Fault prediction model training method, fault prediction method and device

Abstract

The invention discloses a fault prediction model training method, and a fault prediction method and device. Comprising the steps of obtaining a training data set, inputting first operation state index data into a long short-term memory network of a to-be-trained fault prediction model, determining time sequence characteristics of the first operation state index data in a first time period, inputting the time sequence characteristics into a Gaussian process regression network of the fault prediction model, and performing processing according to the time sequence characteristics to obtain a fault prediction result. And the probability distribution data of the operation state indexes of the sample communication network in the second time period are obtained, a global optimization algorithm is adopted to carry out multi-round iterative updating on parameters in the fault prediction model until a preset stop condition is met, the fault prediction model after parameter updating is obtained, and the convergence precision and speed of the model parameters are effectively improved.

Description

Technical Field

[0001] This application belongs to the technical field of fault prediction, and particularly relates to a method for training a fault prediction model, a fault prediction method, and a device. Background Art

[0002] With the rapid development of network technology, the scale of the network environment has gradually become complex. Therefore, in order to ensure the good operation of the network, network faults can be predicted in advance. For example, it can be predicted whether network congestion occurs, whether the network link is interrupted, or whether there is an abnormality in the devices in the network, etc.

[0003] Furthermore, according to the prediction results, network operation and maintenance can be carried out in advance for the network environment where network faults are about to occur, so as to avoid the situation of network paralysis.

[0004] In the prior art, a machine learning model can be used to predict the state indicators of the network in a future time period, and based on this indicator, it is determined whether a network fault occurs. However, during the training process of the neural network model using the prior art, the convergence speed of the model parameters is slow, and the convergence accuracy is low. Summary of the Invention

[0005] Embodiments of this application provide a method for training a fault prediction model, a fault prediction method, and a device, so as to solve the problem that the existing model parameters have a slow convergence speed and a low convergence accuracy.

[0006] In a first aspect, embodiments of this application provide a method for training a fault prediction model, and the method includes:

[0007] Obtain a training data set, where the training data set includes sample data and label data. The sample data includes first running state indicator data of a sample communication network at multiple moments within a first time period, and the label data includes second running state indicator data of the sample communication network at multiple moments within a second time period. The first time period is before the second time period;

[0008] Input the first running state indicator data into the long short-term memory network of the fault prediction model to be trained, and determine the temporal characteristics of the first running state indicator data within the first time period;

[0009] Input the temporal characteristics into the Gaussian process regression network of the fault prediction model, and process according to the temporal characteristics to obtain probability distribution data of the running state indicators of the sample communication network within the second time period;

[0010] Adopt a global optimization algorithm, and based on the probability distribution data of the operation status indicators and the second operation status indicator data, perform multiple rounds of iterative updates on the parameters in the fault prediction model until the preset stop condition is met, obtaining the fault prediction model with updated parameters, and use the fault prediction model with updated parameters as the trained fault prediction model.

[0011] Second, an embodiment of the present application provides a fault prediction method, and the method includes:

[0012] Obtain the target operation status indicator data corresponding to multiple moments within a third time period of the communication network;

[0013] Input the target operation status indicator data into the long short-term memory network of the trained fault prediction model to determine the target temporal characteristics of the target operation status indicator data within the third time period, and the fault prediction model is trained by the above-mentioned fault prediction model training method;

[0014] Input the target temporal characteristics into the Gaussian process regression network of the trained fault prediction model, and perform processing according to the target temporal characteristics to obtain the target probability distribution data of the operation status indicators of the communication network within a fourth time period, and the target probability distribution data includes: the target operation status indicator data interval of the communication network within the fourth time period and the target confidence level of the target operation status indicator data interval;

[0015] Predict whether a network fault occurs in the communication network within the fourth time period according to the target probability distribution data.

[0016] Third, an embodiment of the present application provides a fault prediction model training device, and the device includes:

[0017] An acquisition module, configured to acquire a training data set, where the training data set includes sample data and label data, the sample data includes the first operation status indicator data of the sample communication network at multiple moments within a first time period, and the label data includes the second operation status indicator data of the sample communication network at multiple moments within a second time period, and the first time period is before the second time period;

[0018] A first determination module, configured to input the first operation status indicator data into the long short-term memory network of the fault prediction model to be trained, and determine the temporal characteristics of the first operation status indicator data within the first time period;

[0019] A second determination module, configured to input the temporal characteristics into the Gaussian process regression network of the fault prediction model, and perform processing according to the temporal characteristics to obtain the probability distribution data of the operation status indicators of the sample communication network within the second time period;

[0020] An iterative update module, which is used to adopt a global optimization algorithm and, according to the probability distribution data of the operation status indicators and the second operation status indicator data, perform multiple rounds of iterative updates on the parameters in the fault prediction model until a preset stop condition is met, obtaining a fault prediction model with updated parameters, and using the fault prediction model with updated parameters as the trained fault prediction model.

[0021] Fourthly, an embodiment of the present application provides a fault prediction device, which includes:

[0022] An acquisition module, which is used to acquire the target operation status indicator data corresponding to multiple moments within a third time period of the communication network;

[0023] A first determination module, which is used to input the target operation status indicator data into the long short-term memory network of the trained fault prediction model to determine the target time series characteristics of the target operation status indicator data within the third time period, and the fault prediction model is trained by the above-mentioned fault prediction model training method;

[0024] A second determination module, which is used to input the target time series characteristics into the Gaussian process regression network of the trained fault prediction model, and perform processing according to the target time series characteristics to obtain the target probability distribution data of the operation status indicators of the communication network within a fourth time period, and the target probability distribution data includes: the target operation status indicator data interval of the communication network within the fourth time period and the target confidence level of the target operation status indicator data interval;

[0025] A prediction module, which is used to predict whether a network fault occurs in the communication network within the fourth time period according to the target probability distribution data.

[0026] Fifthly, an embodiment of the present application provides an electronic device, which includes: a processor and a memory storing computer program instructions;

[0027] When the processor executes the computer program instructions, it implements the fault prediction model training method in the first aspect or the fault prediction method in the second aspect.

[0028] Sixthly, an embodiment of the present application provides a computer-readable storage medium, on which computer program instructions are stored, and when the computer program instructions are executed by a processor, they implement the fault prediction model training method in the first aspect or the fault prediction method in the second aspect.

[0029] Seventhly, an embodiment of the present application provides a computer program product, characterized in that when the instructions in the computer program product are executed by the processor of the electronic device, the electronic device is enabled to execute the fault prediction model training method in the first aspect or the fault prediction method in the second aspect.

[0030] The fault prediction model training method provided by the embodiments of the present application can, based on the obtained training data set, adopt a global optimization algorithm, and according to the probability distribution data of the operating state indicators and the second operating state indicator data, perform multiple rounds of iterative updates on the parameters in the fault prediction model. It can significantly improve the convergence speed of the model parameters, and moreover, through the global search ability of the global optimization algorithm, effectively avoid the situation of the fault prediction model falling into local optimum, and improve the convergence accuracy of the model parameters. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] To more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings required to be used in the embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0032] Figure 1 It is a flowchart of a fault prediction model training method provided by the embodiments of the present application;

[0033] Figure 2 It is a flowchart of the specific implementation manner of S102 in a fault prediction model training method provided by the embodiments of the present application;

[0034] Figure 3 It is a flowchart of the specific implementation manner of S103 in a fault prediction model training method provided by the embodiments of the present application;

[0035] Figure 4 It is a flowchart of the specific implementation manner of S104 in a fault prediction model training method provided by the embodiments of the present application;

[0036] Figure 5 It is a flowchart of a fault prediction method provided by the embodiments of the present application;

[0037] Figure 6 It is a flowchart of the specific implementation manner of S504 in a fault prediction method provided by the embodiments of the present application;

[0038] Figure 7 It is a flowchart of the specific implementation manner of S504 in a fault prediction method provided by the embodiments of the present application;

[0039] Figure 8 It is a flowchart of the specific implementation manner before S501 in a fault prediction method provided by the embodiments of the present application;

[0040] Figure 9 It is a flowchart of the specific implementation manner before S501 in a fault prediction method provided by the embodiments of the present application;

[0041] Figure 10 It is a schematic flowchart of the specific implementation manner of the fault prediction method provided by the embodiment of the present application after S504;

[0042] Figure 11 It is a schematic structural diagram of a fault prediction model training device provided by the embodiment of the present application;

[0043] Figure 12 It is a schematic structural diagram of a fault prediction device provided by the embodiment of the present application;

[0044] Figure 13 It is a schematic hardware structure diagram of a fault prediction model training device or a fault prediction device provided by the embodiment of the present application. Detailed implementation manner

[0045] The features and exemplary embodiments of various aspects of the present application will be described in detail below. To make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain the present application, rather than limiting the present application. For those skilled in the art, the present application can be implemented without some of these specific details. The following description of the embodiments is only intended to provide a better understanding of the present application by showing examples of the present application.

[0046] It should be noted that, in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover a non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device. Without further limitation, the elements defined by the statement "including..." do not exclude the existence of additional identical elements in the process, method, article or device including the elements.

[0047] To solve the problems of the prior art, the embodiments of the present application provide a fault prediction model training method, device, electronic device, storage medium and computer program product, which can train a fault prediction model based on the obtained training data set by using a global optimization algorithm, thereby significantly improving the convergence speed of model parameters, and moreover, through the global search ability of the global optimization algorithm, effectively avoiding the situation of local optimality of the fault prediction model and improving the convergence accuracy of model parameters. The following first gives an example of the framework of the data processing method that can apply the embodiments of the present application.

[0048] First, the fault prediction model training method provided by the embodiments of the present application will be introduced below.

[0049] Figure 1 The flowchart of the fault prediction model training method provided by an embodiment of the present application is shown. The execution subject of this fault prediction model training method can be a terminal device such as a desktop computer or a laptop computer, or a client installed in the terminal device, or a server. Hereinafter, only the server is used as the execution subject to illustrate this fault prediction model training method. As Figure 1 shown, the method includes:

[0050] S101: Obtain a training data set, where the training data set includes sample data and label data. The sample data includes the first operation status index data of the sample communication network at multiple moments within a first time period, and the label data includes the second operation status index data of the sample communication network at multiple moments within a second time period. The first time period is before the second time period.

[0051] Exemplarily, the sample communication network may include a wireless local area network, a metropolitan area Ethernet, the Internet, or a satellite communication network, etc., which are networks for transmitting data, voice, video, and other information. The first operation status index data may include the network status index data of the sample communication network and the device status index data of the devices in the sample communication network. For example, the network status index data of the sample communication network may include the transmission delay, network throughput, network packet loss rate, link bandwidth utilization rate, etc. of the sample communication network, and the device status index data of the devices in the sample communication network may include the load rate of the base station in the sample communication network, the signal reception success rate of the base station, the interface traffic of the router, the packet forwarding rate of the router, the temperature of the router, the port traffic of the switch, the port error frame number of the switch, the port packet loss rate of the switch, etc. The concept of the second operation status index data is the same as that of the first operation status index data, and will not be repeated here.

[0052] For example, the first operation status index data may include the temperature data of the router in the sample communication network from April 1, 2025 to April 3, 2025, and the second operation status index data may include the temperature data of the router in the sample communication network from April 4, 2025 to April 6, 2025.

[0053] S102: Input the first operation status index data into the long short-term memory network of the fault prediction model to be trained, and determine the temporal characteristics of the first operation status index data within the first time period.

[0054] Exemplarily, the long short-term memory network is a type of recurrent neural network that captures the characteristics of the first operating state metric data changing over time through a gating mechanism, thereby providing a basis for fault warning. The temporal characteristics of the first operating state metric data within the first time period are used to characterize the changes of the first operating state metric data at multiple moments within the first time period. For example, the temporal characteristics can characterize the trend, periodicity of the first operating state metric data within the first time period, and the mutation point metric data of the first operating state metric data (such as the sudden highest point temperature data).

[0055] S103: Input the temporal characteristics into the Gaussian process regression network of the fault prediction model, process according to the temporal characteristics, and obtain the probability distribution data of the operating state metrics of the sample communication network within the second time period.

[0056] Exemplarily, the Gaussian process regression network is a Bayesian probability network model that can map the temporal characteristics of the first time period and learn the correlation relationship between the temporal characteristics of the first time period to predict the probability distribution data of the operating state metrics of the sample communication network in the future second time period. Continuing with the previous example, the temperature data of the router from April 1, 2025 to April 3, 2025 are 47.3 degrees, 47.8 degrees, and 49.2 degrees respectively. By inputting the temperature data of the router from April 1, 2025 to April 3, 2025 into the fault prediction model, it can be predicted that there is a 95% probability that the temperature data of the router on April 4, 2025 is between 49.8 degrees and 51.4 degrees, a 90% probability that the temperature data on April 5, 2025 is between 52.7 degrees and 53.7 degrees, and a 95% probability that the temperature data on April 6, 2025 is between 54 degrees and 55.6 degrees.

[0057] S104: Adopt a global optimization algorithm to perform multiple rounds of iterative updates on the parameters in the fault prediction model according to the probability distribution data of the operating state metrics and the second operating state metric data until the preset stop condition is met, obtain the fault prediction model with updated parameters, and use the fault prediction model with updated parameters as the trained fault prediction model.

[0058] Exemplarily, the global algorithm can be algorithms such as genetic algorithms and particle swarm optimization algorithms. For example, the genetic algorithm can be used to encode the model parameters to be updated in the fault prediction model as chromosomes, generate an initial population, evaluate the fitness of each chromosome based on the KL divergence between the probability distribution data of the operating state metrics and the second operating state metric data, retain the individuals with high fitness, and generate new model parameters through crossover and mutation to perform iterative updates on the parameters in the fault prediction model. When the preset stop condition is met, obtain the fault prediction model with updated parameters, and use the fault prediction model with updated parameters as the trained fault prediction model.

[0059] For another example, the particle swarm optimization algorithm can be used to calculate the particle fitness based on the probability distribution data of the operating state indicators and the second operating state indicator data, record the historical optimal parameters of each particle and the global optimal parameters of the population, and iteratively update the parameters in the fault prediction model according to these two optimal parameters until the preset stop condition is met, obtaining the fault prediction model with updated parameters, and using the fault prediction model with updated parameters as the trained fault prediction model.

[0060] Therefore, the fault prediction model training method provided by the embodiments of the present application can, based on the obtained training data set, adopt a global optimization algorithm, and iteratively update the parameters in the fault prediction model multiple times according to the probability distribution data of the operating state indicators and the second operating state indicator data. It can significantly improve the convergence speed of the model parameters, and moreover, through the global search ability of the global optimization algorithm, effectively avoid the situation of the fault prediction model falling into a local optimum, and improve the convergence accuracy of the model parameters.

[0061] The above S101 to S104 will be described in detail below, as specifically shown below.

[0062] Regarding S102, in the embodiments of the present application, the long short-term memory network of the fault prediction model can effectively extract the temporal features of the first operating state indicator data through its unique gating mechanism. Figure 2 It is a flowchart of a specific implementation manner of S102 in a fault prediction model training method provided by the embodiments of the present application. As Figure 2 shown, S102 may include S201 to S203.

[0063] The long short-term memory network includes an input gate, a forget gate, and an output gate. The gating mechanism composed of these three gates can learn the dependency relationships between the input indicator data. Among them, the input gate is used to extract the key features of the input indicator data, the forget gate is used to remove the noise information of the input indicator data, and the output gate is used to perform feature fusion and output the temporal features.

[0064] S201, through the forget gate, remove the noise information in the first operating state indicator data to obtain the denoised features.

[0065] In the embodiments of the present application, the server can first remove the noise information in the first operating state indicator data through the forget gate in the long short-term memory network to obtain the denoised features. For example, the forget coefficient can be calculated through the sigmoid function in the forget gate to determine which noise information needs to be removed. Among them, the noise information may refer to the abnormal operating state indicator data in the first operating state indicator data. For example, the temperature data that rises and falls suddenly in a short time, the bandwidth utilization data that is abnormally high at a certain moment, etc.

[0066] S202. Extract the key features from the first operating state index data through the input gate.

[0067] S203. Through the output gate, perform feature fusion on the key features and the denoised features to determine the temporal features of the first operating state index data within the first time period.

[0068] In the embodiment of the present application, the server can extract the key features from the first operating state index data through the sigmoid function and the tanh function in the input gate of the long short-term memory network. Among them, the key features can characterize the key trends when the first operating state index data changes within the first time period. For example, the first time period is from April 1, 2025 to April 5, 2025. During the period from April 1, 2025 to April 3, 2025, the change in the temperature data of the router is relatively gentle. During the period from April 4, 2025 to April 5, 2025, the temperature data of the router rises rapidly. Then, the change in the temperature data of the router during the period from April 4, 2025 to April 5, 2025 can be used as the key trend, and based on the temperature data of the router during the period from April 4, 2025 to April 5, 2025, the key features are extracted. Furthermore, the server can perform feature fusion on the key features and the denoised features through the output gate in the long short-term memory network to determine the temporal features of the first operating state index data within the first time period.

[0069] Thus, in the embodiment of the present application, through the coordinated action of the input gate, forget gate, and output gate in the long short-term memory network, the temporal features of the first operating state index data can be accurately captured, improving the accuracy of extracting temporal features.

[0070] Regarding S103, in the embodiment of the present application, based on the temporal features, the probability distribution data of the operating state index of the sample communication network within the second time period can be obtained to train the fault prediction model to be trained according to the probability distribution data. Figure 3 This is a flowchart of the specific implementation manner of S103 in a fault prediction model training method provided by the embodiment of the present application. As Figure 3 shown, S103 may include S301 to S302.

[0071] S301. Input the temporal features into the Gaussian process regression network of the fault prediction model, and based on the prior probability distribution data of the temporal features, perform Bayesian analysis to determine the posterior probability distribution data of the temporal features.

[0072] S302. According to the posterior probability distribution data, determine the probability distribution data of the operating state index of the sample communication network within the second time period.

[0073] In an embodiment of the present application, the server may input the time series features into the Gaussian process regression network of the fault prediction model, and perform Bayesian analysis based on the prior probability distribution data of the time series features to determine the posterior probability distribution data of the time series features. Among them, the prior probability distribution data of the time series features may be determined in advance based on the historical operation status index data of the sample communication network. For example, it may be the normal distribution data that the load rate of the base stations in the sample communication network conforms to within the past two months. The posterior probability distribution data of the time series features is the probability distribution data obtained after Bayesian analysis. Then, the probability distribution data of the operation status index of the sample communication network in the second time period may be determined according to the posterior probability distribution data.

[0074] The probability distribution data of the operation status index may include the operation status index data interval of the sample communication network in the second time period and the confidence level of the operation status index data interval. The operation status index data interval can represent the range of values of the operation status index data in the second time period, and the confidence level of the operation status index data interval can characterize the probability that the operation status index data is within this interval. For example, the operation status index data interval may be the temperature data interval of the router (49.8 degrees, 51.4 degrees), and the confidence level of this operation status index data interval is 95%. Then, it can be understood that there is a 95% probability that the temperature data of the router is between 49.8 degrees and 51.4 degrees in the second time period.

[0075] Thus, in an embodiment of the present application, Bayesian analysis can be performed on the time series features through the Gaussian process regression network, and the probability distribution data of the operation status index in the future second time period can be predicted, thereby improving the accuracy and practicability of fault prediction.

[0076] Regarding S104, in an embodiment of the present application, the fault prediction model to be trained may be trained based on the probability distribution data of the foregoing operation status index data, so as to perform fault prediction through the trained fault prediction model. Figure 4 It is a flowchart of a specific implementation manner of S104 in a fault prediction model training method provided by an embodiment of the present application. As Figure 4 shown, S104 may include S401 to S404.

[0077] S401, obtain the initial optimization direction of the parameters of the fault prediction model to be trained.

[0078] In an embodiment of the present application, at the beginning of training the fault prediction model to be trained, the optimization direction of the parameters of the fault prediction model may be initialized in advance, so as to obtain the initial optimization direction of the parameters of the fault prediction model to be trained. For example, initialization may be performed by means of orthogonal initialization, random normal distribution initialization, etc.

[0079] S402. Use a global optimization algorithm to update the initial optimization direction based on the probability distribution data of the operation status indicators and the second operation status indicator data, and obtain the updated optimization direction.

[0080] S403. Update the parameters in the fault prediction model to be trained according to the updated optimization direction, and obtain the updated fault prediction model.

[0081] In the embodiment of the present application, global optimization algorithms such as genetic algorithms and particle swarm optimization algorithms can be used to update the initial optimization direction based on the probability distribution data of the operation status indicators and the second operation status indicator data, and obtain the updated optimization direction. Furthermore, the parameters in the fault prediction model to be trained can be updated according to the updated optimization direction, and the updated fault prediction model can be obtained, that is, the parameters in the fault prediction model are adjusted along the updated optimization direction, so as to obtain the updated fault prediction model.

[0082] S404. When the preset stop condition is not satisfied, update the initial optimization direction to the updated optimization direction; and return to input the first operation status indicator data into the long short-term memory network of the fault prediction model to be trained, and determine the temporal characteristics of the first operation status indicator data within the first time period; until the preset stop condition is satisfied, and obtain the trained fault detection model.

[0083] In the embodiment of the present application, when training the fault prediction model, a stop condition can often be set in advance. The stop condition can be to meet the preset number of training rounds, the update amount of model parameters is less than the preset threshold, or the training time reaches the preset duration, etc. When the stop condition is not satisfied, the initial optimization direction can be updated, and the training data is input into the fault prediction model again for multiple rounds of iterative training, that is, return to input the first operation status indicator data into the long short-term memory network of the fault prediction model to be trained, and continue with iterative training until the preset stop condition is satisfied, and obtain the trained fault prediction model.

[0084] Thus, in the embodiment of the present application, the global search ability of the global optimization algorithm can be used to perform multiple rounds of iterative updates on the parameters of the model, effectively avoiding the situation of local optimality of the fault prediction model and improving the convergence accuracy of the model parameters.

[0085] Regarding S402, the probability distribution data of the operation status indicators of the sample communication network determined by the fault prediction model in the second time period may include the operation status indicator data interval and the confidence level of the operation status indicator data interval. Therefore, the server can update the initial optimization direction based on the operation status indicator data interval and the confidence level of the operation status indicator data interval.

[0086] When the probability distribution data of the operating state index includes: the operating state index data interval and the confidence level of the operating state index data interval, S402 may include: using a global optimization algorithm to update the initial optimization direction with the goal of minimizing the difference between the confidence level and a preset reference confidence level, and maximizing the probability that the operating state index data interval covers the second operating state index data, so as to obtain the updated optimization direction.

[0087] In the embodiments of the present application, by gradually reducing the difference between the confidence level and the preset reference confidence level, and gradually increasing the probability that the operating state index data interval covers the second operating state index data, a global optimization algorithm is used to update the initial optimization direction, so as to update the parameters of the fault prediction model based on the updated optimization direction, thereby enabling the fault prediction model to align the confidence level with the reference confidence level and the operating state index data interval to cover the second operating state index data to the greatest extent. Among them, the preset reference confidence level can be set by the user based on the actual situation, such as 85%, 90%, etc.

[0088] Therefore, in the embodiments of the present application, with the goal of minimizing the difference between the confidence level and the preset reference confidence level, and maximizing the probability that the operating state index data interval covers the second operating state index data, the initial optimization direction is updated, thereby improving the generalization ability and adversarial robustness of the fault prediction model, and further improving the accuracy of the fault prediction model.

[0089] In the embodiments of the present application, after the fault prediction model is trained, the trained fault prediction model can be applied to predict the faults of the communication network, such as Figure 5 shown Figure 5 is a schematic flowchart of a fault prediction method provided by an embodiment of the present application. The execution subject of this fault prediction method can be a terminal device such as a desktop computer or a laptop computer, or a client installed in the terminal device, or a server. Hereinafter, only the terminal device is used as the execution subject to illustrate this fault prediction method, as Figure 5 shown, the method includes:

[0090] S501, obtain the target operating state index data corresponding to multiple moments within a third time period of the communication network.

[0091] Exemplarily, a communication network may include a wireless local area network, a metropolitan area Ethernet network, the Internet, or a satellite communication network, etc., which are networks for transmitting information such as data, voice, and video. The target operating status indicator data is data that can represent the operating status of the communication network or the operating status of each device in the communication network, and may include the network status indicator data of the communication network and the device status indicator data of the devices in the sample communication network. For example, the network status indicator data of the communication network may include the transmission delay of the communication network, network throughput, network packet loss rate, link bandwidth utilization rate, etc., and the device status indicator data of the devices in the communication network may include the load rate of the base station in the sample communication network, the signal reception success rate of the base station, the interface traffic of the router, the packet forwarding rate of the router, the temperature of the router, the port traffic of the switch, the number of port error frames of the switch, the port packet loss rate of the switch, etc.

[0092] For example, when performing communication network fault prediction, the terminal device may obtain the temperature data of the router in the communication network during the period from April 1, 2025 to April 5, 2025, so as to predict the temperature data of the router during the period from April 6, 2025 to April 7, 2025 through the trained fault prediction model.

[0093] S502, input the target operating status indicator data into the long short-term memory network of the trained fault prediction model, and determine the target temporal characteristics of the target operating status indicator data within the third time period. The fault prediction model is trained by the above-mentioned fault prediction model training method.

[0094] S503, input the target temporal characteristics into the Gaussian process regression network of the trained fault prediction model, and process according to the target temporal characteristics to obtain the target probability distribution data of the operating status indicators of the communication network within the fourth time period. The target probability distribution data includes: the target operating status indicator data interval of the communication network within the fourth time period and the target confidence level of the target operating status indicator data interval.

[0095] Exemplarily, the target temporal characteristics can characterize the change of the target operating status indicator data at multiple moments within the third time period, the target probability distribution data can represent the distribution of the operating status indicator data of the communication network within the fourth time period, the target operating status indicator data interval can represent the range of the predicted operating status indicator data values, and the target confidence level of the target operating status indicator data interval can characterize the probability that the predicted operating status indicator data is within this interval.

[0096] Continuing with the above example, the terminal device determines the variation of the temperature data of the router over time from April 1, 2025 to April 5, 2025 through the long short-term memory network of the trained fault prediction model. The Gaussian process regression network of the fault prediction model predicts the target probability distribution data of the temperature data of the router on April 6, 2025 based on the variation of the temperature data of the router over time from April 1, 2025 to April 5, 2025. The target probability distribution data includes the target operating state index data interval (54 degrees, 55.6 degrees), and the target confidence level of the target operating state index data interval is 95%. That is, there is a 95% probability that the temperature data of the router on April 6, 2025 is between 54 degrees and 55.6 degrees.

[0097] S504. Predict whether a network failure occurs in the communication network according to the target probability distribution data.

[0098] Exemplarily, after the terminal device determines the target probability distribution data of the operating state index of the communication network in the fourth time period through the fault prediction model, it can predict whether a network failure occurs in the communication network in the fourth time period based on the target probability distribution data. Specifically, corresponding thresholds can be preset in advance, and a threshold mechanism can be used to predict whether a network failure will occur in the future.

[0099] Therefore, in the embodiments of the present application, the fault prediction model can be trained through a global optimization algorithm, so as to obtain a fault prediction model with a relatively high parameter convergence accuracy. Correspondingly, when performing fault prediction through this fault prediction model, the accuracy of the output result of the fault prediction model can be improved, and the accuracy of performing fault prediction is improved.

[0100] Regarding S504, in the embodiments of the present application, a first threshold can be preset in advance, and by comparing the first threshold with the target operating state index data interval, it can be predicted whether a network failure occurs in the communication network in the fourth time period. Figure 6 It is a schematic flowchart of the specific implementation manner of S504 in a fault prediction method provided by an embodiment of the present application. As Figure 6 shown, S504 may include S601 to S603.

[0101] S601. Determine the characteristic value of the target operating state index data interval.

[0102] S602. Predict that a network failure occurs in the communication network when the characteristic value is greater than or equal to the preset first threshold.

[0103] S603. Predict that no network failure occurs in the communication network when the characteristic value is less than the first threshold.

[0104] In the embodiment of the present application, the terminal device may first determine the characteristic value of the target operating state index data range. Specifically, the average value of the upper limit value and the lower limit value of the target operating state index data range may be determined, and the average value of the upper limit value and the lower limit value of the target operating state index data range may be used as the characteristic value of the target operating state index data range. For example, the target operating state index data range of the temperature data is (54 degrees, 55.6 degrees), and the characteristic value of this target operating state index data range is 54.8 degrees.

[0105] Alternatively, the lower limit value of the target operating state index data range may also be directly used as the characteristic value of the target operating state index data range. For example, the target operating state index data range of the temperature data is (50 degrees, 59 degrees), and the characteristic value of this target operating state index data range is 50 degrees.

[0106] After determining the characteristic value, compare the characteristic value with a preset first threshold. In the case where the characteristic value is greater than or equal to the preset first threshold, it is predicted that a network fault occurs in the communication network. In the case where the characteristic value is less than the first threshold, it is predicted that no network fault occurs in the communication network.

[0107] Thus, in the embodiment of the present application, the first threshold may be set based on actual requirements, and based on the magnitudes of the first threshold and the characteristic value, it is predicted whether a network fault will occur in the future period, realizing accurate prediction of communication network faults.

[0108] Regarding S504, in the embodiment of the present application, in addition to predicting whether a network fault occurs in the communication network by comparing the first threshold with the target operating state index data range, it is also possible to predict whether a network fault occurs in the communication network by comparing the target confidence level with a second threshold. Figure 7 It is a schematic flowchart of the specific implementation manner of S504 in a fault prediction method provided by an embodiment of the present application. As Figure 7 shown, S504 may include S701 to S702.

[0109] S701, determine the characteristic value of the target operating state index data range.

[0110] S702, in the case where the characteristic value is greater than or equal to the preset first threshold and the target confidence level is greater than or equal to the preset second threshold, determine that a network fault occurs in the communication network.

[0111] In the embodiments of the present application, a second threshold can be preset based on actual requirements. The second threshold is used to compare with the target confidence level of the target operating state index data range, so as to predict whether a network failure occurs in the communication network. There are four cases for the comparison result: when the eigenvalue is greater than or equal to the preset first threshold and the target confidence level is greater than or equal to the preset second threshold, it is determined that a network failure occurs in the communication network, and the communication network needs to be maintained in a timely manner; when the eigenvalue is less than the preset first threshold and the target confidence level is greater than or equal to the preset second threshold, it is determined that no network failure occurs in the communication network, and it is not necessary to immediately maintain the communication network; when the eigenvalue is greater than or equal to the preset first threshold and the target confidence level is less than the preset second threshold, it is determined that no network failure occurs in the communication network, and it is not necessary to immediately maintain the communication network; when the eigenvalue is less than the preset first threshold and the target confidence level is less than the preset second threshold, it is determined that no network failure occurs in the communication network, and it is not necessary to immediately maintain the communication network.

[0112] Therefore, in the embodiments of the present application, by setting the first threshold and the second threshold, and comparing these two thresholds with the eigenvalue and the target confidence level respectively, it is determined whether a network failure occurs in the future period, further ensuring the accurate prediction of communication network failures.

[0113] In the embodiments of the present application, in order to facilitate users to set corresponding thresholds based on actual requirements for predicting communication network failures, Figure 8 is a schematic flowchart of the specific implementation manner before S501 in a fault prediction method provided by an embodiment of the present application. As Figure 8 shown, before S501, the above fault prediction method further includes:

[0114] S801, display a target interface, and the target interface includes a threshold configuration option.

[0115] S802, receive the user's input to the threshold configuration option.

[0116] S803, in response to the input, determine the threshold corresponding to the input.

[0117] In the embodiments of the present application, the terminal device can display a target interface to the user. The target interface includes a threshold configuration option. The user can configure corresponding thresholds based on actual requirements by operating the threshold configuration option. After the server receives the user's input to the threshold configuration option, it can, in response to the input, determine the threshold corresponding to the input. The threshold includes the first threshold and / or the second threshold.

[0118] Thus, in the embodiments of the present application, the user can set different thresholds in the target interface according to actual needs to apply to different communication networks, ensuring the flexibility of communication network fault prediction. Moreover, by intuitively displaying the threshold configuration options in the target interface, the threshold configuration process is simplified, and the threshold configuration efficiency is improved.

[0119] In the embodiments of the present application, in order to facilitate the user to select the target operation status index data of the corresponding communication network based on actual needs, the target interface may also display index data configuration options. Figure 9 It is a schematic flowchart of a specific implementation manner before S501 in a fault prediction method provided by an embodiment of the present application. As Figure 9 shown, before S501, the above-mentioned fault prediction method further includes:

[0120] S901, display a target interface, and the target interface further includes index data configuration options.

[0121] S902, receive the user's input for the index data configuration options.

[0122] S903, in response to the input, determine the target data system corresponding to the input.

[0123] In the embodiments of the present application, the terminal device can display a target interface to the user, and the target interface further includes index data configuration options, which are used to select the target operation status index data of different communication networks. Specifically, the user can select the target data system where the target operation status index data of the corresponding communication network is located by operating the index data configuration options. The target data system can be a wireless local area network data system, a metropolitan area Ethernet data system, an Internet data system, or a satellite communication network data system, etc.

[0124] Regarding S501, S501 includes: obtaining the target operation status index data corresponding to multiple moments of the communication network in the target time period from the target data system.

[0125] In the embodiments of the present application, the user can also configure the type of the target operation status index data and the time period corresponding to the target operation status index data by operating the index data configuration options. Then, the terminal device can obtain the target operation status index data of the communication network in the corresponding time period from the target data system.

[0126] Thus, in the embodiments of the present application, the user can select the corresponding target data system in the target interface according to actual needs to obtain the target operation status index data from the target data system, improving the coverage breadth of the input data of the fault prediction model to adapt to different communication network scenarios.

[0127] In the embodiments of the present application, to facilitate users to view the results of network fault prediction, the predicted results can be displayed on the result display interface. Moreover, when it is predicted that a network fault occurs in the communication network, the policy information for resolving the network fault can also be displayed on the result display interface. Figure 10 It is a schematic flowchart of the specific implementation manner after S504 of the fault prediction method provided by the embodiments of the present application, as Figure 10 shown. The above-mentioned fault prediction method includes:

[0128] S1001, display the result display interface, where the result display interface includes the results of fault prediction.

[0129] S1002, when it is predicted that a network fault occurs in the communication network, determine the policy information for resolving the network fault.

[0130] S1003, display the policy information on the result display interface.

[0131] In the embodiments of the present application, the result display interface is used to display the results of fault prediction. Moreover, when it is predicted that a network fault occurs in the communication network, the terminal device can also determine the policy information for resolving the network fault and display the policy information on the result display interface.

[0132] For example, when the port packet loss rate of switch A in the predicted communication network is greater than a preset first threshold, the result of fault prediction is that it is predicted that a network fault occurs in the communication network, and the policy information for resolving the current network fault can be "Please restart switch A and replace the faulty network cable".

[0133] Also, for example, when the link bandwidth utilization rate of the predicted communication network is greater than a preset first threshold, the result of fault prediction is that it is predicted that a network fault occurs in the communication network, and the policy information for resolving the current network fault can be "Please enable the standby link or expand the bandwidth".

[0134] Thus, in the embodiments of the present application, based on the predicted network fault, the corresponding policy information for resolving the current network fault can be determined, thereby improving the efficiency of resolving network faults in the future.

[0135] Therefore, the technical solution provided by the embodiments of the present application can, based on the obtained training data set, adopt a global optimization algorithm, and according to the probability distribution data of the running state indicators and the second running state indicator data, perform multiple rounds of iterative updates on the parameters in the fault prediction model. It can significantly improve the convergence speed of the model parameters. Moreover, through the global search ability of the global optimization algorithm, it effectively avoids the situation that the fault prediction model has a local optimum, and improves the convergence accuracy of the model parameters.

[0136] In addition, in the embodiments of the present application, when using the trained fault prediction model to predict network faults, the target operating state index data input to the fault prediction model can also be preprocessed, which can further improve the accuracy and robustness of the prediction. For example, data cleaning, data normalization, etc. can be used for preprocessing.

[0137] As Figure 11 shown, the embodiments of the present application also provide a fault prediction model training device 1100, and the fault prediction model training device 1100 includes:

[0138] An acquisition module 1101, configured to acquire a training data set, where the training data set includes sample data and label data. The sample data includes first operating state index data of a sample communication network at multiple moments within a first time period, and the label data includes second operating state index data of the sample communication network at multiple moments within a second time period. The first time period is before the second time period;

[0139] A first determination module 1102, configured to input the first operating state index data into the long short-term memory network of the to-be-trained fault prediction model, and determine the temporal characteristics of the first operating state index data within the first time period;

[0140] A second determination module 1103, configured to input the temporal characteristics into the Gaussian process regression network of the fault prediction model, and process the temporal characteristics to obtain probability distribution data of the operating state index of the sample communication network within the second time period;

[0141] An iterative update module 1104, configured to adopt a global optimization algorithm, and perform multiple rounds of iterative updates on the parameters in the fault prediction model according to the probability distribution data of the operating state index and the second operating state index data until a preset stop condition is met, to obtain a fault prediction model with updated parameters, and use the fault prediction model with updated parameters as the trained fault prediction model.

[0142] The fault prediction model training device provided by the embodiments of the present application can, based on the acquired training data set, adopt a global optimization algorithm, and perform multiple rounds of iterative updates on the parameters in the fault prediction model according to the probability distribution data of the operating state index and the second operating state index data. It can significantly improve the convergence speed of the model parameters, and through the global search ability of the global optimization algorithm, effectively avoid the situation that the fault prediction model has a local optimum, and improve the convergence accuracy of the model parameters.

[0143] Optionally, the long short-term memory network includes an input gate, a forget gate, and an output gate;

[0144] The first determination module 1102 includes:

[0145] A removal sub-module, configured to remove the noise information in the first operating state index data through a forgetting gate to obtain denoised features;

[0146] An extraction sub-module, configured to extract key features in the first operating state index data through an input gate;

[0147] A determination sub-module, configured to perform feature fusion on the key features and the denoised features through an output gate to determine the temporal features of the first operating state index data in the first time period.

[0148] Optionally, the second determination module 1103 includes:

[0149] An analysis sub-module, configured to input the temporal features into the Gaussian process regression network of the fault prediction model, and perform Bayesian analysis based on the prior probability distribution data of the temporal features to determine the posterior probability distribution data of the temporal features;

[0150] A determination sub-module, configured to determine the probability distribution data of the operating state index of the sample communication network in the second time period according to the posterior probability distribution data.

[0151] Optionally, the iterative update module 1104 includes:

[0152] An acquisition sub-module, configured to acquire the initial optimization direction of the parameters of the fault prediction model to be trained;

[0153] A first update sub-module, configured to use a global optimization algorithm to update the initial optimization direction according to the probability distribution data of the operating state index and the second operating state index data to obtain an updated optimization direction;

[0154] A second update sub-module, configured to update the parameters in the fault prediction model to be trained according to the updated optimization direction to obtain an updated fault prediction model;

[0155] A training sub-module, configured to, when the preset stop condition is not satisfied, update the initial optimization direction to the updated optimization direction; and return to input the first operating state index data into the long short-term memory network of the fault prediction model to be trained to determine the temporal features of the first operating state index data in the first time period; until the preset stop condition is satisfied to obtain a trained fault detection model.

[0156] Optionally, the probability distribution data of the operating state index includes: the operating state index data interval and the confidence level of the operating state index data interval;

[0157] An iterative update module 1104, which uses a global optimization algorithm to update the initial optimization direction with the goal of minimizing the difference between the confidence and a preset reference confidence, and maximizing the probability that the data interval of the operating state index covers the second operating state index data, to obtain an updated optimization direction.

[0158] As Figure 12 shown in the figure, the embodiment of the present application further provides a fault prediction device 1200. The fault prediction model training device 1200 includes:

[0159] An acquisition module 1201, configured to acquire target operating state index data corresponding to multiple moments within a third time period of the communication network;

[0160] A first determination module 1202, configured to input the target operating state index data into the long short-term memory network of the trained fault prediction model to determine the target temporal characteristics of the target operating state index data within the third time period. The fault prediction model is trained by the above-mentioned fault prediction model training method;

[0161] A second determination module 1203, configured to input the target temporal characteristics into the Gaussian process regression network of the trained fault prediction model, and process according to the target temporal characteristics to obtain target probability distribution data of the operating state index of the communication network within a fourth time period. The target probability distribution data includes: the target operating state index data interval of the communication network within the fourth time period and the target confidence of the target operating state index data interval;

[0162] A prediction module 1204, configured to predict whether a network fault occurs in the communication network within the fourth time period according to the target probability distribution data.

[0163] Optionally, the prediction module 1204 includes:

[0164] A determination sub-module, configured to determine the characteristic value of the target operating state index data interval;

[0165] A first prediction sub-module, configured to predict that the communication network has a network fault when the characteristic value is greater than or equal to a preset first threshold;

[0166] A second prediction sub-module, configured to predict that the communication network has no network fault when the characteristic value is less than the first threshold.

[0167] Optionally, the prediction module 1204 includes:

[0168] A first determination sub-module, which determines the characteristic value of the target operating state index data interval;

[0169] A second determination sub-module determines that a network fault has occurred in the communication network when the eigenvalue is greater than or equal to a preset first threshold and the target confidence level is greater than or equal to a preset second threshold.

[0170] Optionally, the above-mentioned fault prediction device 1200 further includes:

[0171] A display module for displaying a target interface, where the target interface includes threshold configuration options;

[0172] A receiving module for receiving user input for the threshold configuration options;

[0173] A response module for determining, in response to the input, the threshold corresponding to the input, where the threshold includes the first threshold and / or the second threshold.

[0174] It should be noted that the fault prediction model training device 1100 is a device corresponding to the above-mentioned fault prediction model training method, and the fault prediction device 1200 is a device corresponding to the above-mentioned fault prediction method. All implementation manners in the above-mentioned fault prediction model training method embodiment are applicable to the fault prediction model training device 1100, and all implementation manners in the above-mentioned fault prediction method embodiment are applicable to the fault prediction device 1200, and the same technical effects can be achieved in the embodiments.

[0175] Figure 13 The figure shows a schematic hardware structure diagram of a fault prediction model training device or a fault prediction device provided by an embodiment of the present application.

[0176] The terminal device may include a processor 1301 and a memory 1302 storing computer program instructions.

[0177] Specifically, the above-mentioned processor 1301 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application.

[0178] The memory 1302 may include a mass storage for data or instructions. By way of example and not limitation, the memory 1302 may include a hard disk drive (HDD), a floppy disk drive, a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a universal serial bus (USB) drive or a combination of two or more of these. In a suitable case, the memory 1302 may include a removable or non-removable (or fixed) medium. In a suitable case, the memory 1302 may be inside or outside the integrated gateway disaster recovery device. In a specific embodiment, the memory 1302 is a non-volatile solid state memory.

[0179] In certain embodiments, the memory 1302 may include a read-only memory (ROM), a random access memory (RAM), a magnetic disk storage media device, an optical storage media device, a flash memory device, an electrical, optical, or other physical / tangible memory storage device. Thus, in general, the memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the method according to one aspect of the present disclosure.

[0180] The processor 1301 reads and executes the computer program instructions stored in the memory 1302 to implement any of the data processing methods in the above embodiments.

[0181] In one example, the device may further include a communication interface 1303 and a bus 1310. Among them, as Figure 13 shown, the processor 1301, the memory 1302, and the communication interface 1303 are connected through the bus 1310 and complete communication with each other.

[0182] The communication interface 1303 is mainly used to implement communication between various modules, devices, units, and / or devices in the embodiments of the present application.

[0183] The bus 1310 includes hardware, software, or both, and couples the components of the online data flow charging device to each other. By way of example and not limitation, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a HyperTransport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an InfiniBand interconnect, a Low Pin Count (LPC) bus, a memory bus, a MicroChannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses or a combination of two or more of these. In a suitable case, the bus 1310 may include one or more buses. Although the embodiments of the present application describe and illustrate specific buses, the present application contemplates any suitable bus or interconnect.

[0184] In addition, in combination with the fault prediction model training method or the above fault prediction method in the above embodiments, the embodiments of the present application may be implemented by providing a computer storage medium. Computer program instructions are stored on the computer storage medium; when the computer program instructions are executed by a processor, the fault prediction model training method or the above fault prediction method in the above embodiments is implemented.

[0185] The embodiment of the present application further provides a computer program product, including a computer program, which when executed by a processor implements the fault prediction model training method or the fault prediction method in the above embodiments.

[0186] It should be clear that the present application is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, the detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present application is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order between steps after understanding the spirit of the present application.

[0187] The functional blocks shown in the above structural block diagrams can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, it can be, for example, an electronic circuit, an application-specific integrated circuit (ASIC), appropriate firmware, a plug-in, a functional card, and so on. When implemented in software, the elements of the present application are programs or code segments used to perform the required tasks. The program or code segment can be stored in a machine-readable medium or transmitted via a data signal carried in a carrier wave on a transmission medium or a communication link. A "machine-readable medium" can include any medium that can store or transmit information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical discs, hard disks, fiber optic media, radio frequency (RF) links, and so on. The code segment can be downloaded via a computer network such as the Internet, an intranet, and so on.

[0188] It should also be noted that the exemplary embodiments mentioned in the present application describe some methods or systems based on a series of steps or devices. However, the present application is not limited to the order of the above steps, that is, the steps can be executed in the order mentioned in the embodiments, can be different from the order in the embodiments, or several steps can be executed simultaneously.

[0189] As described above with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the present disclosure. It should be understood that each block in the flowchart and / or block diagram, and the combination of blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device to produce a machine, such that the instructions executed by the processor of the computer or other programmable data processing device enable the implementation of the functions / actions specified in one or more blocks of the flowchart and / or block diagram. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field-programmable logic circuit. It should also be understood that each block in the block diagram and / or flowchart, and the combination of blocks in the block diagram and / or flowchart, can also be implemented by dedicated hardware that performs the specified functions or actions, or can be implemented by a combination of dedicated hardware and computer instructions.

[0190] The above is only the specific implementation manner of the present application. Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the systems, modules, and units described above can refer to the corresponding processes in the foregoing method embodiments, and will not be described herein again. It should be understood that the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of various equivalent modifications or substitutions within the technical scope disclosed by the present application, and these modifications or substitutions should all be covered by the protection scope of the present application.

Claims

1. A method for training a fault prediction model, characterized in that, The method includes: Obtain a training data set, where the training data set includes sample data and label data. The sample data includes first operating state index data of a sample communication network at multiple moments within a first time period, and the label data includes second operating state index data of the sample communication network at multiple moments within a second time period. The first time period is before the second time period; Input the first operating state index data into the long short-term memory network of the fault prediction model to be trained, and determine the temporal characteristics of the first operating state index data within the first time period; Input the temporal characteristics into the Gaussian process regression network of the fault prediction model, and process according to the temporal characteristics to obtain probability distribution data of the operating state index of the sample communication network within the second time period; Adopt a global optimization algorithm. According to the probability distribution data of the operating state index and the second operating state index data, perform multiple rounds of iterative updates on the parameters in the fault prediction model until a preset stop condition is met, obtain a fault prediction model with updated parameters, and use the fault prediction model with updated parameters as the trained fault prediction model.

2. The method according to claim 1, wherein The long short-term memory network includes an input gate, a forget gate, and an output gate; The step of inputting the first operating state index data into the long short-term memory network of the fault prediction model to be trained and determining the temporal characteristics of the first operating state index data within the first time period includes: Remove the noise information in the first operating state index data through the forget gate to obtain denoised features; Extract the key features in the first operating state index data through the input gate; Through the output gate, perform feature fusion on the key features and the denoised features to determine the temporal characteristics of the first operating state index data within the first time period.

3. The method according to claim 1, wherein The step of inputting the temporal characteristics into the Gaussian process regression network of the fault prediction model, processing according to the temporal characteristics, and obtaining the probability distribution data of the operating state index of the sample communication network within the second time period includes: Input the temporal characteristics into the Gaussian process regression network of the fault prediction model, and perform Bayesian analysis based on the prior probability distribution data of the temporal characteristics to determine the posterior probability distribution data of the temporal characteristics; According to the posterior probability distribution data, determine the probability distribution data of the operating state index of the sample communication network within the second time period.

4. The method according to claim 1, wherein The step of adopting a global optimization algorithm, according to the probability distribution data of the operating state index and the second operating state index data, performing multiple rounds of iterative updates on the parameters in the fault prediction model until a preset stop condition is met, obtaining a fault prediction model with updated parameters, and using the fault prediction model with updated parameters as the trained fault prediction model includes: Obtain the initial optimization direction of the parameters of the fault prediction model to be trained; Using a global optimization algorithm, update the initial optimization direction according to the probability distribution data of the operating status indicators and the second operating status indicator data to obtain an updated optimization direction; Update the parameters in the to-be-trained fault prediction model according to the updated optimization direction to obtain an updated fault prediction model; When the preset stop condition is not met, update the initial optimization direction to the updated optimization direction; and return to input the first operating status indicator data into the long short-term memory network of the to-be-trained fault prediction model to determine the temporal characteristics of the first operating status indicator data within the first time period; until the preset stop condition is met to obtain a trained fault detection model.

5. The method according to claim 4, characterized in that, The probability distribution data of the operating status indicators includes: the operating status indicator data interval and the confidence level of the operating status indicator data interval; The step of using a global optimization algorithm to update the initial optimization direction according to the probability distribution data of the operating status indicators and the second operating status indicator data to obtain an updated optimization direction includes: Taking minimizing the difference between the confidence level and a preset reference confidence level, and maximizing the probability that the operating status indicator data interval covers the second operating status indicator data as the training objective, and using a global optimization algorithm to update the initial optimization direction to obtain an updated optimization direction.

6. A fault prediction method, characterized in that, The method includes: Obtain the target operating status indicator data corresponding to multiple moments within a third time period of the communication network; Input the target operating status indicator data into the long short-term memory network of the trained fault prediction model to determine the target temporal characteristics of the target operating status indicator data within the third time period, where the fault prediction model is trained by the method according to any one of claims 1 to 5; Input the target temporal characteristics into the Gaussian process regression network of the trained fault prediction model, and process according to the target temporal characteristics to obtain the target probability distribution data of the operating status indicators of the communication network within a fourth time period, where the target probability distribution data includes: the target operating status indicator data interval of the communication network within the fourth time period and the target confidence level of the target operating status indicator data interval; Predict whether a network fault occurs in the communication network within the fourth time period according to the target probability distribution data.

7. The method according to claim 6, characterized in that The step of predicting whether a network fault occurs in the communication network within the fourth time period according to the target probability distribution data includes: Determine the characteristic value of the target operating status indicator data interval; When the characteristic value is greater than or equal to a preset first threshold, predict that a network fault occurs in the communication network; When the characteristic value is less than the first threshold, predict that no network fault occurs in the communication network.

8. The method according to claim 6, wherein The step of predicting whether a network fault occurs in the communication network within the fourth time period according to the target probability distribution data includes: Determine the characteristic value of the target operating status indicator data interval; When the eigenvalue is greater than or equal to a preset first threshold and the target confidence level is greater than or equal to a preset second threshold, it is determined that a network failure has occurred in the communication network.

9. The method according to claim 6, characterized in that, Before obtaining the target operating state index data corresponding to multiple moments within a third time period of the communication network, the method further includes: Displaying a target interface, where the target interface includes a threshold configuration option; Receiving an input from a user for the threshold configuration option; In response to the input, determining the threshold corresponding to the input, where the threshold includes a first threshold and / or a second threshold.

10. A fault prediction model training device, characterized in that, The device includes: An acquisition module, configured to acquire a training data set, where the training data set includes sample data and label data, the sample data includes first operating state index data of a sample communication network at multiple moments within a first time period, and the label data includes second operating state index data of the sample communication network at multiple moments within a second time period, and the first time period is before the second time period; A first determination module, configured to input the first operating state index data into a long short-term memory network of a fault prediction model to be trained, and determine the temporal characteristics of the first operating state index data within the first time period; A second determination module, configured to input the temporal characteristics into a Gaussian process regression network of the fault prediction model, and process according to the temporal characteristics to obtain probability distribution data of the operating state index of the sample communication network within the second time period; An iterative update module, configured to adopt a global optimization algorithm, and perform multiple rounds of iterative updates on the parameters in the fault prediction model according to the probability distribution data of the operating state index and the second operating state index data until a preset stop condition is met, to obtain a fault prediction model with updated parameters, and use the fault prediction model with updated parameters as a trained fault prediction model.

11. A fault prediction device, characterized in that, The device includes: An acquisition module, configured to acquire target operating state index data corresponding to multiple moments within a third time period of the communication network; A first determination module, configured to input the target operating state index data into a long short-term memory network of a trained fault prediction model, and determine the target temporal characteristics of the target operating state index data within the third time period, where the fault prediction model is trained by the method according to any one of claims 1 to 5; A second determination module, configured to input the target temporal characteristics into a Gaussian process regression network of the trained fault prediction model, and process according to the target temporal characteristics to obtain target probability distribution data of the operating state index of the communication network within a fourth time period, where the target probability distribution data includes: a target operating state index data interval of the communication network within the fourth time period and a target confidence level of the target operating state index data interval; A prediction module, configured to predict whether a network failure has occurred in the communication network within the fourth time period according to the target probability distribution data.

12. An electronic device, characterized in that, The electronic device includes: a processor and a memory storing computer program instructions; When the processor executes the computer program instructions, it implements the fault prediction model training method according to any one of claims 1-5 or the fault prediction method according to any one of claims 6-9.

13. A computer-readable storage medium, characterized in that, Computer program instructions are stored on the computer-readable storage medium, and when the computer program instructions are executed by a processor, the fault prediction model training method according to any one of claims 1-5 or the fault prediction method according to any one of claims 6-9 is implemented.

Images