution (LTE) log information; performing a first training on a large language model according to a preset hyperparameter and the training set data, and performing a second training on the large language model after the first training according to the test set data; and positioning problems in LTE log information to be analyzed based on the large language model after the second training.

[0007] According to another embodiment of the present application, a positioning device for LTE log problems is provided, comprising: an acquisition module configured to acquire training set data and test set data from a knowledge base of long term evolution (LTE) log information; a training module configured to perform a first training on a large language model according to preset hyperparameters and the training set data, and perform a second training on the large language model after the first training according to the test set data; and a positioning module configured to perform problem positioning on to-be-analyzed LTE log information based on the large language model after the second training.

[0008] According to still another embodiment of the present application, a computer readable storage medium is also provided, which stores a computer program, wherein the computer program is configured to perform the steps in any of the above method embodiments when executed.

[0009] According to still another embodiment of the present application, an electronic device is also provided, which comprises a memory and a processor, the memory stores a computer program, and the processor is configured to execute the computer program to perform the steps in any of the above method embodiments.

[0010] According to still another embodiment of the present application, a computer program product is also provided, which comprises a computer program and instructions, and the computer program and instructions are executed by a processor to implement the steps in any of the above method embodiments.

[0011] Through the above embodiments of the present application, the large language model based on the deep learning mechanism can understand complex technical terms and context information in the LTE log information, and can provide more accurate data analysis results. At the same time, by introducing hyperparameters to train the large language model, the large language model can output higher quality data analysis results. Therefore, the problem of insufficient generalization ability and low processing efficiency of the LTE log processing method in the related art can be solved, and the effect of improving the LTE log processing efficiency is achieved. BRIEF DESCRIPTION OF DRAWINGS

[0012] Figure 1 is a hardware structure block diagram of a computer terminal for implementing the positioning method for LTE log problems according to an embodiment of the present application;

[0013] Figure 2 is a flowchart of the positioning method for LTE log problems according to an embodiment of the present application;

[0014] Figure 3 is a flowchart of the large language model training method according to an embodiment of the present application;

[0015] Figure 4 is a flowchart of the LTE log problem positioning method based on the trained large language model according to an embodiment of the present application;

[0016] Figure 5 is a schematic diagram of an LTE log analysis method based on a large language model according to an embodiment of the present application;

[0017] Figure 6 is a schematic diagram of effect comparison of LTE log analysis under different methods according to an embodiment of the present application;

[0018] Figure 7 is a structural block diagram of an LTE log problem positioning device according to an embodiment of the present application. DETAILED DESCRIPTION

[0019] Hereinafter, embodiments of the present application will be described in detail with reference to the accompanying drawings and in conjunction with embodiments.

[0020] It should be noted that the terms "first", "second" and the like in the specification and claims of the present application and the above-described drawings are used to distinguish similar objects, and do not necessarily have to describe a specific order or sequence.

[0021] The method embodiments provided in the embodiments of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Taking the case of running on a computer terminal, Figure 1 is a hardware structural block diagram of a computer terminal running an LTE log problem positioning method according to an embodiment of the present application. As Figure 1 shown, the computer terminal can include one or more (only one is shown in Figure 1 ) processor 102 (the processor 102 can include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA) and a memory 104 for storing data, wherein the above-mentioned computer terminal can also include a transmission device 106 for communication function and an input and output device 108. Those skilled in the art can understand, Figure 1 the structure shown is only schematic, which does not limit the structure of the above-mentioned computer terminal. For example, the computer terminal can also include more or fewer components than Figure 1 shown, or have a different configuration from Figure 1 shown.

[0022] The memory 104 can be used to store computer programs, such as software programs of application software and modules, such as a computer program corresponding to the positioning method of the LTE log problem in the embodiments of the present application. The processor 102 performs various functional applications and data processing, i.e., implements the above method, by running the computer programs stored in the memory 104. The memory 104 can include a high-speed random access memory, and can further include a non-volatile memory, such as one or more magnetic storage devices, flash memories, or other non-volatile solid-state memories. In some examples, the memory 104 can further include a memory remotely arranged with respect to the processor 102, which can be connected to the computer terminal through a network. Examples of the above network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof.

[0023] The transmission device 106 is used to receive or send data via a network. Specific examples of the above network can include a wireless network provided by a communication provider of the computer terminal. In one example, the transmission device 106 includes a network adapter (NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 can be a radio frequency (RF) module, which is used to communicate with the Internet in a wireless manner.

[0024] In the related art, the LTE log file processing method includes the following steps:

[0025] Step S1: A professional engineer determines the positioning idea of the problem by means of log problem description, personal experience and logical reasoning;

[0026] Step S2: According to the positioning idea obtained, specific information is extracted from the log, which contains specific technical parameters and state information, such as "keyword a" contains "network event A", etc.

[0027] Step S3: The extracted information is viewed, and the log problem is analyzed by means of personal experience, logical reasoning and a plurality of automatic tools relying on predefined rules and pattern matching.

[0028] It can be seen that the above method has high dependence on professionals, and in addition, it has problems such as high complexity, insufficient generalization ability, insufficient context understanding, lack of intelligent decision support, etc.

[0029] For LTE log problem analysis, a large language model can provide a new perspective and method. Based on deep learning mechanism, the large language model has the potential advantage of strong semantic understanding and pattern recognition ability when dealing with the above problems. After training, the large language model can understand technical terms and context information, can process log data in real time and accurately analyze log files, can provide deeper insights, and can also provide suggestions for problem solving. In addition, the real-time analysis capability of the large language model is also crucial for quick response to network failures.

[0030] In the present embodiment, a positioning method for LTE log problems running on the above computer terminal is provided, Figure 2 is a flowchart of the positioning method for LTE log problems according to an embodiment of the present application, as Figure 2 shown, the flow includes the following steps:

[0031] Step S202, obtaining training set data and test set data from a knowledge base of long term evolution LTE log information;

[0032] In one embodiment, the LTE log information is sorted and constructed, and the training set data and test set data are divided. The large language model is trained and fine-tuned using transfer learning technology. The fine-tuning can be full fine-tuning on the entire architecture of the large language model, or only adjusting a part of the large language model (such as the classification layer or the parameters of a specific layer), and evaluating the accuracy of the output data to ensure that the model can effectively run in various situations.

[0033] In one embodiment, the hyperparameters include at least one of the following: temperature coefficient, number of data matches in single training of input data, text similarity matching algorithm, and prompt engineering parameters.

[0034] The temperature coefficient (T) can be set by the user, for example, T is 8.0, 4.0, 2.0, 1.0, or 0.5.

[0035] The temperature coefficient is used to determine the probability q i by the following formula: i The probability q

[0036]

[0037] The number of data matches in single training of input data is the maximum number of items in the returned list when a query request is submitted in a text database. The maximum number, i.e., the number of data matches, can be set by the user or adjusted according to specific application scenarios and requirements.

[0038] Text similarity matching algorithm for determining whether to assign matching weights by cosine similarity, Euclidean distance, or inner product.

[0039] The prompt engineering parameter (i.e., the Prompt engineering parameter) is determined by the log content to be processed, uses delimiters (such as single quotes, double quotes, angle brackets, etc.) to clearly indicate different parts of the input, and explicitly indicates the required output format (for example, JSON format), and by repeatedly analyzing and refining the prompt to generate the output required by the symbol, ensuring that sufficient context and detailed information are provided when asking questions to enable the large language model to understand and generate relevant answers.

[0040] Before step S202 of the embodiment, the method further comprises: preprocessing the historical data samples of the LTE log to obtain a first log problem description, a first log problem positioning process, and a first log problem positioning result; and constructing a knowledge base of the LTE log information according to log features extracted from the log content, the first log problem description, the first log problem positioning process, and the first log problem positioning result. The preprocessing includes at least one of the following: removing irrelevant information, text normalization, removing stop words, and data deduplication.

[0041] Irrelevant information refers to information that is usually not of actual significance to text analysis, but instead interferes with the learning process of the model, leading to inaccurate analysis results. It includes HTML tags, URLs, special symbols, and the like. Removing irrelevant information helps reduce noise and enables the model to focus on the text content itself.

[0042] Text normalization refers to processing that ensures data consistency, making it easier for the model to understand and compare different sources of text. It includes unifying case, expanding abbreviations, and converting character encodings. For example, converting all English text to lowercase can reduce the size of the vocabulary and simplify the learning task of the model.

[0043] Removing stop words refers to the removal of frequently occurring words (such as "of", "is", "in", etc.) that contribute less to the meaning of the text. Removing stop words can reduce the dimensionality of the data and improve processing speed, making the model more focused on informative words and improving its predictive performance.

[0044] Data deduplication, or the removal of duplicate text entries, does not provide additional information to the model and instead wastes computational resources. Deduplication can reduce the size of the data set and improve the efficiency of model training.

[0045] In one embodiment, the knowledge base of the LTE log information is constructed, including: determining first training data and first test data according to the log features and the first log problem description, determining second training data and second test data according to the first log problem positioning process and the first log problem positioning result; constructing the knowledge base according to the first training data, the first test data, the second training data and the second test data, wherein the first training data and the second training data have a corresponding relationship; and the first test data and the second test data have a corresponding relationship.

[0046] In step S204, the large language model is trained for the first time according to the preset hyperparameters and the training set data, and the large language model after the first training is trained for the second time according to the test set data.

[0047] In step S204 of the embodiment, the large language model is trained for the first time according to the preset hyperparameters and the training set data, including: inputting the first training data in the training set data into the large language model to obtain first output data; adjusting the general parameters corresponding to the large language model by a transfer learning method according to the initial value of the hyperparameters, the first output data and the second training data in the training set data, to train the large language model for the first time.

[0048] In step S204 of the embodiment, the large language model is trained for the second time according to the test set data after the first training, including: inputting the first test data in the test set data into the large language model to obtain second output data; comparing the second output data with the second test data in the test set data; and in the case that the difference value between the second output data and the second test data in the test set data is greater than a preset threshold, adjusting the initial value of the hyperparameters to train the large language model after the first training for the second time.

[0049] Figure 3 is a flowchart of the large language model training method according to an embodiment of the present application, as shown in Figure 3 The method comprises the following steps:

[0050] Step S301, data preprocessing.

[0051] Specifically, the historical data samples i of the LTE log are preprocessed to improve the data quality, wherein the data preprocessing includes removing irrelevant information, text normalization, removing stop words, data deduplication, etc.

[0052] Step S302, knowledge base construction and arrangement of LTE log information.

[0053] Specifically, after data preprocessing on the historical data sample i of the LTE log, the log problem description x is obtained 1i , the log problem positioning process y 1i , and the log problem positioning result y 2i .

[0054] The log feature content x 0i extracted from the log content x 2i , let X i = [x 1i , x 2i ], Y i = [y 1i , y 2i ], X = [X1, X2, …, X i ], Y = [Y1, Y2, …, Y i ] as the input of the large language model, each pair of X i , Y i correspond to each other, and the knowledge base construction and sorting of the LTE log information are performed.

[0055] Step S303, setting hyperparameters. The hyperparameters include a temperature coefficient, a data matching number of input data in a single training, a text similarity matching algorithm, and a prompt engineering parameter. The data matching number of input data in a single training, that is, the single matching number, is the number of X i between the input X in this training and the input X in the previous training. The matching weight of the word matching number is distributed by cosine similarity, Euclidean distance, or inner product.

[0056] Step S304, fine-tuning of the large language model; including first training and second training.

[0057] Specifically, the large language model is trained for the first time according to the preset hyperparameters and the training set data.

[0058] For example: assuming that the total number of samples in the knowledge base is M, n% (10≤n≤90) of the data in the knowledge base is extracted as the training set data, wherein the training set data includes first training data X train = [X0, X1, …, X n-1 ] and second training data Y train = [Y0, Y1, …, Y n-1 ]; the first training data is input into the large language model to obtain first output data; the large language model is trained and fine-tuned by using the transfer learning technology according to the initial value of the hyperparameters, the first output data, and the second training data.

[0059] In the embodiment, the large language model is trained and fine-tuned by the initial values of the hyperparameters, the first output data, and the second training data, including: adjusting (fine-tuning) the general parameters of the large language model according to the first output data output by the large language model, so that the output data of the large language model approximates the second training data, and when the approximation degree of the output data of the large language model to the second training data meets the first preset threshold, the first training of the large language model is ended, wherein the general hyperparameters are determined by the type of the large language model.

[0060] In the embodiment, the hyperparameters are pre-set according to the content of the LTE log to be processed.

[0061] Specifically, the large language model is trained for the second time according to the test set data.

[0062] For example: assuming that the total amount of samples in the knowledge base is M, then extract n% (10≤n≤90) data in the knowledge base as training set data, and the remaining (100-n)% data as test set data, wherein the test set data includes first test data X test =[X n , X n+1 ,..., X 99 ] and second test data Y test =[Y n , Y n+1 ,..., Y 99 ]; input the first test data into the large language model which has completed the first training to obtain second output data; compare the similarity of the second output data and the second test data, and adjust the large language model based on the similarity of the second output data and the second test data.

[0063] Step S305, judging whether the output data meets the expectation.

[0064] Specifically, the large language model is adjusted based on the similarity of the second output data and the second test data, including: in the case that the similarity of the second output data and the second test data is lower than the second preset threshold, adjusting the initial values of the hyperparameters of the large language model to make the second output data approximate the second test data, until the similarity of the second output data and the second test data is higher than or equal to the second preset threshold.

[0065] Wherein, the similarity of the second output data and the second test data can be compared by manual judgment, and the accuracy of the output data of the large language model after the second training can be determined through the similarity between the second output data and the second test data, so as to ensure that the large language model can work effectively in various situations.

[0066] In this embodiment, if the output data is too different from the expected result, the initial value of the hyperparameters of the large language model is adjusted: at least one of the temperature coefficient, the number of data matches in a single training of the input data, the text similarity matching algorithm, and the prompt engineering parameters is adjusted to obtain output data with higher quality.

[0067] Through the above two model training, a large language model with stronger generalization ability, higher output quality and more relevant target task can be obtained.

[0068] Step S306, if the output data is less different from the expected result (i.e., meets the expectation), output the trained large language model.

[0069] Step S206, based on the large language model after the second training, perform problem positioning on the LTE log information to be analyzed.

[0070] Figure 4 is a flowchart of an LTE log problem positioning method based on a trained large language model according to an embodiment of the present application, as shown in Figure 4 The method comprises the following steps:

[0071] Step S401, deploy the trained large language model to a production environment.

[0072] Step S402, obtain real-time LTE log information to be analyzed.

[0073] Step S403, perform data cleaning from the LTE log information to be analyzed, obtain log problem description, problem positioning process and problem positioning result, and extract log feature content from the LTE log content.

[0074] Step S404, input the log feature content, log problem description, problem positioning process and problem positioning result into the deployed large language model.

[0075] Step S405, based on the performance of the large language model and user feedback, iteratively optimize the general parameters of the large language model, and further adjust the hyperparameters to improve the accuracy of the large language model analysis and the quality of intelligent decision-making.

[0076] The large language model also has intelligent decision support function, i.e., it can provide solutions based on historical data and pattern recognition, which is of great significance to improve the efficiency of network operation and reduce cost.

[0077] In step S206 of the embodiment, the LTE log information to be analyzed is positioned based on the large language model after the second training, including: acquiring a second log problem description from the LTE log information to be analyzed by the large language model, and acquiring a second log problem positioning process according to the second log problem description, wherein the second log problem positioning process includes: acquiring abnormal situation information from the second log problem description; determining an abnormal reason corresponding to the abnormal situation according to reference information stored in a knowledge base; and determining a second log problem positioning result according to the second log problem positioning process.

[0078] Through the above steps, through the above embodiment of the application, the large language model can understand complex technical terms and context information in the LTE log information based on the deep learning mechanism, and can provide more accurate data analysis results. At the same time, by introducing hyperparameters to train the large language model, the large language model can output higher quality data analysis results. Therefore, the problem of insufficient generalization ability and low processing efficiency of the LTE log processing method in the related art can be solved, and the LTE log processing efficiency can be improved.

[0079] Figure 5 is a schematic diagram of a large language model-based LTE log analysis method according to an embodiment of the application. As shown in Figure 5 , the method includes two processes of training and testing: the first process is to fine-tune the large language model based on historical data, and the second process is to identify analysis, intelligent decision-making and deployment iteration based on new data. Among them, the fine-tuned large language model in the first process is the large language model deployed in the second process.

[0080] (1) Fine-tune the large language model based on historical data: first, clean the historical data samples to improve data quality, including removing irrelevant information, text normalization, removing stop words, and data deduplication. Data preprocessing obtains log content (further processing obtains log features), log problem description (i.e. test phenomenon description), and log problem positioning process and log problem positioning result, and inputs the preprocessed data into the large language model for training, wherein the preprocessed data is text_x1, text_x2, text_y1, and text_y2 in 5. Use transfer learning technology to train and fine-tune the large language model, and evaluate the accuracy of the output data to ensure that the model can work effectively in various situations.

[0081] (2) Based on new data recognition analysis, intelligent decision-making and deployment iteration: the trained large language model is deployed in the production environment to analyze real-time log data. For new data outside the sample, first extract the log feature content (X2) from the log content, and perform data cleaning. The log problem description (i.e. X1), (log) problem positioning process (Y1), and (log) problem positioning result (Y2) are described. Input X1, X2, Y1, and Y2 into the deployed large language model. According to the performance of the large language model and user feedback, the model parameters are continuously iterated and optimized to improve the accuracy of the analysis and the quality of the intelligent decision-making.

[0082] In this embodiment, another key technical feature of the large language model is its intelligent decision-making support function. The deep semantic understanding ability of the large language model not only enables it to identify problems, but also provides solutions based on historical data and pattern recognition, greatly improving the efficiency and accuracy of troubleshooting, and thus has important significance for improving the efficiency and reducing the cost of network operation.

[0083] Figure 6 is a schematic diagram of the effect comparison of LTE log analysis under different methods according to the embodiments of the present application, as shown in Figure 6 Compared with the traditional method, the LTE log problem positioning method provided by the embodiments of the present application takes less time and provides more detailed analysis results. Not only can it quickly identify network fault descriptions, but also can provide the causes of the fault and solutions, so the method proposed in the embodiments of the present application is more detailed, reasonable and fast. Among them, Figure 6 The "1. Extract 'abnormal situation' from 【phenomenon X]' and '2. Refer to the information in the knowledge base to give the reason for the occurrence of 'abnormal situation'" in the analysis result in

[0084] The embodiments of the present application provide a large language model-based LTE log problem positioning method and principle. In addition to being applicable to LTE network environments, it can also be applied to 5G communication log problem analysis and positioning. The large language model can be used to assist in analyzing LTE log files, quickly identify network fault descriptions, and provide solutions to help engineers locate log problems. The large language model can be used as part of technical support by fine-tuning the log data knowledge base built by different engineers to provide personalized solutions.

[0085] Large language models also exhibit significant advantages in terms of user experience and cost-effectiveness. They can provide a more user-friendly interface through natural language processing technology, allowing non-experts to easily query and understand log information. In addition, although the initial development and training costs are high, in the long run, large language models can significantly reduce the need for manual analysis, thereby reducing operating costs. At the same time, through scalable design, it ensures that the analysis capability remains efficient as the amount of network data continues to grow. The combination of these technical features indicates the broad application prospects of large language models in LTE log analysis.

[0086] Through the description of the above embodiments, those skilled in the art can clearly understand that the method according to the above embodiments can be realized by means of software and the necessary general hardware platform, of course, it can also be realized by hardware, but in many cases the former is a better embodiment. Based on such understanding, the technical solutions of the present application or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes a plurality of instructions for making a terminal device (which can be a mobile phone, computer, server, or network device, etc.) execute the methods described in various embodiments of the present application.

[0087] In this embodiment, an LTE log problem positioning device is also provided, which is used to implement the above embodiments and preferred embodiments, and will not be described again. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, hardware or a combination of software and hardware implementation is also possible and contemplated.

[0088] Figure 7 is a structural block diagram of the LTE log problem positioning device according to the embodiments of the present application, as Figure 7 shown, the LTE log problem positioning device 700 includes an acquisition module 710, a training module 720, and a positioning module 730.

[0089] The acquisition module 710 is configured to acquire training set data and test set data from a knowledge base of long term evolution (LTE) log information.

[0090] The training module 720 is configured to perform a first training on a large language model according to preset hyperparameters and the training set data, and perform a second training on the large language model after the first training according to the test set data.

[0091] The positioning module 730 is configured to perform problem positioning on LTE log information to be analyzed based on the large language model after the second training.

[0092] It should be noted that the above modules can be implemented by software or hardware, and the hardware can be implemented in the following manner, but is not limited thereto: all the modules are located in the same processor; or the modules are located in different processors in any combination.

[0093] The embodiment of the present application further provides a computer readable storage medium, which stores a computer program, and the computer program is arranged to execute the steps in any of the method embodiments when running.

[0094] In an example embodiment, the computer readable storage medium can include, but is not limited to, a U disk, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk, and various media that can store computer programs.

[0095] The embodiment of the present application further provides an electronic device, which comprises a memory storing a computer program and a processor arranged to execute the computer program to perform the steps in any of the method embodiments.

[0096] In an example embodiment, the electronic device can further comprise a transmission device connected to the processor and an input and output device connected to the processor.

[0097] The embodiment of the present application further provides a computer program product, which comprises a computer program and instructions, and the computer program and instructions are executed by a processor to implement the steps in any of the method embodiments.

[0098] The specific examples in the embodiment can refer to the examples described in the above embodiments and example embodiments, and the embodiment will not be described here.

[0099] Obviously, those skilled in the art should understand that the modules or steps of the present application described above can be realized by general computing devices, which can be concentrated on a single computing device or distributed on a network composed of multiple computing devices, and they can be realized by program codes executable by computing devices, so that they can be stored in storage devices and executed by computing devices, and in some cases, the steps shown or described can be executed in different order, or they can be manufactured into individual integrated circuit modules, or multiple modules or steps can be manufactured into a single integrated circuit module. Thus, the present application is not limited to any specific combination of hardware and software.

[0100] The above description is only the preferred embodiment of the present application, and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. within the principles of the present application should be included in the protection scope of the present application.

Claims

1. A method for locating an LTE log problem, the method comprising: The method comprises the following steps: obtaining training set data and test set data from a knowledge base of long term evolution (LTE) log information; first training a large language model according to preset hyperparameters and the training set data, and second training the large language model after the first training according to the test set data; performing problem positioning on to-be-analyzed LTE log information based on the large language model after the second training.

2. The method of claim 1, wherein, In the method, the hyperparameters comprise at least one of a temperature coefficient, a data matching quantity of input data in single training, a text similarity matching algorithm, and a prompt engineering parameter.

3. The method of claim 1, wherein, Before obtaining the training set data and the test set data from the knowledge base of LTE log information, the method further comprises the following steps: preprocessing historical data samples of LTE logs to obtain first log problem descriptions, first log problem positioning processes, and first log problem positioning results; constructing the knowledge base of LTE log information according to log features extracted from log content, the first log problem descriptions, the first log problem positioning processes, and the first log problem positioning results.

4. The method of claim 3, wherein, In the method, the preprocessing comprises at least one of removing irrelevant information, text normalization, removing stop words, and data deduplication.

5. The method of claim 3, wherein, Constructing the knowledge base of LTE log information comprises the following steps: determining first training data and first test data according to the log features and the first log problem descriptions, and determining second training data and second test data according to the first log problem positioning processes and the first log problem positioning results; constructing the knowledge base according to the first training data, the first test data, the second training data, and the second test data, wherein the first training data and the second training data have a corresponding relationship; and the first test data and the second test data have a corresponding relationship.

6. The method of claim 1, wherein, The first training a large language model according to preset hyperparameters and the training set data comprises the following steps: inputting first training data in the training set data into the large language model to obtain first output data; adjusting general parameters of the large language model by a transfer learning method according to initial values of the hyperparameters, the first output data, and second training data in the training set data, to perform the first training on the large language model.

7. The method of claim 1, wherein, The second training the large language model after the first training according to the test set data comprises the following steps: inputting first test data in the test set data into the large language model to obtain second output data; comparing the second output data with second test data in the test set data; adjusting the initial values of the hyperparameters to perform the second training on the large language model after the first training, in a case where a difference value between the second output data and the second test data in the test set data is greater than a preset threshold.

8. The method of claim 1, wherein, The problem positioning on to-be-analyzed LTE log information based on the large language model after the second training comprises the following steps: The second log problem description is obtained from the LTE log information to be analyzed by the large language model, and a second log problem positioning process is obtained according to the second log problem description, wherein the second log problem positioning process comprises: obtaining abnormal situation information from the second log problem description; determining an abnormal reason corresponding to the abnormal situation according to reference information stored in a knowledge base; A second log problem positioning result is determined according to the second log problem positioning process.

9. A device for locating LTE log problems, characterized in that, Comprise: An acquisition module is configured to acquire training set data and test set data from a knowledge base of long term evolution (LTE) log information; A training module is configured to perform first training on a large language model according to preset hyperparameters and the training set data, and perform second training on the large language model after the first training according to the test set data; A positioning module is configured to perform problem positioning on LTE log information to be analyzed based on the large language model after the second training.

10. A computer-readable storage medium, characterized in that, The computer program is stored in the computer readable storage medium, and when the computer program is executed by the processor, the steps of the method in any one of claims 1 to 8 are implemented. 11.An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein, The processor executes the computer program to implement the steps of the method in any one of claims 1 to 8.

12. A computer program product comprising a computer program, instructions, characterized in that, The computer program, instructions are executed by the processor to implement the steps of the method in any one of claims 1 to 8.

Images