CN113825165B - 5G slice network congestion early warning method and device based on time diagram network - Google Patents
5G slice network congestion early warning method and device based on time diagram network Download PDFInfo
- Publication number
- CN113825165B CN113825165B CN202010558931.9A CN202010558931A CN113825165B CN 113825165 B CN113825165 B CN 113825165B CN 202010558931 A CN202010558931 A CN 202010558931A CN 113825165 B CN113825165 B CN 113825165B
- Authority
- CN
- China
- Prior art keywords
- slice
- load
- network
- slice network
- matrix
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W24/00—Supervisory, monitoring or testing arrangements
- H04W24/06—Testing, supervising or monitoring using simulated traffic
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W28/00—Network traffic management; Network resource management
- H04W28/02—Traffic management, e.g. flow control or congestion control
- H04W28/0284—Traffic management, e.g. flow control or congestion control detecting congestion or overload during communication
Abstract
The embodiment of the invention relates to the technical field of communication, and discloses a 5G slice network congestion early warning method and device based on a time diagram network, wherein the method comprises the following steps: obtaining a slice load topological graph of the latest first number of moments from NSMF, and converting the slice load topological graph into an adjacent matrix and a feature matrix; calculating and outputting a load predicted value of a slice network node at a second number of moments in the future according to the adjacency matrix and the feature matrix by applying a pre-trained slice network load predicted model based on a time graph network; and judging whether slice network congestion exists or not according to the load predicted value of the slice network nodes at a second number of moments in the future. By means of the method, the device and the system, congestion of the slicing network can be avoided in time, and early warning accuracy is improved.
Description
Technical Field
The embodiment of the invention relates to the technical field of communication, in particular to a 5G slice network congestion early warning method and device based on a time diagram network.
Background
At present, the congestion early warning of the slice network is mainly realized by a mode of respectively setting threshold values for the loads of nodes of each slice network in the slice network, but the mode is easy to have the problems of frequent false alarms, low accuracy and the like. In addition, network congestion often occurs when an alarm is found in this way, and the avoidance of slice network congestion is not helpful. Meanwhile, the slicing network has the advantages of multiple nodes, complex relationship among the nodes, high automatic detection difficulty and no better early warning means for the congestion of the slicing network at present.
Disclosure of Invention
In view of the above problems, embodiments of the present invention provide a 5G slice network congestion early warning method and apparatus based on a time-diagram network, which overcome or at least partially solve the above problems.
According to an aspect of the embodiment of the present invention, there is provided a 5G slice network congestion early warning method based on a time graph network, the method including: obtaining a slice load topological graph of the latest first number of moments from NSMF, and converting the slice load topological graph into an adjacent matrix and a feature matrix; calculating and outputting a load predicted value of a slice network node at a second number of moments in the future according to the adjacency matrix and the feature matrix by applying a pre-trained slice network load predicted model based on a time graph network; and judging whether slice network congestion exists or not according to the load predicted value of the slice network nodes at a second number of moments in the future.
In an alternative manner, the converting the slice load topology map into an adjacency matrix and a feature matrix includes: representing the connection relation of each slice network node as the adjacent matrix of N according to the slice load topological graph, wherein N is the number of the slice network nodes in the slice load topological graph; and representing the characteristics of the load time sequences of the first number of moments of each slice network node as the characteristic matrix of N.M according to the slice load topological graph, wherein M is the load time sequence length of the slice network node attribute.
In an alternative manner, the calculating the load prediction value of the sliced network node outputting the second number of future moments according to the sliced network load prediction model of the pre-trained time graph based network applied by the adjacency matrix and the feature matrix includes: capturing the spatial characteristics of a slice network load topological graph according to the adjacency matrix and the characteristic matrix by using a graph convolution network in the slice network load prediction model; and extracting time characteristics by using a long-term and short-term memory neural network according to the dynamic change of the spatial characteristics learning slice network load, and outputting the load predicted value of the slice network node at a second number of moments in the future.
In an optional manner, before the calculating the load prediction value of the sliced network node outputting the second number of future moments according to the sliced network load prediction model of the pre-trained time-graph-based network applied by the adjacency matrix and the feature matrix comprises: collecting a historical slice load topological graph from NSMF as a total data set, and converting the slice load topological graph into a historical adjacency matrix and a historical feature matrix; acquiring real node attribute values of a slice network node at a second number of moments in the future to be predicted to form a label matrix; and training the slice network load prediction model based on the time graph network by using the total data set to acquire the weight parameters of the converged slice network load prediction model.
In an optional manner, the training the slice network load prediction model based on the time graph network by using the total data set to obtain the weight parameters of the converged slice network load prediction model includes: training the slice network load prediction model according to the history adjacency matrix and the history feature matrix in the total data set to obtain predicted load prediction values of slice network nodes at a second number of moments in the future; calculating an error between the load predicted value of the slice network node and the real node attribute value, and measuring the error by using an objective function; and gradient descent optimization algorithm is applied to enable the slice network load prediction model to descend in gradient, and the optimal weight parameter of the slice network load prediction model with the minimum objective function is obtained, wherein the optimal weight parameter is the weight parameter of the slice network load prediction model after training.
In an optional manner, the training the slice network load prediction model according to the historical adjacency matrix and the historical feature matrix in the total data set to obtain predicted load prediction values of slice network nodes at a second number of future moments includes: the input layer is applied to receive the input history adjacency matrix and the history feature matrix; obtaining a time sequence corresponding to the spatial characteristics of the slice load topological graph by applying two graph convolution layers; obtaining a feature vector comprising spatial features and temporal features of the slice load topological graph by using two long-short-term memory layers; and outputting the predicted load predicted value of the slice network node at a predicted second number of moments in the future according to the feature vector by an application output layer.
In an optional manner, the determining whether there is a slice network congestion according to the load prediction value of the slice network node at the second number of moments in the future includes: judging whether the load predicted value of the continuous third number of moments of the slicing network node exceeds a preset threshold value or not; if so, determining that congestion exists in the slicing network node, and informing NSMF to limit the slicing users related to the slicing network node.
According to another aspect of the embodiment of the present invention, there is provided a 5G slice network congestion early-warning device based on a time graph network, the device including: the data acquisition module is used for acquiring a slice load topological graph of the latest first number of moments from NSMF and converting the slice load topological graph into an adjacent matrix and a feature matrix; the load prediction module is used for calculating and outputting a load prediction value of a slice network node at a second number of moments in the future according to the slice network load prediction model based on the time graph network and pre-trained by applying the adjacency matrix and the feature matrix; and the congestion judging module is used for judging whether the slice network congestion exists according to the load predicted value of the slice network node at a second number of moments in the future.
According to another aspect of an embodiment of the present invention, there is provided a computing device including: the device comprises a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface complete communication with each other through the communication bus;
the memory is configured to store at least one executable instruction, where the executable instruction causes the processor to execute the steps of the 5G slice network congestion pre-warning method based on the time-diagram network.
According to yet another aspect of the embodiments of the present invention, there is provided a computer storage medium having at least one executable instruction stored therein, the executable instruction causing the processor to perform the steps of the above-described 5G slice network congestion early warning method based on a time graph network.
According to the embodiment of the invention, the slice load topological graph of the latest first number of moments is obtained from NSMF, and the slice load topological graph is converted into an adjacent matrix and a feature matrix; calculating and outputting a load predicted value of a slice network node at a second number of moments in the future according to the adjacency matrix and the feature matrix by applying a pre-trained slice network load predicted model based on a time graph network; judging whether the slice network congestion exists according to the load predicted value of the slice network node at a second number of moments in the future, so that the slice network congestion can be avoided in time, and the early warning accuracy is improved.
The foregoing description is only a summary of technical results of the embodiments of the present invention, and it is to be understood that the following detailed description of the present invention will be given for clarity of understanding of technical means of the embodiments of the present invention, and for clarity of understanding of the above and other objects, features and advantages of the embodiments of the present invention.
Drawings
Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The drawings are only for purposes of illustrating the preferred embodiments and are not to be construed as limiting the invention. Also, like reference numerals are used to designate like parts throughout the figures. In the drawings:
fig. 1 is a schematic diagram of a method for 5G slice network congestion early warning based on a time diagram network according to an embodiment of the present invention;
fig. 2 shows a flow diagram of a 5G slice network congestion early warning method based on a time diagram network according to an embodiment of the present invention;
fig. 3 is a schematic structural diagram of a slice network load prediction model of a 5G slice network congestion early warning method based on a time diagram network according to an embodiment of the present invention;
Fig. 4 shows a schematic diagram of LSTM neurons in a slice network load prediction model of a 5G slice network congestion early warning method based on a time graph network according to an embodiment of the present invention;
fig. 5 shows a schematic structural diagram of a 5G slice network congestion early warning device based on a time diagram network according to an embodiment of the present invention;
FIG. 6 illustrates a schematic diagram of a computing device provided by an embodiment of the present invention.
Detailed Description
Exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the present invention are shown in the drawings, it should be understood that the present invention may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.
Network slicing (Network Slice) is an end-to-end logical function and a set of physical or virtual resources required, including access networks, transport networks, core networks, etc., which can be considered as a virtualized "private Network" in a 5G Network; network slicing is based on unified infrastructure construction of network function virtualization (Network Function Virtualization, NFV), and low-cost and efficient operation is achieved. Network slicing techniques may implement logical isolation of a communication network, allowing network elements and functions to be configured and reused in each network slice to meet specific industry application requirements. The slice management architecture is mainly composed of a communication traffic management function (Communication Service Management Function, CSMF), a network slice management function (Network Slice Management Function, NSMF), and a network slice subnet management function (Network Slice Subnet Management Function, NSSMF).
The CSMF completes the order and process of the user business communication service, is responsible for converting the communication service requirement of the operator/third party client into the requirement of the network slice, and sends the requirement of the network slice (such as creating, terminating, modifying the network slice instance request, etc.) to the NSMF through an interface between the CSMF and the NSMF, and obtains the management data (such as performance, fault data, etc.) of the network slice from the NSMF. The NSMF is responsible for receiving network slice requirements sent by the CSMF, managing the life cycle, performance, faults and the like of the network slice examples, arranging the composition of the network slice examples, decomposing the requirements of the network slice examples into requirements of network slice subnet examples or network functions, and sending a network slice subnet example management request to each NSSMF. The NSSMF receives the network slicing subnet deployment requirement issued from the NSMF, manages the network slicing subnet instances, orchestrates the composition of the network slicing subnet instances, maps the SLA requirement of the network slicing subnet to the QoS requirement of the network service, and issues a deployment request of the network service to the NFV orchestrator (NFV Orchestration, NFVO) system of the european telecommunication standardization institute (European Telecommunications Standards Institute, ETSI) NFV domain.
The slicing network nodes are numerous and complex in relation, the slicing network is not a two-dimensional grid, but is in the form of a graph, which means that the convolutional neural network (Convolutional Neural Network, CNN) model cannot reflect the complex topology of the slicing network and thus cannot accurately capture the spatial dependency. Embodiments of the present invention therefore propose to take advantage of the time-diagram convolutional neural network (Temporal Graph Convolutional Network, T-GCN) in capturing both spatial and time-dependent relationships. The time graph convolutional neural network combines a Graph Convolutional Network (GCN) and a long-short-term memory neural network (LSTM). GCN is used for learning complex topological structure to capture space dependence, LSTM is used for learning dynamic change of slice network load to capture time dependence.
In the embodiment of the invention, as shown in fig. 1, a 5G slice network congestion early warning method based on a time map network acquires slice load topological graphs of the latest M times from a network slice management function NSMF at each time; converting the slice load topological graph into an adjacent matrix A and a characteristic matrix X, wherein the adjacent matrix A is the connection relation of each node of the slice network, and the characteristic matrix X is the characteristic representation of the load time sequence of each slice network node M moments in the slice network; carrying out data normalization processing on the adjacent matrix A and the feature matrix X; inputting the preprocessed adjacency matrix A and the feature matrix X into a slice network load prediction model based on a time diagram network; outputting the load predicted value of each slice network node at the future time T after the pre-trained slice network load predicted model is calculated; whether the load predicted value of the continuous K moments exists in the slice network node exceeds a preset threshold value is judged, if so, NSMF is informed of limiting the slice users related to the slice network node. Therefore, congestion of the slice network can be avoided in time, and early warning accuracy is improved.
Fig. 2 shows a flow chart of a 5G slice network congestion early warning method based on a time chart network according to an embodiment of the present invention. The 5G slice network congestion early warning method based on the time diagram network is applied to a server side, and as shown in fig. 2, the 5G slice network congestion early warning method based on the time diagram network comprises the following steps:
step S11: and obtaining a slice load topological graph of the latest first number of moments from NSMF, and converting the slice load topological graph into an adjacency matrix and a feature matrix.
At each moment, a slice load topology map of the latest first number M of moments is acquired from NSMF. Wherein, the slice network load topological graph can be expressed as g= (V, E), V is a set v= { V of slice network nodes 1 ，V 2 ，V 3 ，…，V N N is the number of slice network nodes in the slice network load topology, and E is the set of edges. If slicing network node V i And slice network node V j E is connected with ij =1, otherwise e ij =0. Converting a slice network load topological graph into an adjacent matrix A and a feature matrix X, wherein the adjacent matrix A with the connection relation of each slice network node expressed as N X N according to the slice load topological graph, and N is the number of the slice network nodes in the slice load topological graph; and representing the characteristic of the load time sequence of the first number of moments of each slice network node as N.M according to the slice load topological graph, wherein M is the load time sequence length of the slice network node attribute.
Wherein e ij Representing a slice network node V i And slice network node V j The connection relation between the slice network nodes is 1, otherwise, 0, x it Representing the properties of the i-th slice network node at time t. In embodiments of the present invention, the sliced network node attribute, i.e., network load, may be set up for the number of calls per second (Call Attempts Per Second, CAPS), i.e., network Concurrency (CAPS), or transaction per second (Transaction Per Second, TPS). M represents the time sequence length of the attribute of the slicing network node, and the time sequence length of the feature matrix X is equal to the first number M because the attribute number of the slicing network node is 1 at a certain moment.
The embodiment of the invention also carries out standardized pretreatment on the data, and carries out the following steps for each dimension: (X-mean)/std, i.e., the data is attributed (column-wise) minus its mean and divided by its variance. After the standardized pretreatment, the convergence speed of the subsequent slice network load prediction model can be improved, and the precision of the model is improved.
Step S12: and calculating and outputting the load predicted value of the slice network node at a second number of moments in the future according to the adjacency matrix and the feature matrix by applying a pre-trained slice network load predicted model based on the time graph network.
The slice network load prediction model of the embodiment of the invention is a time graph convolutional neural network (T-GCN) constructed by a Graph Convolution Network (GCN) and a long and short term memory neural network (LSTM). Utilizing a time graph convolutional neural network to simultaneously extract advantages of space and time dependence, and specifically utilizing a graph rolling network GCN in the slice network load prediction model to capture space characteristics of a slice network load topological graph according to the adjacency matrix and the characteristic matrix; inputting the obtained time sequence with the spatial characteristics into an LSTM, applying a long-short-term memory neural network LSTM to learn the dynamic change of the slice network load according to the spatial characteristics to extract the temporal characteristics, and outputting the load predicted value of the slice network node at a second number T of time points in the future. Each graph neural network layer can be written as a nonlinear function:
H (l+1) ＝f(H (l) ,A),
wherein H is (0) =x is input data, H (l) The =z is the output data, L is the number of layers of the neural network, and selecting different f () and parameters also determines different slice network load prediction models. W (W) (l) Is the parameter matrix of the first neural network layer, a non-linear activation function such as ReLU, a is the adjacency matrix, and D is the node degree diagonal matrix of a.
In the embodiment of the present invention, before step S12, the pre-training of the slice network load prediction model is completed, and the weight parameters for convergence of the slice network load prediction model are obtained. Specifically, a historical slice load topological graph is collected from NSMF as a total data set, and the slice load topological graph is converted into a historical adjacency matrix A and a historical feature matrix X; acquiring real node attribute values of a slice network node at a second number of moments in the future to be predicted to form a label matrix Y; and training the slice network load prediction model based on the time graph network by using the total data set to acquire the weight parameters of the converged slice network load prediction model. The shape of the adjacent matrix A is N, the shape of the matrix of the characteristic matrix X is N.M, the shape of the label matrix Y is N.T, wherein the attribute of the ith slicing network node at the time T is expressed as Y it . The preprocessing is performed on the data before the total data set is used to train the slice network load prediction model, and the specific preprocessing method is the same as the foregoing, and will not be described herein. In the embodiment of the invention, 80% of the total data set is divided into a training set, 20% of the total data set is divided into a test set, the training set is used for training a model, and the test set is used for testing the performance of the model.
When the slice network load prediction model is trained, the total data set is usedAnd training the slice network load prediction model by the historical adjacency matrix and the historical feature matrix to obtain predicted load prediction values of the slice network nodes at a second number of moments in the future. The structure of the sliced network load prediction model referring to fig. 3, specifically, the history adjacency matrix a and the history feature matrix X, which are input, are received by the application input layer. X in FIG. 3 t The characteristic attribute of the sliced network node at the time t is input of a history characteristic matrix X, and at the same time, a history adjacency matrix a (not shown) is also input. And (5) obtaining a time sequence corresponding to the spatial characteristics of the slice load topological graph by using two graph roll stacking layers (GCNs). The time series is input to a subsequent long-short-term memory Layer (LSTM), and the two long-term memory Layers (LSTM) are applied to obtain feature vectors comprising spatial features and temporal features of the slice load topology. And outputting the predicted load predicted value of the slice network node at a predicted second number of moments in the future according to the feature vector by an application output layer. Wherein the number of convolution kernels of the two graph convolution layers is 32 (i.e., the dimension of the output), the activation function is set to "relu". The number of neurons of the two long short term memory Layers (LSTM) is set to 64 and the activation function is set to "relu". The long-short-term memory Layer (LSTM) is a special type of cyclic neural network, and long-term information can be remembered by controlling the time for which the values in the cache are stored, so that the method is suitable for forecasting time sequences. As shown in fig. 4, there are four inputs and one output per neuron, and one Cell stores the memorized values in each neuron. Three gates are contained in each LSTM neuron: forgetting door, input door, output door, forgetting door f t Control of the internal state c at the last instant t-1 How much information needs to be forgotten; input gate i t Controlling candidate states at a current timeHow much information needs to be saved; output door o t Control of internal state c at the present instant t How much information needs to be output to the external state h t . Specifically, the method is obtained by calculation according to the following formulas:
forgetting the door:
an input door:
obtaining a state updating function through a nonlinear function:
internal state:
output door:
outputting an external state:
outputting an external state: y is Y t ＝σ(W′h t )。
Wherein x is t For feature vector input at time t, f t Indicating forgetful door, i t Represents an input gate, o t Representing an output gate, c t Indicating the neuron state at time t, h t Representing the external state of the neuron, i.e., the hidden state of the LSTM layer, W is a trainable weight matrix and b is a bias vector. For example W i Is the weight matrix of the input gate, W f Weight matrix representing forgetting gate, W o Is the weight matrix of the output gate, b i Is the bias term of the input gate, b f Is the bias item of the forgetting door, b o Is the bias term of the output gate, the gate activation function is sigmoid (sigma), the value range is (0, 1), and the output activation function is the tanh function. Equation (1) represents a forgetting gate, new information is added in equations (2) and (3), equation (4) fuses new information and old information, equations (5) and (6) output that the current LSTM neuron has learned Is provided for the next time stamp. Each connection line in the LSTM neuron contains a corresponding weight. The long-term and short-term memory neural network has a good effect on long-sequence learning.
The output layer is constituted by a fully connected layer (Dense). And setting the number of neurons included in the output layer as a second number T, and setting an activation function as sigmoid, namely outputting the predicted historical load predicted value of each slice network node of the second number T of predicted future moments.
After obtaining a predicted historical load predicted value, calculating an error between the load predicted value of the slicing network node and the real node attribute value, and measuring the error by using an objective function. The mean square error MSE (Mean Squared Error) is chosen as the loss function or objective function (loss = 'mean square error'),the training objective is to minimize this error. Setting the training round number to 1000 (epochs=1000), batch size to 32 (batch_size=32), gradient descent optimization algorithm selects adam optimizer for improving learning speed of conventional gradient descent (optimizer= 'adam'). The neural network can find the optimal weight parameter which minimizes the objective function through gradient descent, and the neural network can autonomously learn the weight parameter through training. Training is carried out by using a training set, so that the smaller the objective function is, the better the objective function is, and after each round of training, a test set is used for evaluating and verifying a network element function cutting slice network load prediction model. And gradient descent optimization algorithm is applied to enable the slice network load prediction model to descend in gradient, and the optimal weight parameter of the slice network load prediction model with the minimum objective function is obtained, wherein the optimal weight parameter is the weight parameter of the slice network load prediction model after training.
In step S12, specifically, the adjacency matrix a and the feature matrix X are input into a pre-trained slice network load prediction model based on the time graph network, the weight parameter of the slice network load prediction model is the optimal weight parameter obtained by training, and the load prediction values of the slice network nodes at the second number T of moments in the future are calculated and output through the slice network load prediction model.
Step S13: and judging whether slice network congestion exists or not according to the load predicted value of the slice network nodes at a second number of moments in the future.
Specifically, judging whether the load predicted value of the continuous third number K of times of the slice network nodes exceeds a preset threshold value or not; if yes, determining that congestion exists in the slicing network node, and informing NSMF to limit the current of slicing users related to the slicing network node, so that the slicing network congestion can be avoided in time.
According to the embodiment of the invention, the load prediction values of the slice network nodes at T times in the future are predicted at each time by the constructed slice network load prediction model based on the time graph network, so that whether the corresponding slice network node is likely to have congestion is judged according to the load prediction values, and when congestion is likely to exist, NSMF is timely informed of limiting the current of related slice users, so that the slice network congestion can be timely avoided, and the automation of 5G slice network congestion early warning is improved. And when congestion judgment is carried out, congestion possibly exists in a certain slice network node is determined only when the load predicted values of K continuous moments of the slice network node exceed a preset threshold, so that false early warning caused by single abnormal detection data can be eliminated, early warning accuracy can be improved, and too frequent early warning can be prevented.
According to the embodiment of the invention, the slice load topological graph of the latest first number of moments is obtained from NSMF, and the slice load topological graph is converted into an adjacent matrix and a feature matrix; calculating and outputting a load predicted value of a slice network node at a second number of moments in the future according to the adjacency matrix and the feature matrix by applying a pre-trained slice network load predicted model based on a time graph network; judging whether the slice network congestion exists according to the load predicted value of the slice network node at a second number of moments in the future, so that the slice network congestion can be avoided in time, and the early warning accuracy is improved.
Fig. 5 shows a schematic structural diagram of a 5G slice network congestion warning device based on a time diagram network according to an embodiment of the present invention. As shown in fig. 5, the 5G slice network congestion early-warning device based on the time-diagram network includes: a data acquisition module 501, a load prediction module 502, a congestion judgment module 503, and a model training unit 504. Wherein:
the data acquisition module 501 is configured to acquire a slice load topology map of a first number of recent moments from an NSMF, and convert the slice load topology map into an adjacency matrix and a feature matrix; the load prediction module 502 is configured to apply a pre-trained slice network load prediction model based on a time graph network according to the adjacency matrix and the feature matrix to calculate and output a load prediction value of a slice network node at a second number of moments in the future; the congestion judging module 503 is configured to judge whether there is slice network congestion according to the load prediction value of the slice network node at a second number of moments in the future.
In an alternative manner, the data acquisition module 501 is configured to: representing the connection relation of each slice network node as the adjacent matrix of N according to the slice load topological graph, wherein N is the number of the slice network nodes in the slice load topological graph; and representing the characteristics of the load time sequences of the first number of moments of each slice network node as the characteristic matrix of N.M according to the slice load topological graph, wherein M is the load time sequence length of the slice network node attribute.
In an alternative approach, the load prediction module 502 is configured to: capturing the spatial characteristics of a slice network load topological graph according to the adjacency matrix and the characteristic matrix by using a graph convolution network in the slice network load prediction model; and extracting time characteristics by using a long-term and short-term memory neural network according to the dynamic change of the spatial characteristics learning slice network load, and outputting the load predicted value of the slice network node at a second number of moments in the future.
In an alternative way, the model training unit 504 is used to: collecting a historical slice load topological graph from NSMF as a total data set, and converting the slice load topological graph into a historical adjacency matrix and a historical feature matrix; acquiring real node attribute values of a slice network node at a second number of moments in the future to be predicted to form a label matrix; and training the slice network load prediction model based on the time graph network by using the total data set to acquire the weight parameters of the converged slice network load prediction model.
In an alternative way, the model training unit 504 is used to: training the slice network load prediction model according to the history adjacency matrix and the history feature matrix in the total data set to obtain predicted load prediction values of slice network nodes at a second number of moments in the future; calculating an error between the load predicted value of the slice network node and the real node attribute value, and measuring the error by using an objective function; and gradient descent optimization algorithm is applied to enable the slice network load prediction model to descend in gradient, and the optimal weight parameter of the slice network load prediction model with the minimum objective function is obtained, wherein the optimal weight parameter is the weight parameter of the slice network load prediction model after training.
In an alternative way, the model training unit 504 is further configured to: the input layer is applied to receive the input history adjacency matrix and the history feature matrix; obtaining a time sequence corresponding to the spatial characteristics of the slice load topological graph by applying two graph convolution layers; obtaining a feature vector comprising spatial features and temporal features of the slice load topological graph by using two long-short-term memory layers; and outputting the predicted load predicted value of the slice network node at a predicted second number of moments in the future according to the feature vector by an application output layer.
In an alternative manner, the congestion determination module 503 is configured to: judging whether the load predicted value of the continuous third number of moments of the slicing network node exceeds a preset threshold value or not; if so, determining that congestion exists in the slicing network node, and informing NSMF to limit the slicing users related to the slicing network node.
According to the embodiment of the invention, the slice load topological graph of the latest first number of moments is obtained from NSMF, and the slice load topological graph is converted into an adjacent matrix and a feature matrix; calculating and outputting a load predicted value of a slice network node at a second number of moments in the future according to the adjacency matrix and the feature matrix by applying a pre-trained slice network load predicted model based on a time graph network; judging whether the slice network congestion exists according to the load predicted value of the slice network node at a second number of moments in the future, so that the slice network congestion can be avoided in time, and the early warning accuracy is improved.
The embodiment of the invention provides a non-volatile computer storage medium, which stores at least one executable instruction, and the computer executable instruction can execute the 5G slice network congestion early warning method based on the time chart network in any method embodiment.
The executable instructions may be particularly useful for causing a processor to:
obtaining a slice load topological graph of the latest first number of moments from NSMF, and converting the slice load topological graph into an adjacent matrix and a feature matrix;
calculating and outputting a load predicted value of a slice network node at a second number of moments in the future according to the adjacency matrix and the feature matrix by applying a pre-trained slice network load predicted model based on a time graph network;
and judging whether slice network congestion exists or not according to the load predicted value of the slice network nodes at a second number of moments in the future.
In one alternative, the executable instructions cause the processor to:
representing the connection relation of each slice network node as the adjacent matrix of N according to the slice load topological graph, wherein N is the number of the slice network nodes in the slice load topological graph;
and representing the characteristics of the load time sequences of the first number of moments of each slice network node as the characteristic matrix of N.M according to the slice load topological graph, wherein M is the load time sequence length of the slice network node attribute.
In one alternative, the executable instructions cause the processor to:
Capturing the spatial characteristics of a slice network load topological graph according to the adjacency matrix and the characteristic matrix by using a graph convolution network in the slice network load prediction model;
and extracting time characteristics by using a long-term and short-term memory neural network according to the dynamic change of the spatial characteristics learning slice network load, and outputting the load predicted value of the slice network node at a second number of moments in the future.
In one alternative, the executable instructions cause the processor to:
collecting a historical slice load topological graph from NSMF as a total data set, and converting the slice load topological graph into a historical adjacency matrix and a historical feature matrix;
acquiring real node attribute values of a slice network node at a second number of moments in the future to be predicted to form a label matrix;
and training the slice network load prediction model based on the time graph network by using the total data set to acquire the weight parameters of the converged slice network load prediction model.
In one alternative, the executable instructions cause the processor to:
training the slice network load prediction model according to the history adjacency matrix and the history feature matrix in the total data set to obtain predicted load prediction values of slice network nodes at a second number of moments in the future;
Calculating an error between the load predicted value of the slice network node and the real node attribute value, and measuring the error by using an objective function;
and gradient descent optimization algorithm is applied to enable the slice network load prediction model to descend in gradient, and the optimal weight parameter of the slice network load prediction model with the minimum objective function is obtained, wherein the optimal weight parameter is the weight parameter of the slice network load prediction model after training.
In one alternative, the executable instructions cause the processor to:
the input layer is applied to receive the input history adjacency matrix and the history feature matrix;
obtaining a time sequence corresponding to the spatial characteristics of the slice load topological graph by applying two graph convolution layers;
obtaining a feature vector comprising spatial features and temporal features of the slice load topological graph by using two long-short-term memory layers;
and outputting the predicted load predicted value of the slice network node at a predicted second number of moments in the future according to the feature vector by an application output layer.
In one alternative, the executable instructions cause the processor to:
judging whether the load predicted value of the continuous third number of moments of the slicing network node exceeds a preset threshold value or not;
If so, determining that congestion exists in the slicing network node, and informing NSMF to limit the slicing users related to the slicing network node.
According to the embodiment of the invention, the slice load topological graph of the latest first number of moments is obtained from NSMF, and the slice load topological graph is converted into an adjacent matrix and a feature matrix; calculating and outputting a load predicted value of a slice network node at a second number of moments in the future according to the adjacency matrix and the feature matrix by applying a pre-trained slice network load predicted model based on a time graph network; judging whether the slice network congestion exists according to the load predicted value of the slice network node at a second number of moments in the future, so that the slice network congestion can be avoided in time, and the early warning accuracy is improved.
An embodiment of the present invention provides a computer program product, including a computer program stored on a computer storage medium, where the computer program includes program instructions, when executed by a computer, cause the computer to perform the 5G slice network congestion early warning method based on a time graph network in any of the above method embodiments.
The executable instructions may be particularly useful for causing a processor to:
obtaining a slice load topological graph of the latest first number of moments from NSMF, and converting the slice load topological graph into an adjacent matrix and a feature matrix;
calculating and outputting a load predicted value of a slice network node at a second number of moments in the future according to the adjacency matrix and the feature matrix by applying a pre-trained slice network load predicted model based on a time graph network;
and judging whether slice network congestion exists or not according to the load predicted value of the slice network nodes at a second number of moments in the future.
In one alternative, the executable instructions cause the processor to:
representing the connection relation of each slice network node as the adjacent matrix of N according to the slice load topological graph, wherein N is the number of the slice network nodes in the slice load topological graph;
and representing the characteristics of the load time sequences of the first number of moments of each slice network node as the characteristic matrix of N.M according to the slice load topological graph, wherein M is the load time sequence length of the slice network node attribute.
In one alternative, the executable instructions cause the processor to:
Capturing the spatial characteristics of a slice network load topological graph according to the adjacency matrix and the characteristic matrix by using a graph convolution network in the slice network load prediction model;
and extracting time characteristics by using a long-term and short-term memory neural network according to the dynamic change of the spatial characteristics learning slice network load, and outputting the load predicted value of the slice network node at a second number of moments in the future.
In one alternative, the executable instructions cause the processor to:
collecting a historical slice load topological graph from NSMF as a total data set, and converting the slice load topological graph into a historical adjacency matrix and a historical feature matrix;
acquiring real node attribute values of a slice network node at a second number of moments in the future to be predicted to form a label matrix;
and training the slice network load prediction model based on the time graph network by using the total data set to acquire the weight parameters of the converged slice network load prediction model.
In one alternative, the executable instructions cause the processor to:
training the slice network load prediction model according to the history adjacency matrix and the history feature matrix in the total data set to obtain predicted load prediction values of slice network nodes at a second number of moments in the future;
Calculating an error between the load predicted value of the slice network node and the real node attribute value, and measuring the error by using an objective function;
and gradient descent optimization algorithm is applied to enable the slice network load prediction model to descend in gradient, and the optimal weight parameter of the slice network load prediction model with the minimum objective function is obtained, wherein the optimal weight parameter is the weight parameter of the slice network load prediction model after training.
In one alternative, the executable instructions cause the processor to:
the input layer is applied to receive the input history adjacency matrix and the history feature matrix;
obtaining a time sequence corresponding to the spatial characteristics of the slice load topological graph by applying two graph convolution layers;
obtaining a feature vector comprising spatial features and temporal features of the slice load topological graph by using two long-short-term memory layers;
and outputting the predicted load predicted value of the slice network node at a predicted second number of moments in the future according to the feature vector by an application output layer.
In one alternative, the executable instructions cause the processor to:
judging whether the load predicted value of the continuous third number of moments of the slicing network node exceeds a preset threshold value or not;
If so, determining that congestion exists in the slicing network node, and informing NSMF to limit the slicing users related to the slicing network node.
According to the embodiment of the invention, the slice load topological graph of the latest first number of moments is obtained from NSMF, and the slice load topological graph is converted into an adjacent matrix and a feature matrix; calculating and outputting a load predicted value of a slice network node at a second number of moments in the future according to the adjacency matrix and the feature matrix by applying a pre-trained slice network load predicted model based on a time graph network; judging whether the slice network congestion exists according to the load predicted value of the slice network node at a second number of moments in the future, so that the slice network congestion can be avoided in time, and the early warning accuracy is improved.
FIG. 6 is a schematic diagram of a computing device according to an embodiment of the present invention, and the embodiment of the present invention is not limited to the specific implementation of the device.
As shown in fig. 6, the computing device may include: a processor 602, a communication interface (Communications Interface), a memory 606, and a communication bus 608.
Wherein: processor 602, communication interface 604, and memory 606 perform communication with each other via communication bus 608. Communication interface 604 is used to communicate with network elements of other devices, such as clients or other servers. The processor 602 is configured to execute the program 610, and may specifically perform relevant steps in the embodiment of the 5G slice network congestion early warning method based on the time map network.
In particular, program 610 may include program code including computer-operating instructions.
The processor 602 may be a central processing unit CPU, or a specific integrated circuit ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement embodiments of the present invention. The device includes one or each processor, which may be the same type of processor, such as one or each CPU; but may also be different types of processors such as one or each CPU and one or each ASIC.
A memory 606 for storing a program 610. The memory 606 may comprise high-speed RAM memory or may further comprise non-volatile memory (non-volatile memory), such as at least one disk memory.
The program 610 may be specifically operable to cause the processor 602 to:
obtaining a slice load topological graph of the latest first number of moments from NSMF, and converting the slice load topological graph into an adjacent matrix and a feature matrix;
calculating and outputting a load predicted value of a slice network node at a second number of moments in the future according to the adjacency matrix and the feature matrix by applying a pre-trained slice network load predicted model based on a time graph network;
And judging whether slice network congestion exists or not according to the load predicted value of the slice network nodes at a second number of moments in the future.
In an alternative, the program 610 causes the processor to:
representing the connection relation of each slice network node as the adjacent matrix of N according to the slice load topological graph, wherein N is the number of the slice network nodes in the slice load topological graph;
and representing the characteristics of the load time sequences of the first number of moments of each slice network node as the characteristic matrix of N.M according to the slice load topological graph, wherein M is the load time sequence length of the slice network node attribute.
In an alternative, the program 610 causes the processor to:
capturing the spatial characteristics of a slice network load topological graph according to the adjacency matrix and the characteristic matrix by using a graph convolution network in the slice network load prediction model;
and extracting time characteristics by using a long-term and short-term memory neural network according to the dynamic change of the spatial characteristics learning slice network load, and outputting the load predicted value of the slice network node at a second number of moments in the future.
In an alternative, the program 610 causes the processor to:
collecting a historical slice load topological graph from NSMF as a total data set, and converting the slice load topological graph into a historical adjacency matrix and a historical feature matrix;
acquiring real node attribute values of a slice network node at a second number of moments in the future to be predicted to form a label matrix;
and training the slice network load prediction model based on the time graph network by using the total data set to acquire the weight parameters of the converged slice network load prediction model.
In an alternative, the program 610 causes the processor to:
training the slice network load prediction model according to the history adjacency matrix and the history feature matrix in the total data set to obtain predicted load prediction values of slice network nodes at a second number of moments in the future;
calculating an error between the load predicted value of the slice network node and the real node attribute value, and measuring the error by using an objective function;
and gradient descent optimization algorithm is applied to enable the slice network load prediction model to descend in gradient, and the optimal weight parameter of the slice network load prediction model with the minimum objective function is obtained, wherein the optimal weight parameter is the weight parameter of the slice network load prediction model after training.
In an alternative, the program 610 causes the processor to:
the input layer is applied to receive the input history adjacency matrix and the history feature matrix;
obtaining a time sequence corresponding to the spatial characteristics of the slice load topological graph by applying two graph convolution layers;
obtaining a feature vector comprising spatial features and temporal features of the slice load topological graph by using two long-short-term memory layers;
and outputting the predicted load predicted value of the slice network node at a predicted second number of moments in the future according to the feature vector by an application output layer.
In an alternative, the program 610 causes the processor to:
judging whether the load predicted value of the continuous third number of moments of the slicing network node exceeds a preset threshold value or not;
if so, determining that congestion exists in the slicing network node, and informing NSMF to limit the slicing users related to the slicing network node.
According to the embodiment of the invention, the slice load topological graph of the latest first number of moments is obtained from NSMF, and the slice load topological graph is converted into an adjacent matrix and a feature matrix; calculating and outputting a load predicted value of a slice network node at a second number of moments in the future according to the adjacency matrix and the feature matrix by applying a pre-trained slice network load predicted model based on a time graph network; judging whether the slice network congestion exists according to the load predicted value of the slice network node at a second number of moments in the future, so that the slice network congestion can be avoided in time, and the early warning accuracy is improved.
The algorithms or displays presented herein are not inherently related to any particular computer, virtual system, or other apparatus. Various general-purpose systems may also be used with the teachings herein. The required structure for a construction of such a system is apparent from the description above. In addition, embodiments of the present invention are not directed to any particular programming language. It will be appreciated that the teachings of the present invention described herein may be implemented in a variety of programming languages, and the above description of specific languages is provided for disclosure of enablement and best mode of the present invention.
In the description provided herein, numerous specific details are set forth. However, it is understood that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures and techniques have not been shown in detail in order not to obscure an understanding of this description.
Similarly, it should be appreciated that in the above description of exemplary embodiments of the invention, various features of the embodiments of the invention are sometimes grouped together in a single embodiment, figure, or description thereof for the purpose of streamlining the disclosure and aiding in the understanding of one or more of the various inventive aspects. However, the disclosed method should not be construed as reflecting the intention that: i.e., the claimed invention requires more features than are expressly recited in each claim. Rather, as the following claims reflect, inventive aspects lie in less than all features of a single foregoing disclosed embodiment. Thus, the claims following the detailed description are hereby expressly incorporated into this detailed description, with each claim standing on its own as a separate embodiment of this invention.
Those skilled in the art will appreciate that the modules in the apparatus of the embodiments may be adaptively changed and disposed in one or more apparatuses different from the embodiments. The modules or units or components of the embodiments may be combined into one module or unit or component and, furthermore, they may be divided into a plurality of sub-modules or sub-units or sub-components. Any combination of all features disclosed in this specification (including any accompanying claims, abstract and drawings), and all of the processes or units of any method or apparatus so disclosed, may be used in combination, except insofar as at least some of such features and/or processes or units are mutually exclusive. Each feature disclosed in this specification (including any accompanying claims, abstract and drawings), may be replaced by alternative features serving the same, equivalent or similar purpose, unless expressly stated otherwise.
Furthermore, those skilled in the art will appreciate that while some embodiments herein include some features but not others included in other embodiments, combinations of features of different embodiments are meant to be within the scope of the invention and form different embodiments. For example, in the following claims, any of the claimed embodiments can be used in any combination.
It should be noted that the above-mentioned embodiments illustrate rather than limit the invention, and that those skilled in the art will be able to design alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word "comprising" does not exclude the presence of elements or steps not listed in a claim. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention may be implemented by means of hardware comprising several distinct elements, and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by one and the same item of hardware. The use of the words first, second, third, etc. do not denote any order. These words may be interpreted as names. The steps in the above embodiments should not be construed as limiting the order of execution unless specifically stated.
Claims (6)
1. A 5G slice network congestion pre-warning method based on a time diagram network, the method comprising:
Obtaining a slice load topological graph of the latest first number of moments from NSMF, and converting the slice load topological graph into an adjacent matrix and a feature matrix;
collecting a historical slice load topological graph from NSMF as a total data set, and converting the slice load topological graph into a historical adjacency matrix and a historical feature matrix; acquiring real node attribute values of a slice network node at a second number of moments in the future to be predicted to form a label matrix; training the slice network load prediction model based on the time graph network by applying the total data set to acquire the weight parameters of the converged slice network load prediction model;
the training of the slice network load prediction model based on the time graph network by using the total data set, obtaining the weight parameters of the converged slice network load prediction model, includes: training the slice network load prediction model according to the history adjacency matrix and the history feature matrix in the total data set to obtain predicted load prediction values of slice network nodes at a second number of moments in the future; calculating an error between the load predicted value of the slice network node and the real node attribute value, and measuring the error by using an objective function; gradient descent optimization algorithm is applied to enable the slice network load prediction model to descend in gradient, and the optimal weight parameter of the slice network load prediction model with the minimum objective function is obtained, wherein the optimal weight parameter is the weight parameter of the slice network load prediction model after training;
Training the slice network load prediction model according to the history adjacency matrix and the history feature matrix in the total data set, and obtaining predicted load prediction values of slice network nodes at a second number of moments in the future, including: the input layer is applied to receive the input history adjacency matrix and the history feature matrix; obtaining a time sequence corresponding to the spatial characteristics of the slice load topological graph by applying two graph convolution layers; obtaining a feature vector comprising spatial features and temporal features of the slice load topological graph by using two long-short-term memory layers; the application output layer outputs the predicted load predicted value of the slice network node at a predicted second number of moments in the future according to the feature vector;
calculating and outputting the load predicted value of the slice network node at a second number of moments in the future according to the adjacency matrix and the feature matrix by applying a pre-trained slice network load predicted model based on a time graph network, wherein the method comprises the following steps: capturing the spatial characteristics of a slice network load topological graph according to the adjacency matrix and the characteristic matrix by using a graph convolution network in the slice network load prediction model; the long-term and short-term memory neural network is applied to extract time characteristics according to the dynamic change of the load of the learning slice network of the space characteristics, and the load predicted value of the slice network node at a second number of moments in the future is output;
And judging whether slice network congestion exists or not according to the load predicted value of the slice network nodes at a second number of moments in the future.
2. The method of claim 1, wherein said translating the slice load topology map into an adjacency matrix and a feature matrix comprises:
representing the connection relation of each slice network node as the adjacent matrix of N according to the slice load topological graph, wherein N is the number of the slice network nodes in the slice load topological graph;
and representing the characteristics of the load time sequences of the first number of moments of each slice network node as the characteristic matrix of N.M according to the slice load topological graph, wherein M is the load time sequence length of the slice network node attribute.
3. The method of claim 1, wherein said determining whether there is a slice network congestion based on the load predictors of the slice network nodes at a second number of future times comprises:
judging whether the load predicted value of the continuous third number of moments of the slicing network node exceeds a preset threshold value or not;
if so, determining that congestion exists in the slicing network node, and informing NSMF to limit the slicing users related to the slicing network node.
4. A 5G slice network congestion early warning device based on a time diagram network, the device comprising:
the data acquisition module is used for acquiring a slice load topological graph of the latest first number of moments from NSMF and converting the slice load topological graph into an adjacent matrix and a feature matrix;
the model training unit is used for collecting a historical slice load topological graph from NSMF as a total data set and converting the slice load topological graph into a historical adjacency matrix and a historical feature matrix; acquiring real node attribute values of a slice network node at a second number of moments in the future to be predicted to form a label matrix; training the slice network load prediction model based on the time graph network by applying the total data set to acquire the weight parameters of the converged slice network load prediction model;
the training of the slice network load prediction model based on the time graph network by using the total data set, obtaining the weight parameters of the converged slice network load prediction model, includes: training the slice network load prediction model according to the history adjacency matrix and the history feature matrix in the total data set to obtain predicted load prediction values of slice network nodes at a second number of moments in the future; calculating an error between the load predicted value of the slice network node and the real node attribute value, and measuring the error by using an objective function; gradient descent optimization algorithm is applied to enable the slice network load prediction model to descend in gradient, and the optimal weight parameter of the slice network load prediction model with the minimum objective function is obtained, wherein the optimal weight parameter is the weight parameter of the slice network load prediction model after training;
Training the slice network load prediction model according to the history adjacency matrix and the history feature matrix in the total data set, and obtaining predicted load prediction values of slice network nodes at a second number of moments in the future, including: the input layer is applied to receive the input history adjacency matrix and the history feature matrix; obtaining a time sequence corresponding to the spatial characteristics of the slice load topological graph by applying two graph convolution layers; obtaining a feature vector comprising spatial features and temporal features of the slice load topological graph by using two long-short-term memory layers; the application output layer outputs the predicted load predicted value of the slice network node at a predicted second number of moments in the future according to the feature vector;
the load prediction module is used for calculating and outputting the load predicted value of the slice network node at a second number of time points in the future according to the slice network load predicted model of the pre-trained time-map-based network applied by the adjacency matrix and the feature matrix, and comprises the following steps: capturing the spatial characteristics of a slice network load topological graph according to the adjacency matrix and the characteristic matrix by using a graph convolution network in the slice network load prediction model; the long-term and short-term memory neural network is applied to extract time characteristics according to the dynamic change of the load of the learning slice network of the space characteristics, and the load predicted value of the slice network node at a second number of moments in the future is output;
And the congestion judging module is used for judging whether the slice network congestion exists according to the load predicted value of the slice network node at a second number of moments in the future.
5. A computing device, comprising: the device comprises a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface complete communication with each other through the communication bus;
the memory is configured to store at least one executable instruction that causes the processor to perform the steps of the 5G slice network congestion warning method based on a time graph network according to any one of claims 1 to 3.
6. A computer storage medium having stored therein at least one executable instruction for causing a processor to perform the steps of the time graph network based 5G slice network congestion warning method of any of claims 1-3.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010558931.9A CN113825165B (en) | 2020-06-18 | 2020-06-18 | 5G slice network congestion early warning method and device based on time diagram network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010558931.9A CN113825165B (en) | 2020-06-18 | 2020-06-18 | 5G slice network congestion early warning method and device based on time diagram network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113825165A CN113825165A (en) | 2021-12-21 |
| CN113825165B true CN113825165B (en) | 2023-08-04 |
Family
ID=78911692
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010558931.9A Active CN113825165B (en) | 2020-06-18 | 2020-06-18 | 5G slice network congestion early warning method and device based on time diagram network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113825165B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114358449A (en) * | 2022-03-21 | 2022-04-15 | 国网天津市电力公司电力科学研究院 | Electric vehicle charging load space-time distribution prediction method based on graph neural network |
| CN115296984B (en) * | 2022-08-08 | 2023-12-19 | 中国电信股份有限公司 | Abnormal network node detection method and device, equipment and storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106549806A (en) * | 2016-10-26 | 2017-03-29 | 清华大学 | A kind of network section manager and its management method |
| CN109154990A (en) * | 2016-06-03 | 2019-01-04 | 英特尔公司 | Lookup convolutional layer in convolutional neural networks |
| CN110383292A (en) * | 2017-04-07 | 2019-10-25 | 英特尔公司 | The method and system through budget and simplified training for deep neural network |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10986010B2 (en) * | 2018-08-09 | 2021-04-20 | At&T Intellectual Property I, L.P. | Mobility network slice selection |
| US11363031B2 (en) * | 2018-08-27 | 2022-06-14 | Ciena Corporation | Network architecture providing device identification and redirection using whitelisting traffic classification |
| US10743211B2 (en) * | 2018-09-25 | 2020-08-11 | Verizon Patent And Licensing, Inc. | Diversified ran architecture with congestion control based on spectrum re-allocation |
-
2020
- 2020-06-18 CN CN202010558931.9A patent/CN113825165B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109154990A (en) * | 2016-06-03 | 2019-01-04 | 英特尔公司 | Lookup convolutional layer in convolutional neural networks |
| CN106549806A (en) * | 2016-10-26 | 2017-03-29 | 清华大学 | A kind of network section manager and its management method |
| CN110383292A (en) * | 2017-04-07 | 2019-10-25 | 英特尔公司 | The method and system through budget and simplified training for deep neural network |
Non-Patent Citations (1)
| Title |
|---|
| 5G网络切片在智能电网领域应用探讨;李亚光;《通讯世界》;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113825165A (en) | 2021-12-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109120462B (en) | Method and device for predicting opportunistic network link and readable storage medium | |
| CN113825165B (en) | 5G slice network congestion early warning method and device based on time diagram network | |
| CN113746696A (en) | Network flow prediction method, equipment, storage medium and device | |
| CN114297036A (en) | Data processing method and device, electronic equipment and readable storage medium | |
| CN113541980B (en) | Network slice fault root cause positioning method, device, computing equipment and storage medium | |
| US11197175B2 (en) | Forcasting time series data | |
| CN111027591B (en) | Node fault prediction method for large-scale cluster system | |
| CN116489038A (en) | Network traffic prediction method, device, equipment and medium | |
| CN112651534A (en) | Method, device and storage medium for predicting resource supply chain demand | |
| CN113824575B (en) | Method and device for identifying fault node, computing equipment and computer storage medium | |
| CN113541986B (en) | Fault prediction method and device for 5G slice and computing equipment | |
| CN111935005A (en) | Data transmission method, device, processing equipment and medium | |
| CN113822454B (en) | Method and device for predicting slice complaint processing resources | |
| CN113825161B (en) | 5G slice anomaly detection method and device based on depth self-coding neural network | |
| CN113825148B (en) | Method and device for determining alarm grade of network node and computing equipment | |
| CN116804964A (en) | Data processing method, device, electronic equipment and readable storage medium | |
| CN113762978B (en) | Complaint delimiting method and device for 5G slicing user and computing equipment | |
| CN113949633A (en) | 5G network slice disaster recovery pool resource management method and device based on machine learning | |
| CN113825151B (en) | Method and device for predicting relationship between slice alarms based on relationship graph convolution network | |
| CN112203311B (en) | Network element abnormity diagnosis method, device, equipment and computer storage medium | |
| CN114143163B (en) | Slice false alarm identification method and device based on graph attention network | |
| CN112990425A (en) | Automatic classification method of 5G network slices, device thereof, electronic equipment and computer storage medium | |
| CN111767934A (en) | Image identification method and device and electronic equipment | |
| CN113806498A (en) | Online question-answering method and device for network slice instance running state | |
| CN113810212B (en) | Root cause positioning method and device for 5G slice user complaints |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |