CN-118827560-B - Network congestion early warning method, device, equipment, storage medium and program product

CN118827560BCN 118827560 BCN118827560 BCN 118827560BCN-118827560-B

Abstract

The application discloses a network congestion early warning method, a device, equipment, a storage medium and a program product. The method comprises the steps of obtaining current network performance parameters of different devices in different areas, inputting the current network performance parameters into a network congestion model, calculating a connection weight value of each neuron of the current network performance parameters in the network congestion model to obtain an optimal connection weight value, taking the neuron corresponding to the optimal connection weight value as a winning neuron, clustering the winning neurons to obtain a neighborhood neuron set, adjusting the connection weight value of each domain neuron in the neighborhood neuron set to obtain a plurality of future network states corresponding to the current network performance parameters, and carrying out maximum pooling treatment on the plurality of future network states to obtain the future network states in each time period in the future. According to the embodiment of the application, the network congestion can be early warned in advance, the network congestion phenomenon is improved, and the user experience is improved.

Inventors

YANG HAITAO
DENG XING
HE JUN
Lv Chuanyu
LUO JIAN

Assignees

中移物联网有限公司
中国移动通信集团有限公司

Dates

Publication Date: 20260512
Application Date: 20240603

Claims (10)

1. The network congestion early warning method is characterized by comprising the following steps of: acquiring current network performance parameters of different devices in a plurality of different areas; Inputting the current network performance parameter into a network congestion model, and executing the following steps on the current network performance parameter to obtain a future network state corresponding to the network performance parameter: In the network congestion model, calculating a connection weight value of each neuron of the current network performance parameter in the network congestion model to obtain an optimal connection weight value, and taking the neuron corresponding to the optimal connection weight value as a winning neuron; clustering the winning neurons to obtain a neighborhood neuron set; performing connection weight value adjustment on each neighborhood neuron in a neighborhood neuron set to obtain a plurality of future network states corresponding to the current network performance parameters; classifying the future network states according to a preset time period to obtain a plurality of classified future network states; Performing maximum pooling treatment on the classified multiple future network states to obtain future network states in each time period in the future; Calculating a connection weight value of each neuron of the current network performance parameter in the network congestion model, and obtaining an optimal connection weight value comprises the following steps: Vector conversion is carried out on the current network performance parameters to obtain a first characteristic vector of the current network performance; Acquiring an initialized connection weight value; squaring the difference between the first eigenvector and the initialized connection weight value to obtain a similarity value of each neuron; comparing the sizes of the similarity values, taking the smallest similarity value as an optimal similarity value, and taking a connection weight value corresponding to the optimal similarity value as an optimal connection weight value; The clustering of the winning neurons to obtain a neighborhood neuron set includes: Acquiring an initial influence range of the network congestion model; calculating a first distance value of each neuron from the winning neuron; Calculating an influence distance value according to the initial influence range, the preset time step number, the preset attenuation coefficient and the first distance value, wherein the influence distance value is the distance value of the winning neuron on the adjacent neuron which generates influence; Determining a neighborhood according to the winning neuron and the influence distance value, wherein the neighborhood comprises a region taking the winning neuron as a circle center and the influence distance value as a radius; taking all neurons in the neighborhood as neighborhood neurons to form a neighborhood neuron set; the adjusting the connection weight value of each neighborhood neuron in the neighborhood neuron set comprises: acquiring an adaptive learning rate, a first feature vector and a current connection weight value of a current neuron; Calculating a weight change amount according to the adaptive learning rate, the influence distance value, the first feature vector and the connection weight value; and adjusting the current connection weight value of each neuron according to the weight change quantity to obtain the future network state of the current neuron mapping.
2. The method of claim 1, wherein the obtaining current network performance parameters for a plurality of different area different devices comprises: Acquiring original network performance parameters by using a preset network tool, wherein the original network performance parameters comprise bandwidth utilization rate, data packet loss rate, data packet delay and flow size; And carrying out data preprocessing on the original network performance parameters to obtain the current network performance parameters, wherein the data preprocessing comprises at least one processing mode of standardization processing, missing value processing and normalization processing.
3. The method of claim 1, wherein prior to calculating the connection weight value for each neuron of the current network performance parameter in the network congestion model to obtain an optimal connection weight value, the method further comprises: initializing the connection weight value of each neuron in the network congestion model according to a preset connection weight value to obtain an initialized connection weight value.
4. A method according to claim 3, wherein before calculating the connection weight value for each neuron of the current network performance parameter in the network congestion model, the method further comprises: acquiring an initial learning rate; and carrying out attenuation adjustment on the initial learning rate according to a preset learning rate attenuation speed to obtain an adaptive learning rate.
5. The method of claim 1, wherein after said maximizing said plurality of categorized future network states to obtain future network states for each time period in the future, said method further comprises: Acquiring first historical network state values of first preset historical periods of different devices in the plurality of different areas, wherein the first historical network state values comprise network state values of the first preset historical periods of the current device; Calculating an autoregressive coefficient of the historical network state value according to the first historical network state value; Acquiring first historical white noise, wherein the first historical white noise comprises white noise of a first preset historical period of current equipment; calculating a moving average coefficient of white noise according to the first historical white noise; acquiring a current network state value and a white noise error term; And calculating a network state value at a future time according to the autoregressive coefficient, the moving average coefficient, the current network state value and the white noise error value.
6. The method of claim 5, wherein after calculating the network state value for the future time instant from the autoregressive coefficients, moving average coefficients, current network state value, and white noise error value, the method further comprises: Under the condition that the number of the preset early warning thresholds is one, the network state value at the future moment is larger than the preset early warning threshold, and the network state value at the future moment is used as early warning information and sent to a preset terminal, or Dividing a plurality of early warning intervals according to the preset early warning threshold values under the condition that the number of the preset early warning threshold values is a plurality of; Determining the early warning level of the future moment according to the network state value of the future moment and a plurality of early warning intervals; and taking the early warning grade and the network state value at the future moment as early warning information and sending the early warning information to a preset terminal.
7. A network congestion warning device, the device comprising: the acquisition module is used for acquiring current network performance parameters of different devices in a plurality of different areas; the input module is used for inputting the current network performance parameters into a network congestion model, and executing the following steps on the current network performance parameters to obtain the future network state corresponding to the network performance parameters: The input module comprises: the computing unit is used for computing the connection weight value of each neuron of the current network performance parameter in the network congestion model to obtain an optimal connection weight value, and taking the neuron corresponding to the optimal connection weight value as a winning neuron, and is also used for: vector conversion is carried out on the current network performance parameters to obtain a first feature vector of the current network performance; Acquiring an initialized connection weight value; The first eigenvector and the initialized connection weight value are subjected to difference square to obtain a similarity value of each neuron; the method comprises the steps of comparing the sizes of all similarity values, taking the smallest similarity value as the best similarity value and taking the connection weight value corresponding to the best missing value as the best connection weight value, clustering the winning neurons to obtain a neighborhood neuron set, and further comprising the following steps: acquiring an initial influence range of a network congestion model; calculating a first distance value of each neuron from the winning neuron; calculating an influence distance value according to the initial influence range, the preset time step number, the preset attenuation coefficient and the first distance value, wherein the influence distance value is the distance value of the winning neuron on the adjacent neuron which generates influence; Determining a neighborhood according to the winning neuron and the influence distance value, wherein the neighborhood comprises a region taking the winning neuron as a circle center and the influence distance value as a radius; taking all neurons in the neighborhood as neighborhood neurons to form a neighborhood neuron set; the adjusting unit is used for adjusting the connection weight value of each neighborhood neuron in the neighborhood neuron set to obtain a plurality of future network states corresponding to the current network performance parameters, and is also used for: acquiring an adaptive learning rate, a first feature vector and a current connection weight value of a current neuron; Calculating weight variation according to the adaptive learning rate, the influence distance value, the first feature vector and the connection weight value; adjusting the current connection weight value of each neuron according to the weight variation to obtain the future network state of the current neuron mapping; The classifying unit is used for classifying the future network states according to a preset time period to obtain a plurality of classified future network states; And the pooling unit is used for carrying out maximum pooling treatment on the classified multiple future network states to obtain the future network state in each time period in the future.
8. An electronic device comprising a processor and a memory storing computer program instructions; The processor, when executing the computer program instructions, implements the network congestion warning method according to any of claims 1-6.
9. A computer readable storage medium, wherein computer program instructions are stored on the computer readable storage medium, which when executed by a processor, implement the network congestion warning method according to any of claims 1-6.
10. A computer program product, characterized in that instructions in the computer program product, when executed by a processor of an electronic device, cause the electronic device to perform the network congestion warning method according to any of claims 1-6.

Description

Network congestion early warning method, device, equipment, storage medium and program product Technical Field The application belongs to the technical field of network detection, and particularly relates to a network congestion early warning method, device, equipment, storage medium and program product. Background At present, with the development of internet technology, the internet is an indispensable part of people's daily life. When users using the network gradually increase, network resources are impacted, so that network congestion is caused, and the use of people's network is affected. Network congestion refers to a continuously overloaded network state, and due to factors such as limited bandwidth, storage space, processor capacity and the like, the requirement of users on network resources exceeds inherent processing capacity, and when network congestion occurs, the problems of resource shortage, packet loss, delay and the like occur in the process of data packet transmission. There is a need for a network congestion warning method, apparatus, device, storage medium and program product. Disclosure of Invention The embodiment of the application provides a network congestion early warning method, a device, equipment, a storage medium and a program product, which can early warn network congestion in advance, improve network congestion phenomenon and improve user experience. In one aspect, an embodiment of the present application provides a network congestion early warning method, where the method includes: acquiring current network performance parameters of different devices in a plurality of different areas; Inputting the current network performance parameter into a network congestion model, and executing the following steps on the current network performance parameter to obtain a future network state corresponding to the network performance parameter: In the network congestion model, calculating a connection weight value of each neuron of the current network performance parameter in the network congestion model to obtain an optimal connection weight value, and taking the neuron corresponding to the optimal connection weight value as a winning neuron; clustering the winning neurons to obtain a neighborhood neuron set; performing connection weight value adjustment on each domain neuron in a neighborhood neuron set to obtain a plurality of future network states corresponding to the current network performance parameters; classifying the future network states according to a preset time period to obtain a plurality of classified future network states; And carrying out maximum pooling treatment on the classified multiple future network states to obtain the future network states in each time period in the future. Optionally, the obtaining the current network performance parameter includes: Acquiring original network performance parameters by using a preset network tool, wherein the original network performance parameters comprise bandwidth utilization rate, data packet loss rate, data packet delay and flow size; And carrying out data preprocessing on the original network performance parameters to obtain the current network performance parameters, wherein the data preprocessing comprises at least one processing mode of standardization processing, missing value processing and normalization processing. Optionally, before calculating the connection weight value of each neuron of the current network performance parameter in the network congestion model to obtain the optimal connection weight value, the method further includes: initializing the connection weight value of each neuron in the network congestion model according to a preset connection weight value to obtain an initialized connection weight value. Optionally, before calculating the connection weight value of each neuron of the current network performance parameter in the network congestion model to obtain the optimal connection weight value, the method further includes: acquiring an initial learning rate; and carrying out attenuation adjustment on the initial learning rate according to a preset learning rate attenuation speed to obtain an adaptive learning rate. Optionally, the calculating a connection weight value of each neuron of the current network performance parameter in the network congestion model, and obtaining an optimal connection weight value includes: Vector conversion is carried out on the current network performance parameters to obtain a first characteristic vector of the current network performance; Acquiring an initialized connection weight value; squaring the difference between the first eigenvector and the initialized connection weight value to obtain a similarity value of each neuron; and comparing the sizes of the similarity values, taking the smallest similarity value as an optimal similarity value, and taking the connection weight value corresponding to the optimal similarity value as an optimal connection weight value. Optiona