A network bandwidth overrun analysis method, device, apparatus and storage medium

Abstract

The application provides a network bandwidth over-limit analysis method and device, equipment and a storage medium, and relates to the technical field of network bandwidth analysis.The method comprises the following steps: acquiring first network index data of a target network, wherein the first network index data represents the bandwidth condition of a network node; inputting the first network index data into a bandwidth over-limit analysis model; and acquiring bandwidth over-limit probability information output by the bandwidth over-limit analysis model, wherein the bandwidth over-limit probability information represents the probability of bandwidth over-limit of the network node.Based on the actual bandwidth condition of the target network, the bandwidth over-limit condition is predicted and known in advance by analyzing the probability of bandwidth over-limit, the network configuration is adjusted in time according to the size of the probability, the probability of bandwidth over-limit is reduced, the effect of early prevention is achieved by avoiding bandwidth over-limit, effective quantitative analysis data is provided for network operation, and the quality and stability of network bandwidth are improved.

Description

Technical Field

[0001] This invention relates to the field of network bandwidth analysis technology, and in particular to a method, apparatus, device, and storage medium for analyzing network bandwidth overruns. Background Technology

[0002] Packet-based networks are currently the most common network type in communication networks. Their bandwidth statistical multiplexing method greatly improves the transmission efficiency of the entire communication network, leading to their widespread use. When carrying services, packet-based networks do not allow all services to reach their peak bandwidth simultaneously, thus enabling a larger number of services to be carried with a smaller bandwidth.

[0003] In practical packet networks, the bandwidth used by various services is random. Each service may use peak bandwidth, but the probability of all services using peak bandwidth simultaneously is very small. Therefore, when operating and maintaining packet networks, the bandwidth of various services is generally configured according to the bandwidth convergence ratio. For example, in a PON network, when a single GPON port connects 64 100M broadband users, the bandwidth convergence ratio for the 100M broadband service is set to 33%. Thus, the average bandwidth allocated to each 100M broadband user is 100M * 33% = 33M, and the 64 users share 33M * 64 = 2112M of bandwidth. Since most of the time the bandwidth used by users is less than 33M, when a small number of users simultaneously use 100M, the 2112M bandwidth capacity can still meet the bandwidth needs of all users.

[0004] However, due to the current highly unpredictable nature of business operations, although the probability of multiple users and multiple services simultaneously reaching peak bandwidth is small, it still exists. In live networks with a large number of users and long operating times, it is possible for a large number of users to simultaneously reach peak bandwidth usage, i.e., network bandwidth exceeding limits. This directly leads to the loss of a large number of user data packets, thereby affecting service quality and user experience. Existing bandwidth convergence ratio configuration methods cannot assess the probability of exceeding the combined configured bandwidth for multiple users in a live network, i.e., the probability of network bandwidth exceeding limits.

[0005] Therefore, timely analysis of the probability of network bandwidth exceeding limits, in order to adjust network configuration in a timely manner and improve the quality and stability of network bandwidth, is an urgent technical problem that needs to be solved. Summary of the Invention

[0006] This invention provides a method, apparatus, device, and storage medium for analyzing network bandwidth limits, thereby addressing the shortcomings of existing technologies that cannot assess the probability of network bandwidth exceeding limits.

[0007] This invention provides a method for analyzing network bandwidth overruns, comprising:

[0008] Obtain the first network indicator data of the target network, whereby the first network indicator data represents the bandwidth status of network nodes;

[0009] The first network indicator data is input into the bandwidth overrun analysis model to obtain the bandwidth overrun probability information output by the bandwidth overrun analysis model. The bandwidth overrun probability information represents the probability that a network node will exceed the bandwidth limit.

[0010] The bandwidth over-limit analysis model is obtained by training an artificial intelligence model based on a network over-limit dataset.

[0011] According to a network bandwidth overrun analysis method provided by the present invention, the steps for obtaining the bandwidth overrun analysis model include:

[0012] According to the sampling period, second network index data is obtained from the target network as a training dataset;

[0013] Obtain the network node topology data of the target network;

[0014] Based on the preset bandwidth over-limit standard and the training dataset, the over-limit time information is determined, which represents the time and duration when the network node is in the bandwidth over-limit state.

[0015] Based on the sampling period and the over-limit time information, a probability density function is calculated as training label data;

[0016] The artificial intelligence model is trained based on the training dataset, the training label data, and the network node topology data to obtain the bandwidth overrun analysis model.

[0017] According to a network bandwidth overrun analysis method provided by the present invention, the artificial intelligence model includes a first network layer, a second network layer, and a third network layer; the step of training the artificial intelligence model based on the training dataset, the training label data, and the network node topology relationship data to obtain the bandwidth overrun analysis model includes:

[0018] The data from the training dataset that correspond to the same time point and the same network node are combined to form the first sample data;

[0019] The first sample data is input into the first network layer for integration processing, so that the format of each first sample data is unified and the first sample data with adjacent time sequence are correlated to form the second sample data;

[0020] The second sample data and the network node topology relationship data are input into the second network layer to construct the dependency relationship between the second sample data according to the dependency relationship between network nodes, thereby forming the third sample data;

[0021] The third sample data is input into the third network layer to perform feature extraction and dimensionality reduction on the third sample data, and obtain the bandwidth over-limit probability prediction result.

[0022] Based on a preset loss function, the bandwidth over-limit probability prediction result and the training label data are calculated to obtain error information;

[0023] The first network layer, the second network layer, and the third network layer are optimized based on the error information until the model converges or reaches a preset number of iterations, and the artificial intelligence model is used as the bandwidth over-limit analysis model.

[0024] According to a network bandwidth overrun analysis method provided by the present invention, the optimization of the first network layer, the second network layer, and the third network layer based on the error information includes:

[0025] Calculate the gradient of the loss function with respect to the model parameters using the chain rule;

[0026] The parameters of the first network layer, the second network layer, and the third network layer are updated using stochastic gradient descent based on the gradient.

[0027] According to the network bandwidth overrun analysis method provided by the present invention, the first network layer includes a multi-layer bidirectional long short-term memory network, the second network layer includes a multi-layer graph attention network, and the third network layer includes a multi-layer perceptron network.

[0028] According to the network bandwidth overrun analysis method provided by the present invention, the step of obtaining second network indicator data from the target network as a training dataset according to the sampling period includes:

[0029] Obtain the sampling period and sampling frequency;

[0030] Within the sampling period, the index data of each network node in the target network is recorded according to the sampling frequency to form the second network index data.

[0031] The present invention also provides a network bandwidth overrun analysis device, comprising:

[0032] The input module is used to obtain the first network metric data of the target network;

[0033] The bandwidth over-limit analysis module is used to input the first network indicator data into the bandwidth over-limit analysis model and obtain the bandwidth over-limit probability information output by the bandwidth over-limit analysis model.

[0034] The bandwidth over-limit analysis model is obtained by training an artificial intelligence model based on a network over-limit dataset.

[0035] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement a network bandwidth overrun analysis method as described above.

[0036] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a network bandwidth overrun analysis method as described above.

[0037] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements a network bandwidth overrun analysis method as described above.

[0038] This invention provides a network bandwidth overrun analysis method, apparatus, device, and storage medium, which has at least the following beneficial effects: During the actual operation of the target network, first network indicator data reflecting the bandwidth situation is acquired. This first network indicator data is processed through a bandwidth overrun analysis model to obtain bandwidth overrun probability information, enabling the prediction of the probability of bandwidth overruns. This facilitates timely adjustments to network configuration to avoid bandwidth overrun problems. Therefore, based on the actual bandwidth situation of the target network, by analyzing the probability of bandwidth overruns, the possibility of bandwidth overruns can be predicted in advance. The network configuration can be adjusted in a timely manner according to the probability, reducing the probability of bandwidth overruns and achieving a preventative effect. This provides effective quantitative analysis data for network operation, improving the quality and stability of network bandwidth. Attached Figure Description

[0039] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0040] Figure 1 This is one of the flowcharts of a network bandwidth over-limit analysis method provided by the present invention;

[0041] Figure 2This is the second flowchart of a network bandwidth over-limit analysis method provided by the present invention;

[0042] Figure 3 This is the third flowchart of a network bandwidth over-limit analysis method provided by the present invention;

[0043] Figure 4 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0044] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0045] The following is combined Figures 1-3 A network bandwidth overrun analysis method according to the present invention includes:

[0046] S100: Obtain the first network indicator data of the target network, wherein the first network indicator data represents the bandwidth status of the network nodes;

[0047] S110: Input the first network indicator data into the bandwidth overrun analysis model, and obtain the bandwidth overrun probability information output by the bandwidth overrun analysis model. The bandwidth overrun probability information represents the probability that a network node will exceed the bandwidth limit.

[0048] The bandwidth over-limit analysis model is obtained by training an artificial intelligence model based on a network over-limit dataset.

[0049] During the actual operation of the target network, primary network indicator data reflecting bandwidth status is acquired. This data is then processed using a bandwidth overrun analysis model to obtain bandwidth overrun probability information. This allows for prediction of the probability of bandwidth overruns, facilitating timely adjustments to network configuration to prevent such issues. Therefore, based on the actual bandwidth of the target network, by analyzing the probability of bandwidth overruns, situations can be predicted in advance. Adjusting network configuration according to the probability reduces the likelihood of bandwidth overruns, achieving a preventative effect. This provides effective quantitative analysis data for network operation, improving network bandwidth quality and stability.

[0050] The bandwidth over-limit probability information can include the bandwidth over-limit probability distribution of each network node in the target network, so that the network configuration of a specific network node can be adjusted based on the bandwidth over-limit probability information, or bandwidth configuration suggestions can be generated based on the bandwidth over-limit probability information.

[0051] refer to Figure 2 In some embodiments of the network bandwidth overrun analysis method of the present invention, the step of obtaining the bandwidth overrun analysis model includes:

[0052] S200: Obtain second network index data from the target network as a training dataset according to the sampling period;

[0053] S210: Obtain the network node topology data of the target network;

[0054] S220: Based on the preset bandwidth over-limit standard and the training dataset, determine the over-limit time information, wherein the over-limit time information represents the time and duration when the network node is in the bandwidth over-limit state;

[0055] S230: Calculate the probability density function as training label data based on the sampling period and the over-limit time information;

[0056] S240: Train the artificial intelligence model based on the training dataset, the training label data, and the network node topology relationship data to obtain the bandwidth over-limit analysis model.

[0057] The second network indicator data of the target network is acquired during the sampling period and used as the training set data for training a bandwidth overrun analysis model. Based on a preset bandwidth standard, it is determined whether the target network is in an overrun state. When an overrun state is detected, the time of occurrence and duration are recorded, forming overrun time information. A probability density function is calculated based on the sampling period and overrun time information and used as training label data. This actual overrun probability is compared with the model's predicted overrun probability during subsequent model training to optimize the model. In this way, the artificial intelligence model is trained using the training dataset, training label data, and network node topology data to obtain a bandwidth overrun analysis model capable of performing network bandwidth overrun analysis based on the first network indicator data of the target network.

[0058] It is important to emphasize that network node topology data was used during model training. Regarding the issue of bandwidth exceeding limits, since there are convergence and distribution relationships between network nodes, the dependencies between network nodes in the target network were further considered. That is, network topology data was used to comprehensively consider the bandwidth and traffic relationships between network nodes. For a single network node in the target network, data from the network nodes associated with it were also used for analysis. This allows for a more accurate analysis and prediction of the probability of bandwidth exceeding limits for that network node, thereby improving the accuracy of the overall target network bandwidth analysis.

[0059] It is understandable that the data types included in the first network indicator data and the second network indicator data are actually the same. The terms "first" and "second" are only used to distinguish the data obtained in the actual detection process from the training process.

[0060] The first and second network indicator data may include user maximum bandwidth, OLT device uplink and downlink rates, OLT device PON port uplink and downlink rates, ONU device uplink and downlink rates, OLT device uplink and downlink packet loss rate (or packet loss duration), OLT device PON port uplink and downlink packet loss rate (or packet loss duration), ONU device uplink and downlink packet loss rate (or packet loss duration), etc.

[0061] Network nodes can include OLT devices, ONU devices, user equipment, etc.

[0062] In some embodiments of the present invention, the preset bandwidth over-limit standard may be traffic exceeding 70% of the bandwidth. When the uplink and downlink traffic of a network node exceeds 70% of the bandwidth, the over-limit time and duration of the network node are recorded, and the over-limit time and duration of each network node form over-limit time information.

[0063] refer to Figure 3 In some embodiments of the network bandwidth overrun analysis method of the present invention, the artificial intelligence model includes a first network layer, a second network layer, and a third network layer; S240 includes:

[0064] S241: Combine the data in the training dataset that correspond to the same time point and the same network node to form the first sample data;

[0065] S242: Input the first sample data into the first network layer for integration processing, so that the format of each first sample data is unified and the first sample data with adjacent time sequence are correlated to form the second sample data;

[0066] S243: Input the second sample data and the network node topology relationship data into the second network layer to construct the dependency relationship between the second sample data according to the dependency relationship between network nodes, and form the third sample data;

[0067] S244: Input the third sample data into the third network layer to perform feature extraction and dimensionality reduction on the third sample data, and obtain the bandwidth over-limit probability prediction result;

[0068] S245: Calculate the bandwidth over-limit probability prediction result and the training label data according to the preset loss function to obtain error information;

[0069] S246: Optimize the first network layer, the second network layer, and the third network layer based on the error information until the model converges or reaches a preset number of iterations, and use the artificial intelligence model as the bandwidth over-limit analysis model.

[0070] The process of training a model based on data to build a bandwidth overrun analysis model:

[0071] First, different types of data corresponding to the same time point and the same network node in the training dataset are combined to form the first sample data, which forms a data foundation model that is easy to input into the artificial intelligence model for processing.

[0072] Then, the first sample data is input into the artificial intelligence model. The first network layer in the artificial intelligence model processes the first sample data to make the format of the combined and spliced ​​first sample data uniform. Based on the time sequence, the first sample data with adjacent time sequences are correlated to form the second sample data. Based on the second sample data, the probability of bandwidth exceeding the limit can be analyzed according to the data changes in the time sequence, which helps to improve the accuracy of analyzing the probability of bandwidth exceeding the limit.

[0073] Then, the second network layer in the artificial intelligence model constructs the dependency relationship between the second sample data based on the network node topology relationship data. That is, based on the working dependency relationship between network nodes, the second sample data corresponding to the network nodes are also associated to form the third sample data. This allows for the comprehensive consideration of the data of network nodes with dependency relationships when analyzing the probability of bandwidth exceeding the limit of network nodes, which is conducive to further improving the accuracy of the analysis of the probability of bandwidth exceeding the limit.

[0074] Finally, the third network layer in the AI ​​model performs feature extraction, dimensionality reduction, and analysis on the third sample data to obtain the bandwidth overrun probability prediction result. The training label data, i.e., the density probability function corresponding to actual bandwidth overrun in the target network, is compared with the bandwidth overrun probability prediction result. The error information between the two is calculated based on the loss function, which can be understood as the accuracy of the bandwidth overrun probability prediction result output by the AI ​​model. Then, based on the error information, the parameters of the AI ​​model, such as the parameters of the first, second, and third network layers, are optimized to improve the accuracy of the AI ​​model. Training is considered complete when the model converges or reaches the preset number of iterations, and the AI ​​model is then used as a bandwidth overrun analysis model.

[0075] Model convergence can be defined as the model meeting the following conditions: the error information obtained from the loss function calculation is stable, the accuracy of the artificial intelligence model is stable, and the parameter optimization of the artificial intelligence model has small fluctuations.

[0076] In some embodiments of the network bandwidth overrun analysis method of the present invention, the step according to S246 includes:

[0077] Calculate the gradient of the loss function with respect to the model parameters using the chain rule;

[0078] The parameters of the first network layer, the second network layer, and the third network layer are updated using stochastic gradient descent based on the gradient.

[0079] In the process of optimizing artificial intelligence models, the chain rule is used to calculate the gradient of the loss function with respect to the model parameters, and the stochastic gradient descent method is used to update the parameters of the artificial intelligence model. This can improve training efficiency, convergence speed and generalization ability, thereby enabling the final bandwidth overrun analysis model to more accurately predict the probability of network bandwidth overrun.

[0080] In some embodiments of the network bandwidth overrun analysis method of the present invention, the first network layer includes a multi-layer bidirectional long short-term memory network, the second network layer includes a multi-layer graph attention network, and the third network layer includes a multi-layer perceptron network.

[0081] Multi-layer bidirectional long short-term memory (LSTM) networks can capture bidirectional information in data, that is, they consider both past and future information at each time step, thereby obtaining more comprehensive contextual information, improving the model's expressive power and generalization ability, and meeting the need to associate temporally adjacent first sample data to form second sample data.

[0082] Multi-layer graph attention networks can learn global relationships. The attention mechanism of each layer can adaptively learn the dependencies and importance between nodes, which improves the model's ability to understand graph data and meets the need to construct the dependencies between second sample data based on the network node topology relationship data.

[0083] Multi-layer perceptron (MLP) networks offer flexible network structures and powerful nonlinear modeling capabilities, making them suitable for various data types and tasks. By stacking multiple hidden layers and introducing nonlinear activation functions, MLP networks can effectively fit complex data distributions and patterns, exhibiting good adaptability and generalization ability. Therefore, MLP networks can meet the need to fit the relationship between second-level network index data and the network bandwidth exceedance probability. Consequently, after training with large datasets, MLP networks can accurately output bandwidth exceedance probability prediction results based on third-level sample data generated from the second-level network index data.

[0084] Therefore, based on the required functional requirements, a multi-layer bidirectional long short-term memory network, a multi-layer graph attention network, and a multi-layer perceptron network were selected as the first, second, and third network layers of the artificial intelligence model, respectively. This ensures that the performance of the artificial intelligence model meets the requirements and makes the bandwidth over-limit probability information obtained from the final analysis based on the first network index data more accurate and reliable.

[0085] In some embodiments of the network bandwidth overrun analysis method of the present invention, step S200 includes:

[0086] Obtain the sampling period and sampling frequency;

[0087] Within the sampling period, the index data of each network node in the target network is recorded according to the sampling frequency to form the second network index data.

[0088] Based on the input sampling period and sampling frequency, data generated by each network node during the actual operation of the target network is collected, such as uplink and downlink traffic and corresponding bandwidth configurations, to form second network indicator data. Therefore, different sampling periods and frequencies can be set according to the characteristics of the target network, such as its type and the services it carries, to selectively collect and form second network indicator data. For example, a smaller sampling frequency can be set for a relatively stable target network, while a longer sampling period can be set for a target network with greater variability, in order to more accurately obtain the probability distribution function as training label data.

[0089] To better understand the sampling period and sampling frequency, let's take an example. If the sampling period is one week and the sampling frequency is 1 second, then 1 week × 7 days / week × 24 hours / day × 60 minutes / hour × 60 seconds / minute = 604,800 data points will be collected during the sampling period to form the second network indicator data.

[0090] The following describes a network bandwidth overrun analysis device provided by the present invention. The network bandwidth overrun analysis device described below and the network bandwidth overrun analysis method described above can be referred to in correspondence.

[0091] The present invention also provides a network bandwidth overrun analysis device, comprising:

[0092] The input module is used to obtain the first network metric data of the target network;

[0093] The bandwidth over-limit analysis module is used to input the first network indicator data into the bandwidth over-limit analysis model and obtain the bandwidth over-limit probability information output by the bandwidth over-limit analysis model.

[0094] The bandwidth over-limit analysis model is obtained by training an artificial intelligence model based on a network over-limit dataset.

[0095] Figure 4 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 4 As shown, the electronic device may include a processor 810, a communications interface 820, a memory 830, and a communication bus 840. The processor 810, communications interface 820, and memory 830 communicate with each other via the communication bus 840. The processor 810 can call logical instructions from the memory 830 to execute the aforementioned network bandwidth over-limit analysis method.

[0096] Furthermore, the logical instructions in the aforementioned memory 830 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0097] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer is able to execute a network bandwidth overrun analysis method provided by the above methods.

[0098] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform a network bandwidth overrun analysis method provided by the methods described above.

[0099] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0100] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0101] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for analyzing network bandwidth exceeding limits, characterized in that, include: Obtain the first network indicator data of the target network, whereby the first network indicator data represents the bandwidth status of network nodes; The first network indicator data is input into the bandwidth overrun analysis model to obtain the bandwidth overrun probability information output by the bandwidth overrun analysis model. The bandwidth overrun probability information represents the probability that a network node will exceed the bandwidth limit. The bandwidth over-limit analysis model is obtained by training an artificial intelligence model based on a network over-limit dataset.

2. The network bandwidth overrun analysis method according to claim 1, characterized in that, The steps for obtaining the bandwidth over-limit analysis model include: According to the sampling period, second network index data is obtained from the target network as a training dataset; Obtain the network node topology data of the target network; Based on the preset bandwidth over-limit standard and the training dataset, the over-limit time information is determined, which represents the time and duration when the network node is in the bandwidth over-limit state. Based on the sampling period and the over-limit time information, a probability density function is calculated as training label data; The artificial intelligence model is trained based on the training dataset, the training label data, and the network node topology data to obtain the bandwidth overrun analysis model.

3. The network bandwidth overrun analysis method according to claim 2, characterized in that, The artificial intelligence model includes a first network layer, a second network layer, and a third network layer; the step of training the artificial intelligence model based on the training dataset, the training label data, and the network node topology data to obtain the bandwidth overrun analysis model includes: The data from the training dataset that correspond to the same time point and the same network node are combined to form the first sample data; The first sample data is input into the first network layer for integration processing, so that the format of each first sample data is unified and the first sample data with adjacent time sequence are correlated to form the second sample data; The second sample data and the network node topology relationship data are input into the second network layer to construct the dependency relationship between the second sample data according to the dependency relationship between network nodes, thereby forming the third sample data; The third sample data is input into the third network layer to perform feature extraction and dimensionality reduction on the third sample data, and obtain the bandwidth over-limit probability prediction result. Based on a preset loss function, the bandwidth over-limit probability prediction result and the training label data are calculated to obtain error information; The first network layer, the second network layer, and the third network layer are optimized based on the error information until the model converges or reaches a preset number of iterations, and the artificial intelligence model is used as the bandwidth over-limit analysis model.

4. The network bandwidth overrun analysis method according to claim 3, characterized in that, The optimization of the first network layer, the second network layer, and the third network layer based on the error information includes: Calculate the gradient of the loss function with respect to the model parameters using the chain rule; The parameters of the first network layer, the second network layer, and the third network layer are updated using stochastic gradient descent based on the gradient.

5. The network bandwidth overrun analysis method according to claim 3, characterized in that, The first network layer includes a multi-layer bidirectional long short-term memory network, the second network layer includes a multi-layer graph attention network, and the third network layer includes a multi-layer perceptron network.

6. The network bandwidth overrun analysis method according to claim 2, characterized in that, The step of obtaining second network metric data from the target network as a training dataset according to the sampling period includes: Obtain the sampling period and sampling frequency; Within the sampling period, the index data of each network node in the target network is recorded according to the sampling frequency to form the second network index data.

7. A network bandwidth overrun analysis device, characterized in that, include: The input module is used to obtain the first network metric data of the target network; The bandwidth over-limit analysis module is used to input the first network indicator data into the bandwidth over-limit analysis model and obtain the bandwidth over-limit probability information output by the bandwidth over-limit analysis model. The bandwidth over-limit analysis model is obtained by training an artificial intelligence model based on a network over-limit dataset.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements a network bandwidth over-limit analysis method as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements a network bandwidth overrun analysis method as described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements a network bandwidth overrun analysis method as described in any one of claims 1 to 6.

Images