Log data processing methods, electronic device and computer-readable storage medium

Abstract

The present disclosure relates to the technical field of large models and cloud computing. Disclosed are log data processing methods, an electronic device and a computer-readable storage medium. A method comprises: acquiring log data from a physical server, the log data at least comprising: at least one type of exception log collected during a historical time period; performing data segmentation on the log data to obtain a segmentation result; and using a target fault prediction model to perform fault prediction on the segmentation result, so as to obtain a target fault prediction result, the target fault prediction model being used for predicting whether the physical server will crash within a future time period due to a fault. The present disclosure solves the technical problem in the prior art that predicting physical machine faults on the basis of BERT models causes high computation complexity, long computation consumption time and low accuracy of prediction results.

Description

[0001] TECHNICAL FIELD: The present disclosure relates to the fields of large model technology and cloud computing technology, and more specifically, to a log data processing method, electronic device, and computer-readable storage medium. Background: In cloud computing systems, a large number of virtual machines (VMs) run on physical machines (NCs). NC failures can severely impact the availability of the VMs, and thus the stability of the cloud computing system. Currently, the LogBERT model has been proposed based on the BERT model. The LogBERT model can predict faults based on system exception logs generated by NCs, thereby predicting NC failures. However, pre-training this model is very time-consuming and computationally complex. Furthermore, the XgBoost model has been proposed. This model can also predict faults based on system exception logs generated by NCs, but this model only uses statistical features of the number of times system exception logs are generated, which is computationally very time-consuming. Furthermore, since the statistical features only contain information about the frequency of system exception log occurrences, the prediction accuracy is low. Currently, no effective solution has been proposed to address the above-mentioned issues. SUMMARY OF THE INVENTION Embodiments of the present disclosure provide a log data processing method, electronic device, and computer-readable storage medium to at least address the technical issues in the related art of predicting physical machine faults based on the BERT model, which result in high computational complexity, long computation time, and low prediction accuracy. According to one aspect of an embodiment of the present disclosure, a log data processing method is provided, comprising: obtaining log data from a physical server, wherein the log data includes at least at least one type of abnormal log collected within a historical time period; performing data segmentation on the log data to obtain segmentation results; and performing fault prediction on the segmentation results using a target fault prediction model to obtain a target fault prediction result, wherein the target fault prediction model is used to predict whether the physical server will crash due to a fault within a future time period.According to another aspect of an embodiment of the present disclosure, a log data processing method is also provided, including: obtaining a fault prediction request through a first application programming interface, wherein the request data carried in the fault prediction request includes: log data obtained from a physical server, the log data including at least: at least one type of abnormal log collected within a historical time period; returning a fault prediction response through a second application programming interface, wherein the response data carried in the fault prediction response includes: a target fault prediction result, wherein the target fault prediction result is obtained by performing fault prediction on a segmentation result using a target fault prediction model, the segmentation result is obtained by performing data segmentation on the log data, and the target fault prediction model is used to predict whether the physical server will crash due to a fault within a future time period. According to another aspect of an embodiment of the present disclosure, a log data processing method is provided, comprising: obtaining a currently input fault prediction dialogue request, wherein the request data carried in the fault prediction dialogue request includes log data obtained from a physical server, wherein the log data includes at least at least one type of exception log collected within a historical time period; returning a fault prediction dialogue reply in response to the fault prediction dialogue request, wherein the information carried in the fault prediction dialogue reply includes a target fault prediction result, wherein the target fault prediction result is obtained by performing fault prediction on a segmentation result using a target fault prediction model, wherein the segmentation result is obtained by segmenting the log data, and the target fault prediction model is used to predict whether the physical server will crash due to a fault within a future time period; and displaying the target fault prediction result in a graphical user interface. According to another aspect of an embodiment of the present disclosure, an electronic device is provided, comprising: a memory storing an executable program; and a processor for executing the program, wherein when the program executes, any one of the aforementioned log data processing methods is executed. According to another aspect of an embodiment of the present disclosure, a computer-readable storage medium is provided. The computer-readable storage medium includes a stored executable program. When the executable program is executed, the device containing the computer-readable storage medium is controlled to execute any one of the aforementioned log data processing methods. According to another aspect of an embodiment of the present disclosure, a computer program product is provided. The computer program, when executed by a processor, implements any one of the aforementioned log data processing methods.In an embodiment of the present disclosure, log data, including at least one type of exception log collected within a historical time period, is acquired from a physical server. The log data is then segmented to divide long sequences in the log data into multiple shorter subsequences. Finally, a target fault prediction model is used to perform fault prediction on the multiple shorter subsequences, resulting in a target fault prediction result indicating whether the corresponding physical server is about to crash. This achieves the goal of accurately predicting faults based on extremely long exception logs reported by the physical server, thereby reducing the computational complexity and computation time of the model, improving the computational efficiency of the model, and enhancing the prediction accuracy of the model, as well as improving the stability of the cloud computing system. This addresses the technical problem in related art of predicting physical machine faults based on the BERT model, which results in high computational complexity, long computation time, and low prediction accuracy. It should be noted that the general description above and the detailed description that follow are merely examples and explanations of the present disclosure and do not constitute limitations of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS The accompanying drawings described herein are intended to provide a further understanding of the present disclosure and constitute a part of the present disclosure. The exemplary embodiments of the present disclosure and their descriptions are intended to explain the present disclosure and do not constitute undue limitations of the present disclosure. In the accompanying drawings: Figure 1 is a schematic diagram of an application scenario of a log data processing method according to Example 1 of the present disclosure; Figure 2 is a flow chart of a log data processing method according to Example 1 of the present disclosure; Figure 3 is a flow chart of a long sequence processing method according to Example 1 of the present disclosure; Figure 4 is a flow chart of a log data processing method according to Example 2 of the present disclosure; Figure 5 is a flow chart of a log data processing method according to Example 3 of the present disclosure; Figure 6 is a schematic diagram of the structure of a log data processing device according to Example 4 of the present disclosure; Figure 7 is a schematic diagram of the structure of another log data processing device according to Example 4 of the present disclosure; Figure 8 is a schematic diagram of the structure of yet another log data processing device according to Example 4 of the present disclosure; Figure 9 is a block diagram of the structure of a computer terminal according to an embodiment of the present disclosure. DETAILED DESCRIPTION To help those skilled in the art better understand the disclosed solutions, the following will provide a clear and complete description of the technical solutions in the disclosed embodiments, in conjunction with the accompanying drawings. Obviously, the described embodiments represent only a portion of the disclosed embodiments, and are not exhaustive. Based on the disclosed embodiments, all other embodiments devised by persons of ordinary skill in the art without inventive effort should fall within the scope of protection of the disclosed embodiments. It should be noted that the terms "first," "second," and so on, in the specification and claims of the disclosed embodiments, and in the accompanying drawings, are used to distinguish similar objects and are not necessarily used to describe a specific order or precedence.It should be understood that the terms used in this manner are interchangeable where appropriate, such that the embodiments of the present disclosure described herein can be implemented in sequences other than those illustrated or described herein. Furthermore, the terms "including," "having," and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or apparatus comprising a series of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, product, or apparatus. The technical solutions provided by this disclosure are primarily implemented using large-scale model technology. Large-scale models herein refer to deep learning models with large-scale model parameters, typically including hundreds of millions, tens of billions, hundreds of billions, trillions, or even more than ten trillion model parameters. Large models, also known as cornerstone models or foundation models, are pre-trained on large, unlabeled corpora, producing pre-trained models with over 100 million parameters. These models are adaptable to a wide range of downstream tasks and exhibit good generalization capabilities. Examples include large language models (LLMs) and multi-modal pre-training models. It should be noted that in practical applications, large models can be fine-tuned using a small number of samples, allowing them to be applied to different tasks. For example, large models are widely applicable in fields such as natural language processing (NLP), computer vision, and speech processing. Specifically, they can be applied to computer vision tasks such as visual question answering (VQA), image captioning (IC), and image generation. They can also be widely used in natural language processing tasks such as text-based sentiment classification, text summarization, and machine translation. Therefore, the main application scenarios of large models include but are not limited to digital assistants, intelligent robots, search, online education, office software, e-commerce, intelligent design, etc. First, some nouns or terms that appear in the description of the embodiments of this disclosure are subject to the following explanations: Network Computer (NC): This refers to a physical server and is a single unit in a cloud computing system that counts failures and downtime.

[0002] BERT (Bidirectional Encoder Representations from Transformers) model: A bidirectional encoding Transformer model, commonly used in natural language processing and image processing tasks.

[0003] XgBoost (extreme Gradient Boosting) model: an ensemble learning model based on decision trees.

[0004] The Transformer model: A machine learning model based on the attention mechanism, commonly used in natural language processing and image processing tasks. The attention mechanism is used to mine correlation information between elements in a sequence. In cloud computing systems, a large number of virtual machines (VMs) run on the NC. NC failures can significantly impact VM availability. Therefore, predicting NC failures and implementing timely maintenance are crucial to the stability of cloud computing systems. Different NC failure causes generate different system exception log patterns (such as type and frequency), which in turn generate different system exception logs on the NC. For example, monitors record NC operating system logs. System exception logs provide important information for predicting NC failures. Therefore, learning the system exception log patterns of existing faulty NCs to predict NC failures is key to solving the problem. However, because NCs can have multiple failure causes, they generate a large number of exception logs. This often results in a large number of exception events being reported in a short period of time, resulting in extremely long exception logs. These different types of exception events may be highly correlated with the occurrence of the failure. Furthermore, NCs also report a large number of redundant exception logs. Numerous critical and redundant exceptions are often mixed together, resulting in extremely long exception logs. Therefore, a powerful feature extractor is needed to capture key system exception log patterns and learn semantic information from them.

[0005] The Transformer model has powerful capabilities for mining semantic information in continuous sequence data. The attention computation model can highly summarize key sequence information to obtain deep semantic information, thereby capturing important information from the entire input sequence. However, the time and space complexity of traditional attention computation methods scales squarely with the length of the input sequence, significantly limiting the model's input sequence length. In the case of downtime prediction based on exception logs, extremely long exception sequence inputs result in extremely high computational costs, making it impossible to meet the requirements of online real-time prediction and even difficult to deploy offline due to resource constraints. For example, the LogBERT model, currently proposed based on the BERT model, uses the BERT model as a feature extractor. It first constructs an auxiliary learning task using a large amount of abnormal log text data. For example, masking a word in the text allows the BERT model to guess the masked word, or predicting whether two log text data are reported consecutively. This pre-trains the BERT model, allowing it to learn the composition logic of the log text data. The model is then used for learning and testing fault prediction tasks. However, there are the following disadvantages: (1) Pre-training is very time-consuming; (2) The BERT model is highly complex, and directly processing log text data requires complex encoding steps, which is inefficient; (3) It cannot process extremely long log text data. In addition, the XgBoost model is proposed. The XgBoost model counts the number of times each abnormal log is generated by each NC within a fixed time window, uses the number as the statistical feature of the NC, and then feeds this feature into the classifier of the XgBoost model for training, and then tests it online. However, there are the following disadvantages: (1) A large amount of statistical feature calculation is required for extremely long log text data, which is very time-consuming; (2) The statistical features only contain the frequency information of abnormal log generation, and the learned system abnormal log pattern is limited. It can be seen that the related technology has the following defects in processing system abnormal logs to predict NC failures. Drawback 1: Model pre-training is very time-consuming, the model complexity is high, the computation time is long, and the computational efficiency is low. Drawback 2: It cannot process extremely long log texts, that is, it cannot perform calculations on long sequences. Drawback 3: The accuracy and recall of fault prediction are low. Prior to this disclosure, no effective solutions to these drawbacks had been proposed.Example 1 According to an embodiment of the present disclosure, a log data processing method is provided. It should be noted that the steps illustrated in the flowcharts of the accompanying drawings can be executed in a computer system, such as a set of computer-executable instructions. Furthermore, although the flowcharts illustrate a logical order, in some cases, the steps illustrated or described may be executed in a different order. Considering the large number of model parameters in a large model and the limited computing resources of mobile terminals, the log data processing method provided in the embodiment of the present disclosure can be applied to, but is not limited to, the application scenario shown in FIG1 . In the application scenario shown in FIG1 , the large model is deployed on a server 10. The server 10 can be connected to one or more client devices 20 via a local area network (LAN), a wide area network (WAN), the Internet, or other types of data networks. Client devices 20 herein may include, but are not limited to, smartphones, tablet computers, laptops, PDAs, personal computers, smart home devices, and in-vehicle devices. Client devices 20 can interact with users via a graphical user interface to invoke the large model and thereby implement the method provided in the embodiment of the present disclosure. In an embodiment of the present disclosure, a system consisting of a client device and a server can perform the following steps: the client device obtains log data from a physical server, performs data segmentation on the log data, and obtains segmentation results. The server performs fault prediction on the segmentation results using a target fault prediction model to obtain a target fault prediction result, and returns the target fault prediction result to the client device. It should be noted that if the operating resources of the client device meet the deployment and operating conditions of a large model, the present disclosure can be performed on the client device. In this operating environment, the present disclosure provides a log data processing method as shown in Figure 2. Figure 2 is a flow chart of a log data processing method according to Example 1 of the present disclosure. As shown in Figure 2, the method may include the following steps: Step S21, obtaining log data from a physical server, wherein the log data includes at least one type of exception log collected within a historical time period; Step S22, segmenting the log data to obtain a segmentation result; Step S23, performing fault prediction on the segmentation result using a target fault prediction model to obtain a target fault prediction result, wherein the target fault prediction model is used to predict whether the physical server will crash due to a fault within a future time period. A physical server is a physical machine (Network Computer, NC) running multiple virtual machines (VMs) and providing computing, storage, and network resources.Log data can be understood as system exception logs reported in real time when an NC system exception occurs, indicating the cause of the NC exception. It is understood that log data can include various types of exception logs. For example, log data can include exception logs caused by memory errors, disk errors, and central processing unit (CPU) errors, without limitation. In the disclosed embodiments, log data can be at least one type of exception log collected within a historical time period. For example, exception logs can be collected at 5-minute intervals and within a 3-day sampling window, thereby obtaining log data. Given that NC systems can generate a large number of exception logs due to various fault causes, it is common for a large number of exception logs to be reported in a short period of time, resulting in extremely long exception logs. Because processing extremely long exception logs increases computational complexity, the disclosed embodiments segment the log data to break the extremely long exception logs into shorter ones, thereby obtaining segmentation results. For example, log data can be abstracted into sequences, specifically, extremely long exception logs can be abstracted into extremely long exception sequences. Sequence splitting (Sequence Splitting) is then performed on the long sequence, dividing it into multiple subsequences to obtain segmentation results. It is understood that the lengths of the multiple subsequences obtained by segmentation are much smaller than the length of the long sequence. Therefore, by segmenting the long sequence into multiple subsequences and subsequently using a model to perform fault prediction on the resulting subsequences, the model's computational complexity can be significantly reduced, thereby improving its computational efficiency. The target fault prediction model can be understood as a model used to predict whether an NC will crash due to a fault within a future time period. It can be a Transformer model capable of processing long sequences, a large model, or other machine learning model, without limitation. In the disclosed embodiments, the segmentation results are input into the target fault prediction model, which then performs fault prediction on the segmentation results to obtain a target fault prediction result indicating whether the corresponding NC will crash. This effectively reduces the model's computational complexity and yields more accurate prediction results. That is, the target fault prediction model of the embodiment of the present disclosure can accurately predict whether the NC is about to crash based on the extremely long system abnormality log reported by the NC, so as to enable timely operation and maintenance and improve the stability of the cloud computing system.In an embodiment of the present disclosure, log data, including at least one type of exception log collected within a historical time period, is obtained from a physical server. The log data is then segmented to divide long sequences in the log data into multiple shorter subsequences. Finally, a target fault prediction model is used to perform fault prediction on the multiple shorter subsequences, thereby obtaining a target fault prediction result indicating whether the corresponding physical server is about to fail. As can be seen, the present disclosure transforms the fault prediction problem of exception logs into a natural language processing problem, abstracts extremely long exception logs into sequence form, and segments the long sequence to obtain multiple shorter subsequences. These subsequences are then used as input to a Transformer model. This reduces the model's computational cost and computation time, improving the model's computational efficiency and prediction accuracy, enabling timely operation and maintenance, and ultimately enhancing the stability of the cloud computing system. The log data processing method provided by the embodiments of the present disclosure can be applied, but is not limited to, to application scenarios involving the prediction of physical server failures in the fields of e-commerce services, educational services, legal services, medical services, conference services, social network services, financial product services, logistics services, and navigation services. For example, scenarios involving the prediction of NC failures for providing e-commerce services, the prediction of NC failures for providing educational services, the prediction of NC failures for providing legal services, and the prediction of NC failures for providing medical services, etc., are not limited here. According to the disclosed embodiments, log data, including at least one type of exception log collected within a historical time period, is acquired from a physical server. The log data is then segmented to divide long sequences in the log data into multiple shorter subsequences. Finally, a target fault prediction model is used to perform fault prediction on the multiple shorter subsequences, resulting in a target fault prediction result indicating whether the corresponding physical server is about to crash. This achieves the goal of accurately predicting faults based on extremely long exception logs reported by the physical server, thereby reducing the computational complexity and computation time of the model, improving the computational efficiency of the model, and enhancing the prediction accuracy of the model, as well as improving the stability of the cloud computing system. Furthermore, the present invention addresses the technical problem in related technologies of training a video generation model based on network data, which results in poor video quality generated by the trained video generation model and fails to meet user expectations.In an optional embodiment, in step S22, data segmentation is performed on the log data to obtain a segmentation result, including the following method steps: Step S221: Obtaining an abnormal event sequence corresponding to the log data, wherein the abnormal event sequence is obtained by arranging the type identifiers corresponding to at least one type of abnormal log in a preset order; Step S222: Segmenting the abnormal event sequence to obtain a segmentation result, wherein the segmentation result includes: a plurality of abnormal event subsequences obtained by segmenting the abnormal event sequence. The abnormal event sequence can be understood as an abnormal event sequence, which is obtained by arranging the type identifiers corresponding to at least one type of abnormal log in a preset order. The type identifier can be understood as an identifier used to identify different types of abnormal logs. For example, an integer can be used as the type identifier, for example, an abnormal log caused by a CPU error is identified as the integer 20, and an abnormal log caused by a memory error is identified as the integer 30, without limitation. It is understood that the types of abnormal logs and their corresponding identifiers can be further divided, without limitation. The preset order can be the chronological order in which the NC reports the exception logs. For example, the category identifiers corresponding to the exception logs reported by each NC can be arranged in reverse chronological order to form a sequence, thereby obtaining an exception event sequence. For example, the category identifiers corresponding to the exception logs reported by each NC can be arranged in reverse chronological order, with 5-minute intervals and a sampling window length of 3 days. Overlapping sampling is performed across the entire sequence to obtain an exception event sequence. This exception event sequence can be represented as: [start], 32, 91, >, 256, where [start] is a start symbol and has no specific meaning. The representation format of the exception event sequence is not limited here. In the disclosed embodiments, when performing data segmentation on log data to obtain the segmentation results, the exception logs can be abstracted into a sequence to obtain the exception event sequence corresponding to the log data, i.e., the long exception event sequence. The exception event sequence is then segmented to obtain multiple exception event subsequences, i.e., the segmentation results. This allows us to transform the fault prediction problem of exception logs into a natural language processing problem. We abstract extremely long exception logs into sequence inputs, segment the long sequences, and obtain multiple subsequence inputs as inputs to the Transformer model, thereby reducing computational costs and improving efficiency.In an optional embodiment, obtaining an abnormal event sequence corresponding to the log data in step S221 includes the following method steps: Step S2211: Data abstraction is performed on multiple abnormal logs contained in the log data to obtain multiple abnormal events; Step S2212: Based on preset abnormal event types, the multiple abnormal events are converted into type identifiers corresponding to at least one type of abnormal log to obtain a conversion result, where the preset abnormal event type is used to determine multiple log formats and the encodings corresponding to different log formats; Step S2213: Generate an abnormal event sequence based on the conversion result. In the disclosed embodiment, when obtaining an abnormal event sequence corresponding to the log data, data abstraction can be performed on the multiple abnormal logs contained in the log data to obtain multiple abnormal events, that is, each abnormal log is abstracted into an abnormal event in a one-to-one correspondence. Exemplarily, abnormal logs reported by the NC can be obtained in real time by a detector, and each abnormal log is abstracted into a phrase in a one-to-one correspondence, i.e., an abnormal event, based on expert knowledge and regular expressions. For example, the original log "mce: [Hardware Error]: Machine check events logged" can be abstracted into the abnormal event "dmesg_unrecover_mce", which is not limited here. The preset abnormal event type can be understood as a preset abnormal event library, which is used to determine multiple log formats and the encodings corresponding to different log formats. It is understood that the abnormal event library includes all abnormal event types that may be reported by the NC, and each abnormal event in the abnormal event library corresponds to an encoding result, which can be understood as a type identifier. Therefore, after obtaining multiple abnormal events, based on the preset abnormal event types, the multiple abnormal events can be converted into type identifiers corresponding to at least one type of abnormal log to obtain a conversion result. Exemplarily, each abnormal event can be converted into a number for training based on the abnormal event library, for example, the abnormal event "dmesg_unrecover_mce" can be converted into 32, which is not limited here. Finally, based on the obtained conversion results, an abnormal event sequence is generated, such as [start], 32, 91, ..., 256, which is not limited here.In an optional embodiment, in step S2213, generating an abnormal event sequence based on the conversion result includes the following method steps: Step S22131: Determining positive and negative samples based on the conversion result, wherein positive samples are samples that trigger a physical server crash, and negative samples are samples that do not trigger a physical server crash; Step S22132: Arranging the positive samples according to a preset order to obtain an arrangement result; Step S22133: Using the arrangement result and the negative samples to generate an abnormal event sequence. It is understood that multiple NCs operate in a cloud computing system, and among the multiple abnormal logs reported by each NC, there are both abnormal logs that may cause the NC to crash and abnormal logs that will not cause the NC to crash. In the disclosed embodiment, positive samples are samples that trigger a physical server crash, also known as crash samples, and negative samples are samples that do not trigger a physical server crash, also known as normal samples. It can also be understood that NCs predicted to crash are positive samples, and NCs predicted not to crash are negative samples. In the disclosed embodiment, when generating an abnormal event sequence based on the transformation results, positive and negative samples can be determined based on the transformation results. The positive samples are then arranged in a preset order to obtain an arrangement result. For example, the arrangement result can be obtained by arranging the positive samples in reverse chronological order of the reporting time of the abnormal logs corresponding to the positive samples. Finally, the negative samples are inserted into the arrangement result to obtain an abnormal event sequence. For example, for downtime samples (positive samples), only data with a sampling start within a preset time period at the time of the downtime can be retained, for example, only data with a sampling start within three days of the time of the downtime can be retained. For normal samples (negative samples), a certain percentage of samples can be randomly retained from the entire sequence to obtain an abnormal event sequence. This is not a limitation. In an optional embodiment, in step S222, the abnormal event sequence is segmented to obtain a segmentation result. The method includes the following steps: Step S2221: Sliding segmentation is performed on the abnormal event sequence using the length of the abnormal event subsequence as a segmentation window to obtain a segmentation result. In the disclosed embodiments, when segmenting an abnormal event sequence to obtain a segmentation result, the abnormal event sequence may be subjected to sliding segmentation using the length of the abnormal event subsequence as a segmentation window to obtain the segmentation result. For example, if the abnormal event sequence (i.e., the long sequence) is of length L, it may be subjected to sliding segmentation using windows of abnormal event subsequences (i.e., subsequences) of length m to obtain the segmentation result. This is not a limitation. It is understood that the segmentation result includes multiple abnormal event subsequences of length m.In an optional embodiment, the target fault prediction model includes a feature encoding sub-model, a feature extraction sub-model, a feature fusion sub-model, and a classifier. In step S23, the target fault prediction model is used to perform fault prediction on the segmentation results to obtain a target fault prediction result. The method includes the following steps: Step S231, embedding the segmentation results using the feature encoding sub-model to obtain an encoding result; Step S232, extracting features from the encoding result using the feature extraction sub-model to obtain an extraction result; Step S233, fusing features from the extraction result using the feature fusion sub-model to obtain a fusion result; and Step S234, classifying the fusion result using a classifier to obtain a target fault prediction result. In the disclosed embodiment, the target fault prediction model may include four components: the feature encoding sub-model, the feature extraction sub-model, the feature fusion sub-model, and the classifier. The feature encoding sub-model is used to encode log data. For example, the segmentation results obtained by segmenting the log data may be encoded, and multiple abnormal event sub-sequences may be encoded into vector form to obtain an encoding result. The feature extraction sub-model is used to extract features from the encoding results output by the feature encoding sub-model to obtain extraction results. Exemplarily, feature extraction can be performed using a Transformer model, but this is not a limitation here. Taking the encoder structure of the BERT model as an example, the encoding results of multiple abnormal event sub-sequences are passed through the BERT model to obtain feature extraction results corresponding to the abnormal event at each position in the sequence. The feature fusion sub-model is used to perform feature fusion on the extraction results output by the feature extraction sub-model, that is, to fuse the features corresponding to each abnormal event sub-sequence to obtain a fusion result. The classifier is used to classify the fusion results output by the feature fusion sub-model to obtain a target fault prediction result. It is understood that the classifier is used to classify the fusion results into positive samples indicating a predicted downtime fault and negative samples indicating a predicted non-downtime fault, thereby determining which NCs are likely to experience downtime in the future due to faults. In an optional embodiment, in step S231, the segmentation results are embedded-encoded using the feature encoding sub-model to obtain an encoded result. The method includes the following steps: Step S2311: Adding a sequence identifier to each abnormal event subsequence in the segmentation results; Step S2312: Embedding-encoding the segmentation results and the sequence identifier using the feature encoding sub-model to obtain an encoded result. The sequence identifier can be understood as an identifier for identifying each abnormal event subsequence. For example, a [cis] identifier can be added before each abnormal event subsequence.In the disclosed embodiment, when embedding the segmentation results using the feature encoding sub-model, a sequence identifier can be added to each abnormal event sub-sequence in the segmentation results. The segmentation results and the sequence identifiers are then embedded using the feature encoding sub-model to obtain an encoded result. For example, before training the feature encoding sub-model, a mapping matrix E with M > D can be randomly initialized, where M represents the type of abnormal event in the corpus and D represents the dimension of the mapped feature vector. During training, each abnormal event sub-sequence is converted into a feature vector according to the corresponding mapping and input into the feature encoding sub-model as model input. The gradient of the loss corresponding to the abnormal event sub-sequence is back-propagated to the corresponding position of matrix E for update. After training is complete, matrix E serves as part of the feature encoding sub-model parameters and can be used to encode the input. In an optional embodiment, in step S233, a feature fusion sub-model is used to perform feature fusion on the extraction results to obtain a fusion result, including the following method steps: Step S2331: Using the feature fusion sub-model, feature fusion is performed on the extracted features corresponding to the sequence identifiers included in the extraction results to obtain a fusion result. In the disclosed embodiment, when using the feature fusion sub-model to perform feature fusion on the extraction results, the feature fusion sub-model can be used to perform feature fusion on the extracted features corresponding to the sequence identifiers included in the extraction results, that is, to fuse the features corresponding to the sequence identifier [cis] of each abnormal event subsequence to obtain a fusion result. In an optional embodiment, in step S2331, a feature fusion sub-model is used to perform feature fusion on the extracted features corresponding to the sequence identifier to obtain a fusion result. The method includes the following steps: Step S23311: Using the feature fusion sub-model, an attention weight corresponding to the sequence identifier is obtained, and a weighted fusion is performed on the subsequence extracted features corresponding to the sequence identifier and the attention weight corresponding to the sequence identifier to obtain a fusion result. The attention weight is used to determine the importance of each abnormal event subsequence. In the embodiment of the present disclosure, when using the feature fusion sub-model to perform feature fusion on the extracted features corresponding to the sequence identifier, the feature fusion sub-model can be used to obtain the attention weight corresponding to the sequence identifier, that is, to determine the importance of each abnormal event subsequence. Then, a weighted fusion is performed on the subsequence extracted features corresponding to the sequence identifier and the attention weight corresponding to the sequence identifier to obtain a fusion result. Formula (1) is used to calculate and obtain the fusion result Rep.

[0006] Rep In Formula (1), w = Softmax(l / / £,), where Softmax represents the activation function. Among them, e严 represents the feature extraction of the subsequence. In an optional embodiment, in step S2331, a feature fusion sub-model is used to fuse the extracted features corresponding to the sequence identifier to obtain a fusion result, including the following method steps: Eigenvector, that is, obtain the eigenvector corresponding to the [cis] identifier, and determine the fusion result Rep based on the eigenvector corresponding to the obtained sequence identifier. Exemplarily, when using the top-level Transformer model (Top Transformer) to fuse the information between abnormal event subsequences, that is, using the Transformer encoder structure at the top level to fuse the features of multiple subsequences and learn the correlation between abnormal event subsequences. The feature corresponding to the [cis] identifier can be concatenated in front of the features of m abnormal event subsequences as the input, that is,狎 = (e"s,劳”,..., £新), and then the top-level Transformer encoder encodes 狎 to obtain the fusion result* = Transformer(% ‘). Among them, X l+1 e R( m+1 ) xD, that is, the feature vector corresponding to the [cis] identifier is used as the final output fusion result Rep. In an optional embodiment, in step S234, a classifier is used to classify the fusion result to obtain a target fault prediction result, including the following method steps: Step S2341: Classify the fusion result using a classifier to obtain a confidence score corresponding to the fusion result, wherein, in response to the confidence score being greater than a preset threshold, the target fault prediction result indicates that the physical server will experience a downtime within a future time period, and in response to the confidence score being less than or equal to the preset threshold, the target fault prediction result indicates that the physical server will operate normally within a future time period. In the disclosed embodiment, when a classifier is used to classify the fusion result, the classifier can be used to classify the fusion result to obtain a confidence score corresponding to the fusion result. This can be understood as obtaining the confidence score that the model predicts that the fusion result is a positive sample, that is, the confidence level or probability value that the fusion result is predicted to be a sample that triggers a physical server downtime. It is understood that if the confidence level is greater than a preset threshold, the target fault prediction result indicates that the physical server will experience downtime within a future time period. If the confidence level is less than or equal to the preset threshold, the target fault prediction result indicates that the physical server will operate normally within a future time period. Exemplarily, the feature fusion sub-model output Rep is also input into a classifier for classification to predict whether the NC will experience a fault and downtime. It is understood that if the loss function used during training is the cross entropy loss, the gradient of the loss is backpropagated using a gradient descent algorithm to update the model parameters. Figure 3 is a flowchart of long sequence processing according to Example 1 of the present disclosure. As shown in Figure 3, the extremely long abnormal event sequence obtained by abstracting the extremely long abnormal log is first subjected to sequence splitting (Sequence Splitting). The multiple abnormal event subsequences obtained by the splitting are then encoded into vector form using the feature encoding sub-model. The [cis] identifier is added before each abnormal event subsequence, thereby obtaining multiple encoded results. Then, the feature extraction sub-model can be used to extract features from the encoding results to obtain an extraction result. It will be understood that the extraction result includes extracted features such as EmbCl and EmbCn corresponding to the sequence identifier, as well as extracted features such as Emb1 and Embmm corresponding to each abnormal event subsequence. Next, the feature fusion sub-model can be used to fuse the extracted results to obtain a fusion result. For example, if attention weighting is used for feature fusion, the importance of different abnormal event subsequences can be learned by setting a fully connected layer to obtain the attention weight w.The features of the m abnormal event subsequences are then weighted and fused to obtain a fusion result. If a Transformer encoder structure is used at the top level for feature fusion, the features corresponding to the [cis] identifier are appended to the front of the m abnormal event subsequence features as input. The input is then encoded by the top-level Transformer encoder to obtain the fusion result. As can be seen, the present disclosure transforms the fault prediction problem of abnormal logs into a natural language processing problem, abstracts extremely long abnormal logs into sequence form, and segments the long sequence into multiple shorter abnormal event subsequences. These multiple abnormal event subsequences are then used as input for the Transformer model. This reduces the model's computational cost and computation time, improving its computational efficiency and prediction accuracy, enabling timely operation and maintenance, and ultimately enhancing the stability of the cloud computing system. Furthermore, the present disclosure designs multiple subsequence feature fusion methods to learn the correlations between abnormal event subsequences, thereby capturing the global key information of the complete long sequence and more thoroughly mining the abnormal patterns in the long sequence. Ultimately, it is possible to train an accurate and efficient fault prediction model. It is easy to understand that the beneficial effects of the log data processing method provided by the present disclosure include the following points. Beneficial effect (1): The present disclosure converts the abnormal logs produced by the cloud computing platform into natural language processing problems, extracts semantic information from long log sequences, and thus can accurately predict whether the physical server will fail. Beneficial effect (2): Compared with the original BERT model, the fault prediction model of the present disclosure can process longer input sequences, reduce computational complexity, and greatly improve the accuracy and recall rate of fault prediction, greatly improving the stability of the cloud computing system. Beneficial effect (3): Compared with the LogBERT model, the present disclosure constructs a separate abnormal event library for encoding, eliminating the complex process of the BERT model directly encoding the log text. At the same time, there is no pre-training process, the model size is smaller, the training and testing speed is faster, and the efficiency is higher. In addition, the present disclosure improves the method of directly processing the input in the BERT model, so that the model can process longer input sequences and obtain a larger amount of information, which is more conducive to the full learning of the model. Beneficial effect (4): Compared with the XgBoost model, the present invention uses the Transformer encoder as the basic feature extractor. The Transformer model has proven its ability to extract complex feature expressions in many fields such as natural language processing and images, while XgBoost uses manually extracted statistical features, which can only reflect the frequency of each abnormal log occurring within a fixed time, but cannot reflect information such as the order in which the abnormal logs occur.Furthermore, the XgBoost model requires extensive manual feature calculation for long sequence inputs. The present disclosure can directly use long sequences of abnormal events as input, simultaneously extracting and classifying features. In summary, the long-sequence Transformer fault prediction model of the present disclosure can fully leverage the semantic and temporal information in ultra-long abnormal sequences to extract more representative feature expressions, thereby improving the performance of the trained classifier and achieving higher precision and recall in fault prediction. This, in turn, enables cloud computing systems to more quickly and accurately identify impending NCs for maintenance, thereby reducing downtime rates and minimizing the losses caused to users by downtime, effectively improving the stability and reliability of cloud computing systems. It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, storage, and display, etc.) involved in this disclosure are all authorized by the user or fully authorized by all parties. The collection, use, and processing of such data must comply with the relevant laws, regulations, and standards of relevant countries and regions, and corresponding entry points are provided for users to authorize or deny access. Furthermore, it should be noted that, for simplicity of description, the aforementioned method embodiments are described as a series of combined actions. However, those skilled in the art should be aware that the present disclosure is not limited by the order of the actions described, as certain steps may be performed in a different order or simultaneously, according to the present disclosure. Furthermore, those skilled in the art should also be aware that the embodiments described in this specification are preferred embodiments, and the actions and modules involved are not necessarily required for the present disclosure. Through the above description of the embodiments, those skilled in the art will clearly understand that the methods according to the aforementioned embodiments can be implemented using software and a required general-purpose hardware platform, or alternatively, hardware. Based on this understanding, the technical solution of the present disclosure, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (e.g., ROM / RAM, a magnetic disk, or an optical disk) and includes instructions for enabling a terminal device (which may be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in the various embodiments of the present disclosure.Example 2: In the operating environment of Example 1, the present disclosure provides a log data processing method as shown in FIG4 . FIG4 is a flowchart of a log data processing method according to Example 2 of the present disclosure. As shown in FIG4 , the method includes: Step S41: Obtaining a fault prediction request through a first application programming interface, wherein the request data carried in the fault prediction request includes log data obtained from a physical server, wherein the log data includes at least one type of abnormal log collected within a historical time period; Step S42: Returning a fault prediction response through a second application programming interface, wherein the response data carried in the fault prediction response includes a target fault prediction result, which is obtained by performing fault prediction on a segmentation result using a target fault prediction model, wherein the segmentation result is obtained by performing data segmentation on the log data, and the target fault prediction model is used to predict whether the physical server will crash due to a fault within a future time period. The fault prediction request can be understood as a fault prediction request for predicting whether an NC in a cloud computing system will crash due to a fault. The request data carried in the fault prediction request includes log data obtained from the physical server. A physical server is a physical machine (Network Computer, NC) running multiple virtual machines (VMs) and providing computing, storage, and network resources. Log data can be understood as system exception logs reported in real time when an NC exception occurs, indicating the cause of the NC exception. It is understood that log data can include various types of exception logs. For example, log data can include exception logs caused by memory errors, disk errors, and central processing unit (CPU) errors, without limitation. In the disclosed embodiments, log data can be at least one type of exception log collected within a historical time period. For example, exception logs can be collected at 5-minute intervals and within a 3-day sampling window, thereby obtaining log data. This is not a limitation. The fault prediction response can be understood as the response to the fault prediction request. The response data carried in the fault prediction response includes the target fault prediction result. The target fault prediction result is obtained by using the target fault prediction model to predict the segmentation result. The segmentation result is obtained by segmenting the log data. The target fault prediction model is used to predict whether the physical server will crash due to a fault in the future time period.Considering that NCs generate a large number of exception logs due to various fault causes, it is common for a large number of exception logs to be reported in a short period of time, resulting in extremely long exception logs. Because processing extremely long exception logs increases computational complexity, in embodiments of the present disclosure, data segmentation is performed on the log data to break the extremely long exception logs into shorter ones, thereby obtaining segmentation results. For example, the log data can be abstracted into sequences, specifically extremely long exception logs, which can then be abstracted into extremely long exception sequences. Sequence splitting is then performed on the long sequence to divide the long sequence into multiple subsequences, thereby obtaining segmentation results. It is understood that the lengths of the multiple subsequences obtained by segmentation are much shorter than the length of the long sequence. Therefore, by segmenting the long sequence into multiple subsequences and subsequently applying a model to fault prediction on the resulting subsequences, the computational complexity of the model can be greatly reduced, thereby improving the computational efficiency of the model. The target fault prediction model can be understood as a model used to predict whether an NC will crash due to a fault within a future time period. It can be a Transformer model capable of processing long sequences, a large model, or other machine learning model, without limitation. In the disclosed embodiment, the segmentation results are input into the target fault prediction model, which then performs fault prediction on the segmentation results to obtain a target fault prediction result indicating whether the corresponding NC will crash. This effectively reduces the complexity of the model calculation and yields a more accurate prediction result. Specifically, the target fault prediction model in the disclosed embodiment can accurately predict whether an NC will crash based on the extremely long system exception logs reported by the NC, enabling timely maintenance and improving the stability of the cloud computing system. In the disclosed embodiment, a fault prediction request is obtained via a first application programming interface, wherein the request data carried in the fault prediction request includes log data obtained from the physical server, which at least includes at least one type of exception log collected within a historical time period. Then, a fault prediction response is returned through the second application programming interface, wherein the response data carried in the fault prediction response includes: a target fault prediction result, the target fault prediction result is obtained by performing fault prediction on the segmentation result using a target fault prediction model, the segmentation result is obtained by performing data segmentation on the log data, and the target fault prediction model is used to predict whether the physical server will crash due to a fault in a future time period.As can be seen, the present disclosure transforms the fault prediction problem of exception logs into a natural language processing problem, abstracts extremely long exception logs into sequence form, and segments the long sequence into multiple shorter subsequences. These subsequences are then used as input to the Transformer model. This reduces the model's computational cost and computation time, improving the model's computational efficiency and prediction accuracy, enabling timely operation and maintenance, and thereby enhancing the stability of the cloud computing system. The log data processing methods provided in the embodiments of the present disclosure can be applied, but are not limited to, to application scenarios involving physical server fault prediction in fields such as e-commerce services, education services, legal services, medical services, conference services, social networking services, financial product services, logistics services, and navigation services. For example, scenarios involving prediction of NC faults for e-commerce services, education services, legal services, and medical services, without limitation. In an embodiment of the present disclosure, a fault prediction request is obtained through a first application programming interface (API). The request data carried in the fault prediction request includes log data obtained from a physical server, the log data including at least one type of exception log collected within a historical time period. A fault prediction response is then returned through a second application programming interface (API). The response data carried in the fault prediction response includes a target fault prediction result. The target fault prediction result is obtained by performing fault prediction on a segmentation result using a target fault prediction model. The segmentation result is obtained by segmenting the log data. The target fault prediction model is used to predict whether the physical server will crash due to a fault within a future time period. This achieves the goal of accurately predicting faults based on extremely long exception logs reported by the physical server, thereby achieving the technical effects of reducing the model's computational complexity and computation time, improving the model's computational efficiency, improving the model's prediction accuracy, and enhancing the stability of the cloud computing system. This solves the technical problem in related technologies of predicting physical machine faults based on the BERT model, which results in high computational complexity, long computation time, and low prediction accuracy. It should be noted that the preferred implementation of this embodiment can refer to the relevant description in Example 1 and will not be repeated here. Example 3 In the operating environment as in Example 1, the present disclosure provides a log data processing method as shown in FIG5.FIG5 is a flowchart of a log data processing method according to Embodiment 3 of the present disclosure. As shown in FIG5 , the method includes: Step S51: Retrieving a currently input fault prediction dialog request, wherein the request data carried in the fault prediction dialog request includes log data obtained from a physical server, wherein the log data includes at least one type of exception log collected within a historical time period; Step S52: Returning a fault prediction dialog reply in response to the fault prediction dialog request, wherein the information carried in the fault prediction dialog reply includes a target fault prediction result, which is obtained by performing fault prediction on a segmentation result using a target fault prediction model. The segmentation result is obtained by segmenting the log data. The target fault prediction model is used to predict whether a physical server will crash due to a fault within a future time period; Step S53: Displaying the target fault prediction result in a graphical user interface. The fault prediction dialog request can be understood as a fault prediction dialog request for predicting whether an NC in a cloud computing system will crash due to a fault, and can be understood as a request input by a user when communicating with a smart device. The request data carried in the fault prediction dialog request includes log data obtained from a physical server. A physical server is a physical machine (Network Computer, NC) running multiple virtual machines (VMs) that provides computing, storage, and network resources. The log data can be understood as system exception logs reported in real time when an NC exception occurs, indicating the cause of the NC exception. It is understood that the log data can include various types of exception logs. For example, the log data may include exception logs caused by memory errors, disk errors, and central processing unit (CPU) errors, without limitation. In the disclosed embodiments, the log data may be at least one type of exception log collected within a historical time period. For example, the exception logs may be collected at 5-minute intervals and within a 3-day sampling window, without limitation. The fault prediction dialogue reply may be a reply corresponding to the fault prediction dialogue request. The fault prediction dialogue reply carries information including a target fault prediction result. The target fault prediction result is obtained by performing fault prediction on the segmentation result using a target fault prediction model. The segmentation result is obtained by performing data segmentation on the log data. The target fault prediction model is used to predict whether the physical server will crash due to a fault in a future time period.Considering that NCs generate a large number of exception logs due to various fault causes, it is common for a large number of exception logs to be reported in a short period of time, resulting in extremely long exception logs. Because processing extremely long exception logs increases computational complexity, in embodiments of the present disclosure, data segmentation is performed on the log data to break the extremely long exception logs into shorter ones, thereby obtaining segmentation results. For example, the log data can be abstracted into sequences, specifically extremely long exception logs, which can then be abstracted into extremely long exception sequences. Sequence splitting is then performed on the long sequence to divide the long sequence into multiple subsequences, thereby obtaining segmentation results. It is understood that the lengths of the multiple subsequences obtained by segmentation are much shorter than the length of the long sequence. Therefore, by segmenting the long sequence into multiple subsequences and subsequently applying a model to fault prediction on the resulting subsequences, the computational complexity of the model can be greatly reduced, thereby improving the computational efficiency of the model. The target fault prediction model can be understood as a model used to predict whether an NC will crash due to a fault within a future time period. It can be a Transformer model capable of processing long sequences, a large model, or other machine learning model, without limitation. In the disclosed embodiment, the segmentation results are input into the target fault prediction model, which then performs fault prediction on the segmentation results to obtain a target fault prediction result indicating whether the corresponding NC will crash. This effectively reduces the complexity of the model calculation and yields a more accurate prediction result. Specifically, the target fault prediction model in the disclosed embodiment can accurately predict whether an NC will crash based on the extremely long system exception logs reported by the NC, enabling timely maintenance and improving the stability of the cloud computing system. After obtaining the target fault prediction result, it can also be displayed in a graphical user interface to provide user feedback. In the disclosed embodiment, the currently input fault prediction dialog request is obtained, where the request data carried in the fault prediction dialog request includes log data obtained from the physical server, which at least includes at least one type of exception log collected within a historical time period. Then, in response to the fault prediction dialog request, a fault prediction dialog reply is returned. The fault prediction dialog reply includes a target fault prediction result. The target fault prediction result is obtained by performing fault prediction on the segmentation result using a target fault prediction model. The segmentation result is obtained by segmenting the log data. The target fault prediction model is used to predict whether the physical server will crash due to a fault in the future. After obtaining the target fault prediction result, it is displayed in a graphical user interface to provide feedback to the user.As can be seen, the present disclosure transforms the fault prediction problem of exception logs into a natural language processing problem, abstracts extremely long exception logs into sequence form, and segments the long sequence into multiple shorter subsequences. These subsequences are then used as input to the Transformer model. This reduces the model's computational cost and computation time, improving the model's computational efficiency and prediction accuracy, enabling timely operation and maintenance, and thereby enhancing the stability of the cloud computing system. The log data processing methods provided in the embodiments of the present disclosure can be applied, but are not limited to, to application scenarios involving physical server fault prediction in fields such as e-commerce services, education services, legal services, medical services, conference services, social networking services, financial product services, logistics services, and navigation services. For example, scenarios involving prediction of NC faults for e-commerce services, education services, legal services, and medical services, without limitation here. According to an embodiment of the present disclosure, a currently input fault prediction dialog request is obtained, wherein the request data carried in the fault prediction dialog request includes log data obtained from a physical server, wherein the log data includes at least one type of exception log collected within a historical time period. Then, in response to the fault prediction dialog request, a fault prediction dialog reply is returned, wherein the fault prediction dialog reply includes a target fault prediction result. The target fault prediction result is obtained by performing fault prediction on a segmentation result using a target fault prediction model. The segmentation result is obtained by performing data segmentation on the log data. The target fault prediction model is used to predict whether the physical server will crash due to a fault within a future time period. After obtaining the target fault prediction result, it is displayed in a graphical user interface to provide user feedback. This achieves the goal of accurately predicting faults based on the extremely long exception logs reported by the physical server, thereby achieving the technical effects of reducing the model's computational complexity, reducing the model's calculation time, improving the model's computational efficiency, improving the model's prediction accuracy, and enhancing the stability of the cloud computing system. This further solves the technical problem in the related art of predicting physical machine faults based on the BERT model, which results in high computational complexity, long calculation time, and low prediction accuracy. It should be noted that the preferred implementation of this embodiment can be found in the relevant description of Example 1 and will not be repeated here. Example 4 According to the embodiments of the present disclosure, an embodiment of an apparatus for implementing the above-mentioned log data processing method is also provided.FIG6 is a schematic structural diagram of a log data processing device according to Embodiment 4 of the present disclosure. As shown in FIG6 , the device includes: a first acquisition module 601 configured to acquire log data from a physical server, wherein the log data includes at least one type of abnormal log collected within a historical time period; a segmentation module 602 configured to segment the log data to obtain segmentation results; and a prediction module 603 configured to perform fault prediction on the segmentation results using a target fault prediction model to obtain a target fault prediction result, wherein the target fault prediction model is used to predict whether the physical server will crash due to a fault within a future time period. Optionally, the segmentation module 602 is further configured to: acquire an abnormal event sequence corresponding to the log data, wherein the abnormal event sequence is obtained by arranging the category identifiers corresponding to the at least one type of abnormal log in a predetermined order; and segment the abnormal event sequence to obtain a segmentation result, wherein the segmentation result includes: multiple abnormal event subsequences obtained by segmenting the abnormal event sequence. Optionally, the segmentation module 602 is further configured to: perform data abstraction on multiple exception logs contained in the log data to obtain multiple exception events; based on preset exception event types, convert the multiple exception events into type identifiers corresponding to at least one type of exception log to obtain a conversion result, wherein the preset exception event type is used to determine multiple log formats and the encodings corresponding to different log formats; and generate an exception event sequence based on the conversion result. Optionally, the segmentation module 602 is further configured to: determine positive samples and negative samples based on the conversion result, wherein positive samples are samples that trigger a physical server crash, and negative samples are samples that do not trigger a physical server crash; arrange the positive samples in a preset order to obtain an arrangement result; and generate an exception event sequence using the arrangement result and the negative samples. Optionally, the segmentation module 602 is further configured to: perform sliding segmentation on the exception event sequence using the length of the exception event subsequence as a segmentation window to obtain a segmentation result. Optionally, the target fault prediction model includes: a feature encoding sub-model, a feature extraction sub-model, a feature fusion sub-model, and a classifier. The prediction module 603 is further configured to: use the feature encoding sub-model to perform embedding encoding on the segmentation result to obtain an encoding result; use the feature extraction sub-model to perform feature extraction on the encoding result to obtain an extraction result; use the feature fusion sub-model to perform feature fusion on the extraction result to obtain a fusion result; and use the classifier to classify the fusion result to obtain a target fault prediction result.Optionally, the prediction module 603 is further configured to: add a sequence identifier to each abnormal event subsequence in the segmentation result; and embed the segmentation result and the sequence identifier using a feature encoding sub-model to obtain an encoding result. Optionally, the prediction module 603 is further configured to: use a feature fusion sub-model to perform feature fusion on the extracted features corresponding to the sequence identifier contained in the extraction result to obtain a fusion result. Optionally, the prediction module 603 is further configured to: use the feature fusion sub-model to obtain an attention weight corresponding to the sequence identifier, and perform a weighted fusion of the subsequence extracted features corresponding to the sequence identifier with the attention weight corresponding to the sequence identifier to obtain a fusion result, wherein the attention weight is used to determine the importance of each abnormal event subsequence. Optionally, the prediction module 603 is further configured to: use a feature fusion sub-model to perform feature fusion on the splicing result to obtain a feature vector corresponding to the sequence identifier, and determine a fusion result based on the feature vector corresponding to the sequence identifier, wherein the splicing result is obtained by splicing the identifier extracted features corresponding to the sequence identifier with the subsequence extracted features corresponding to the sequence identifier. Optionally, the prediction module 603 is further configured to classify the fusion result using a classifier to obtain a confidence level corresponding to the fusion result, wherein, in response to the confidence level being greater than a preset threshold, the target fault prediction result indicates that the physical server will crash within a future time period, and in response to the confidence level being less than or equal to the preset threshold, the target fault prediction result indicates that the physical server will operate normally within a future time period. According to the embodiments of the present disclosure, log data, including at least one type of abnormal log collected within a historical time period, is acquired from a physical server. The log data is then segmented to divide long sequences in the log data into multiple shorter subsequences. Finally, a target fault prediction model is used to perform fault prediction on the multiple shorter subsequences, thereby obtaining a target fault prediction result indicating whether the corresponding physical server is about to crash. This achieves the goal of accurately predicting faults based on extremely long abnormal logs reported by the physical server, thereby reducing the computational complexity and computation time of the model, improving the computational efficiency of the model, improving the prediction accuracy of the model, and enhancing the stability of the cloud computing system. This further addresses the technical issues in related technologies related to predicting physical machine faults based on the BERT model, which result in high computational complexity, long computation time, and low prediction accuracy. It should be noted that the first acquisition module 601, segmentation module 602, and prediction module 603 correspond to steps S21 to S23 in Example 1. The examples and application scenarios implemented by these three modules and the corresponding steps are the same, but are not limited to those disclosed in Example 1.It should be noted that the above modules or units may be hardware components or software components stored in a memory and processed by one or more processors. The above modules may also run in the server 10 provided in Example 1. According to an embodiment of the present disclosure, another embodiment of a device for implementing the above-mentioned log data processing method is also provided. FIG7 is a schematic structural diagram of another log data processing device according to Example 4 of the present disclosure. As shown in FIG7 , the device includes: a second acquisition module 701, configured to acquire a fault prediction request via a first application programming interface, wherein the request data carried in the fault prediction request includes: log data acquired from a physical server, wherein the log data includes at least: at least one type of abnormal log collected within a historical time period; a first return module 702, configured to return a fault prediction response via the second application programming interface, wherein the response data carried in the fault prediction response includes: a target fault prediction result, which is obtained by performing fault prediction on a segmentation result using a target fault prediction model. The segmentation result is obtained by performing data segmentation on the log data. The target fault prediction model is used to predict whether the physical server will crash due to a fault within a future time period. In an embodiment of the present disclosure, a fault prediction request is obtained through a first application programming interface (API). The request data carried in the fault prediction request includes log data obtained from a physical server, the log data including at least one type of exception log collected within a historical time period. A fault prediction response is then returned through a second application programming interface (API). The response data carried in the fault prediction response includes a target fault prediction result. The target fault prediction result is obtained by performing fault prediction on a segmentation result using a target fault prediction model. The segmentation result is obtained by segmenting the log data. The target fault prediction model is used to predict whether the physical server will crash due to a fault within a future time period. This achieves the goal of accurately predicting faults based on extremely long exception logs reported by the physical server, thereby achieving the technical effects of reducing the model's computational complexity and computation time, improving the model's computational efficiency, improving the model's prediction accuracy, and enhancing the stability of the cloud computing system. This solves the technical problem in related technologies of predicting physical machine faults based on the BERT model, which results in high computational complexity, long computation time, and low prediction accuracy. It should be noted that the second acquisition module 701 and the first return module 702 correspond to step S41 and step S42 in Example 2. The examples and application scenarios implemented by the two modules and the corresponding steps are the same, but are not limited to the contents disclosed in Example 1.It should be noted that the above modules or units may be hardware components or software components stored in a memory and processed by one or more processors. The above modules may also be run in the server 10 provided in Example 1. According to an embodiment of the present disclosure, another device embodiment for implementing the above log data processing method is also provided. FIG8 is a schematic structural diagram of another log data processing device according to Embodiment 4 of the present disclosure. As shown in FIG8 , the device includes: a third acquisition module 801, configured to acquire a currently input fault prediction dialogue request, wherein the request data carried in the fault prediction dialogue request includes log data acquired from a physical server, wherein the log data includes at least at least one type of abnormality log collected within a historical time period; a second return module 802, configured to return a fault prediction dialogue reply in response to the fault prediction dialogue request, wherein the information carried in the fault prediction dialogue reply includes a target fault prediction result, wherein the target fault prediction result is obtained by performing fault prediction on a segmentation result using a target fault prediction model, wherein the segmentation result is obtained by performing data segmentation on the log data, and the target fault prediction model is used to predict whether the physical server will crash due to a fault within a future time period; and a display module 803, configured to display the target fault prediction result in a graphical user interface. According to an embodiment of the present disclosure, a currently input fault prediction dialog request is obtained, wherein the request data carried in the fault prediction dialog request includes log data obtained from a physical server, wherein the log data includes at least one type of exception log collected within a historical time period. Then, in response to the fault prediction dialog request, a fault prediction dialog reply is returned, wherein the fault prediction dialog reply includes a target fault prediction result. The target fault prediction result is obtained by performing fault prediction on a segmentation result using a target fault prediction model. The segmentation result is obtained by performing data segmentation on the log data. The target fault prediction model is used to predict whether the physical server will crash due to a fault within a future time period. After obtaining the target fault prediction results, they are displayed in a graphical user interface to provide user feedback. This achieves the goal of accurately predicting faults based on the extremely long exception logs reported by the physical server, thereby achieving the technical effects of reducing the model's computational complexity and calculation time, improving the model's computational efficiency, and improving the model's prediction accuracy, as well as enhancing the stability of the cloud computing system. This solves the technical problem in related technologies of predicting physical machine faults based on the BERT model, which results in high computational complexity, long calculation time, and low prediction accuracy.It should be noted that the third acquisition module 801, second return module 802, and display module 803 described above correspond to steps S51 to S53 in Example 3. The examples and application scenarios implemented by these three modules and the corresponding steps are the same, but are not limited to the content disclosed in Example 1. It should be noted that the above modules or units can be hardware components or software components stored in a memory and processed by one or more processors. The above modules can also run on the server 10 provided in Example 1. It should be noted that the preferred implementation schemes involved in the above embodiments of the present disclosure are the same as the schemes, application scenarios, and implementation processes provided in Example 1, but are not limited to the schemes provided in Example 1. Example 5 The embodiments of the present disclosure may provide a computer terminal, which can be any computer terminal device in a computer terminal group. Optionally, in this embodiment, the computer terminal can be replaced by a terminal device such as a mobile terminal. Optionally, in this embodiment, the computer terminal can be located in at least one of multiple network devices in a computer network. In this embodiment, the computer terminal can execute program code for the following steps in the log data processing method: obtaining log data from a physical server, where the log data includes at least one type of abnormal log collected within a historical time period; performing data segmentation on the log data to obtain a segmentation result; and performing fault prediction on the segmentation result using a target fault prediction model to obtain a target fault prediction result, where the target fault prediction model is used to predict whether the physical server will crash due to a fault within a future time period. Optionally, FIG9 is a block diagram of a computer terminal according to an embodiment of the present disclosure. As shown in FIG9 , the computer terminal 9 may include one or more (only one is shown) processors 902, a memory 904, a storage controller, and a peripheral interface, where the peripheral interface is connected to a radio frequency module, an audio module, and a display. The memory can be used to store software programs and modules, such as program instructions / modules corresponding to the log data processing method and apparatus in the embodiments of the present disclosure. The processor executes the stored software programs and modules to execute various functional applications and data processing, thereby implementing the aforementioned log data processing method. The memory may include high-speed random access memory and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory may further include memory remotely located relative to the processor, and such remote memory may be connected to the computer terminal 9 via a network.Examples of the aforementioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof. The processor can access information and applications stored in the memory via a transmission device to execute the following steps: obtaining log data from the physical server, wherein the log data includes at least: at least one type of abnormal log collected within a historical time period; performing data segmentation on the log data to obtain a segmentation result; and performing fault prediction on the segmentation result using a target fault prediction model to obtain a target fault prediction result, wherein the target fault prediction model is used to predict whether the physical server will crash due to a fault within a future time period. Optionally, the processor can further execute program code for the following steps: obtaining an abnormal event sequence corresponding to the log data, wherein the abnormal event sequence is obtained by arranging the category identifiers corresponding to the at least one type of abnormal log in a predetermined order; and segmenting the abnormal event sequence to obtain a segmentation result, wherein the segmentation result includes: a plurality of abnormal event subsequences obtained by segmenting the abnormal event sequence. Optionally, the processor may further execute program code for the following steps: performing data abstraction on multiple exception logs contained in the log data to obtain multiple exception events; converting the multiple exception events into category identifiers corresponding to at least one type of exception log based on preset exception event categories to obtain conversion results, wherein the preset exception event categories are used to determine multiple log formats and the encodings corresponding to the different log formats; and generating an exception event sequence based on the conversion results. Optionally, the processor may further execute program code for the following steps: determining positive samples and negative samples based on the conversion results, wherein positive samples are samples that trigger a physical server crash, and negative samples are samples that do not trigger a physical server crash; arranging the positive samples in a preset order to obtain an arrangement result; and generating an exception event sequence using the arrangement result and the negative samples. Optionally, the processor may further execute program code for the following steps: performing sliding segmentation on the exception event sequence using the length of the exception event subsequence as a segmentation window to obtain a segmentation result. Optionally, the processor may further execute program code for the following steps: embedding and encoding the segmentation results using a feature encoding sub-model to obtain an encoded result; extracting features from the encoded results using a feature extraction sub-model to obtain an extracted result; fusing features from the extracted results using a feature fusion sub-model to obtain a fused result; and classifying the fused result using a classifier to obtain a target fault prediction result. Optionally, the processor may further execute program code for the following steps: adding a sequence identifier to each abnormal event subsequence in the segmentation result; and embedding and encoding the segmentation results and the sequence identifier using the feature encoding sub-model to obtain an encoded result.Optionally, the processor may further execute program code for the following steps: using a feature fusion sub-model to perform feature fusion on the extracted features corresponding to the sequence identifier contained in the extraction result to obtain a fusion result. Optionally, the processor may further execute program code for the following steps: using the feature fusion sub-model to obtain an attention weight corresponding to the sequence identifier, and performing weighted fusion on the subsequence extracted features corresponding to the sequence identifier with the attention weight corresponding to the sequence identifier to obtain a fusion result, wherein the attention weight is used to determine the importance of each abnormal event subsequence. Optionally, the processor may further execute program code for the following steps: using the feature fusion sub-model to perform feature fusion on the splicing result to obtain a feature vector corresponding to the sequence identifier, and determining a fusion result based on the feature vector corresponding to the sequence identifier, wherein the splicing result is obtained by splicing the identifier extracted features corresponding to the sequence identifier with the subsequence extracted features corresponding to the sequence identifier. Optionally, the processor may further execute program code of the following steps: using a classifier to classify the fusion result to obtain a confidence level corresponding to the fusion result, wherein, in response to the confidence level being greater than a preset threshold, the target fault prediction result indicates that the physical server will crash within a future time period; and in response to the confidence level being less than or equal to the preset threshold, the target fault prediction result indicates that the physical server will operate normally within a future time period. According to the embodiments of the present disclosure, log data, including at least one type of exception log collected within a historical time period, is acquired from a physical server. The log data is then segmented to divide long sequences in the log data into multiple shorter subsequences. Finally, a target fault prediction model is used to perform fault prediction on the multiple shorter subsequences, resulting in a target fault prediction result indicating whether the corresponding physical server is about to crash. This achieves the goal of accurately predicting faults based on extremely long exception logs reported by the physical server, thereby reducing the computational complexity and computation time of the model, improving the computational efficiency of the model, improving the prediction accuracy of the model, and enhancing the stability of the cloud computing system. This solves the technical problem in related art of predicting physical machine faults based on the BERT model, which results in high computational complexity, long computation time, and low prediction accuracy. Those skilled in the art will appreciate that the structure shown in FIG9 is merely illustrative, and that the computer terminal 9 may also be a smartphone (e.g., an Android phone, an iOS phone, etc.), a tablet computer, a PDA, or a mobile internet device (MID), a PAD, or other terminal device. FIG9 does not limit the structure of the electronic device.For example, the computer terminal 9 may include more or fewer components (such as a network interface, a display device, etc.) than those shown in FIG9 , or may have a configuration different from that shown in FIG9 . Those skilled in the art will appreciate that all or part of the steps in the various methods of the above embodiments can be completed by instructing the hardware associated with the terminal device through a program. The program can be stored in a computer-readable storage medium, which may include a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk. Example 6 The embodiments of the present disclosure also provide a computer-readable storage medium. Optionally, in this embodiment, the computer-readable storage medium can be used to store the program code executed by the log data processing method provided in Example 1. Optionally, in this embodiment, the computer-readable storage medium can be located in any computer terminal in a computer terminal group in a computer network, or in any mobile terminal in a mobile terminal group. Optionally, in this embodiment, the computer-readable storage medium is configured to store program code for executing the following steps: obtaining log data from a physical server, wherein the log data includes at least one type of abnormal log collected within a historical time period; performing data segmentation on the log data to obtain a segmentation result; and performing fault prediction on the segmentation result using a target fault prediction model to obtain a target fault prediction result, wherein the target fault prediction model is used to predict whether the physical server will crash due to a fault within a future time period. Optionally, in this embodiment, the computer-readable storage medium is configured to store program code for executing the following steps: obtaining an abnormal event sequence corresponding to the log data, wherein the abnormal event sequence is obtained by arranging category identifiers corresponding to the at least one type of abnormal log in a predetermined order; and segmenting the abnormal event sequence to obtain a segmentation result, wherein the segmentation result includes a plurality of abnormal event subsequences obtained by segmenting the abnormal event sequence. Optionally, in this embodiment, the computer-readable storage medium is configured to store program code for executing the following steps: performing data abstraction on multiple exception logs contained in log data to obtain multiple exception events; converting the multiple exception events into type identifiers corresponding to at least one type of exception log based on preset exception event types to obtain conversion results, wherein the preset exception event types are used to determine multiple log formats and codes corresponding to different log formats; and generating an exception event sequence based on the conversion results.Optionally, in this embodiment, the computer-readable storage medium is configured to store program code for executing the following steps: determining positive samples and negative samples based on the transformation results, wherein positive samples are samples that trigger a physical server downtime, and negative samples are samples that do not trigger a physical server downtime; arranging the positive samples in a preset order to obtain an arrangement result; and generating an abnormal event sequence using the arrangement result and the negative samples. Optionally, in this embodiment, the computer-readable storage medium is configured to store program code for executing the following steps: performing sliding segmentation on the abnormal event sequence using the length of the abnormal event subsequence as a segmentation window to obtain a segmentation result. Optionally, in this embodiment, the computer-readable storage medium is configured to store program code for executing the following steps: embedding encoding the segmentation result using a feature encoding sub-model to obtain an encoding result; extracting features from the encoding result using a feature extraction sub-model to obtain an extraction result; fusing features on the extraction result using a feature fusion sub-model to obtain a fusion result; and classifying the fusion result using a classifier to obtain a target fault prediction result. Optionally, in this embodiment, the computer-readable storage medium is configured to store program code for executing the following steps: adding a sequence identifier to each abnormal event subsequence in the segmentation result; and embedding encoding the segmentation result and the sequence identifier using a feature encoding sub-model to obtain an encoding result. Optionally, in this embodiment, the computer-readable storage medium is configured to store program code for executing the following steps: using a feature fusion sub-model to fuse the extracted features corresponding to the sequence identifier contained in the extraction result to obtain a fusion result. Optionally, in this embodiment, the computer-readable storage medium is configured to store program code for executing the following steps: using a feature fusion sub-model to obtain an attention weight corresponding to the sequence identifier, and performing a weighted fusion of the subsequence extracted features corresponding to the sequence identifier and the attention weight corresponding to the sequence identifier to obtain a fusion result, wherein the attention weight is used to determine the importance of each abnormal event subsequence. Optionally, in this embodiment, the computer-readable storage medium is configured to store program code for executing the following steps: performing feature fusion on the splicing result using a feature fusion sub-model to obtain a feature vector corresponding to the sequence identifier, and determining a fusion result based on the feature vector corresponding to the sequence identifier, wherein the splicing result is obtained by splicing the identifier-extracted features corresponding to the sequence identifier with the subsequence-extracted features corresponding to the sequence identifier.Optionally, in this embodiment, a computer-readable storage medium is configured to store program code for executing the following steps: using a classifier to classify the fusion result and obtaining a confidence score corresponding to the fusion result, wherein, in response to the confidence score being greater than a preset threshold, the target fault prediction result indicates that the physical server will experience downtime within a future time period, and in response to the confidence score being less than or equal to the preset threshold, the target fault prediction result indicates that the physical server will operate normally within a future time period. Embodiments of the present disclosure also provide a computer program product, including a computer program. When executed by a processor, the computer program implements any of the aforementioned log data processing methods. This description is omitted here. The serial numbers of the aforementioned embodiments of the present disclosure are for descriptive purposes only and do not represent the merits of any particular embodiment. In the aforementioned embodiments of the present disclosure, the descriptions of each embodiment are mutually exclusive. For portions not detailed in a particular embodiment, reference should be made to the relevant descriptions of other embodiments. In the several embodiments provided in this disclosure, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is merely a logical functional division. In actual implementation, other divisions may be employed. For example, multiple units or components may be combined or integrated into another system, or some features may be omitted or not implemented. Furthermore, the couplings or direct couplings or communication connections shown or discussed may be through interfaces, or indirect couplings or communication connections between units or modules, and may be electrical or other. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of these units may be selected to achieve the objectives of the present embodiments as needed. Furthermore, the functional units in the various embodiments of the present disclosure may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. These integrated units may be implemented in either hardware or software functional units. If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium.Based on this understanding, the technical solution of this disclosure, or the portion that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product, stored in a storage medium, includes instructions for causing a computer device (such as a personal computer, server, or network device) to execute all or part of the steps of the methods described in various embodiments of this disclosure. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memories (ROMs), random access memories (RAMs), removable hard drives, magnetic disks, or optical disks. The above description is merely a preferred embodiment of this disclosure. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of this disclosure, and such improvements and modifications should also be considered within the scope of protection of this disclosure.

Claims

Claims 1. A log data processing method, comprising: Obtaining log data from a physical server, wherein the log data includes at least: at least one type of abnormal log collected within a historical time period; performing data segmentation on the log data to obtain a segmentation result; and performing fault prediction on the segmentation result using a target fault prediction model to obtain a target fault prediction result, wherein the target fault prediction model is used to predict whether the physical server will crash due to a fault within a future time period.

2. The log data processing method according to claim 1, wherein: Performing data segmentation on the log data to obtain the segmentation result includes: obtaining an abnormal event sequence corresponding to the log data, wherein the abnormal event sequence is obtained by arranging the type identifiers corresponding to the at least one type of abnormal log in a preset order; and segmenting the abnormal event sequence to obtain the segmentation result, wherein the segmentation result includes: multiple abnormal event subsequences obtained by segmenting the abnormal event sequence.

3. The log data processing method according to claim 2, wherein: Obtaining the abnormal event sequence corresponding to the log data includes: performing data abstraction on multiple abnormal logs contained in the log data to obtain multiple abnormal events; based on preset abnormal event types, converting the multiple abnormal events into type identifiers corresponding to the at least one type of abnormal log to obtain a conversion result, wherein the preset abnormal event type is used to determine multiple log formats and codes corresponding to different log formats; and generating the abnormal event sequence based on the conversion result.

4. The log data processing method according to claim 3, wherein: Generating the abnormal event sequence based on the conversion result includes: determining positive samples and negative samples based on the conversion result, wherein the positive samples are samples that trigger the physical server to crash, and the negative samples are samples that do not trigger the physical server to crash; arranging the positive samples according to the preset order to obtain an arrangement result; and generating the abnormal event sequence using the arrangement result and the negative samples. 27 5. The log data processing method according to claim 2, wherein: Segmenting the abnormal event sequence to obtain the segmentation result includes: performing sliding segmentation on the abnormal event sequence using the length of the abnormal event subsequence as a segmentation window to obtain the segmentation result.

6. The log data processing method according to claim 1, wherein: The target fault prediction model includes: a feature encoding sub-model, a feature extraction sub-model, a feature fusion sub-model, and a classifier. Using the target fault prediction model to perform fault prediction on the segmentation result to obtain the target fault prediction result includes: using the feature encoding sub-model to perform embedded encoding on the segmentation result to obtain an encoding result; using the feature extraction sub-model to perform feature extraction on the encoding result to obtain an extraction result; using the feature fusion sub-model to perform feature fusion on the extraction result to obtain a fusion result; and using the classifier to classify the fusion result to obtain the target fault prediction result.

7. The log data processing method according to claim 6, wherein: Using the feature coding sub-model to embed-code the segmentation result to obtain the encoding result includes: adding a sequence identifier to each abnormal event subsequence in the segmentation result; and using the feature coding sub-model to embed-code the segmentation result and the sequence identifier to obtain the encoding result.

8. The log data processing method according to claim 7, wherein: Using the feature fusion sub-model to perform feature fusion on the extraction result to obtain the fusion result includes: using the feature fusion sub-model to perform feature fusion on the extracted features corresponding to the sequence identifier included in the extraction result to obtain the fusion result.

9. The log data processing method according to claim 8, wherein: Using the feature fusion sub-model to perform feature fusion on the extracted features corresponding to the sequence identifier to obtain the fusion result includes: using the feature fusion sub-model to obtain an attention weight corresponding to the sequence identifier, and performing weighted fusion on the subsequence extracted features corresponding to the sequence identifier and the attention weight corresponding to the sequence identifier to obtain the fusion result, wherein the attention weight is used to determine the importance of each abnormal event subsequence.

10. The log data processing method according to claim 8, wherein: Using the feature fusion sub-model to perform feature fusion on the extracted features corresponding to the sequence identifier to obtain the fusion result includes: using the feature fusion sub-model to perform feature fusion on the splicing result to obtain the features corresponding to the sequence identifier; The fusion result is determined based on the feature vector corresponding to the sequence identifier, wherein the splicing result is obtained by splicing the identifier extraction feature corresponding to the sequence identifier and the subsequence extraction feature corresponding to the sequence identifier.

11. The log data processing method according to claim 6, wherein: Using the classifier to classify the fusion result to obtain the target fault prediction result includes: using the classifier to classify the fusion result to obtain a confidence level corresponding to the fusion result, wherein, in response to the confidence level being greater than a preset threshold, the target fault prediction result indicates that the physical server will crash within the future time period, and in response to the confidence level being less than or equal to the preset threshold, the target fault prediction result indicates that the physical server will operate normally within the future time period.

12. A log data processing method, comprising: A fault prediction request is obtained through a first application programming interface, wherein request data carried in the fault prediction request includes: log data obtained from a physical server, wherein the log data includes at least: at least one type of exception log collected within a historical time period; a fault prediction response is returned through a second application programming interface, wherein response data carried in the fault prediction response includes: a target fault prediction result, wherein the target fault prediction result is obtained by performing fault prediction on a segmentation result using a target fault prediction model, wherein the segmentation result is obtained by performing data segmentation on the log data, and the target fault prediction model is used to predict whether the physical server will crash due to a fault within a future time period.

13. A log data processing method, comprising: Obtaining a currently input fault prediction dialogue request, wherein request data carried in the fault prediction dialogue request includes: log data obtained from a physical server, wherein the log data includes at least: at least one type of exception log collected within a historical time period; returning a fault prediction dialogue reply in response to the fault prediction dialogue request, wherein information carried in the fault prediction dialogue reply includes: a target fault prediction result, wherein the target fault prediction result is obtained by performing fault prediction on a segmentation result using a target fault prediction model, wherein the segmentation result is obtained by performing data segmentation on the log data, and the target fault prediction model is used to predict whether the physical server will crash due to a fault within a future time period; and displaying the target fault prediction result in a graphical user interface.

14. An electronic device, comprising: a memory storing an executable program; A processor is configured to run the program, wherein the program executes the log data processing method according to any one of claims 1 to 13 when running.

15. A computer-readable storage medium comprising a stored executable program, wherein: When the executable program is running, the device where the computer-readable storage medium is located is controlled to execute the log data processing method according to any one of claims 1 to 13.

16. A computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements the log data processing method according to any one of claims 1 to 13.

Images