CN116916463A

CN116916463A - Network resource scheduling method, device, computer equipment and storage medium

Info

Publication number: CN116916463A
Application number: CN202311068269.9A
Authority: CN
Inventors: 陈麒; 庄一嵘; 尹之帆; 曾其妙; 潘庆
Original assignee: China Telecom Technology Innovation Center; China Telecom Corp Ltd
Current assignee: China Telecom Technology Innovation Center; China Telecom Corp Ltd
Priority date: 2023-08-23
Filing date: 2023-08-23
Publication date: 2023-10-20

Abstract

The application relates to a network resource scheduling method, a network resource scheduling device, computer equipment and a storage medium. The method comprises the following steps: determining target graph characteristics of the target network based on the current space-time diagrams of all target position nodes contained in the target network; the current time-space diagram is determined based on the current time, position data corresponding to each target position node and network topology structure data of the target network; inputting the target graph characteristics into a target QoS prediction model to obtain target QoS prediction parameters respectively corresponding to each target position node at a target time point; the target QoS prediction model is obtained based on sample graph characteristics corresponding to each target position node in a training mode; and determining a network resource scheduling strategy corresponding to each target position node based on the target QoS prediction parameters corresponding to each target position node at the target time point. By adopting the method, the QoS scheduling efficiency in the network can be improved.

Description

Network resource scheduling method, device, computer equipment and storage medium

Technical Field

The present application relates to the field of wireless network communication technology, and in particular, to a network resource scheduling method, apparatus, computer device, storage medium, and computer program product.

Background

With the development of communication technology, wi-Fi 6 technology appears, and Wi-Fi 6 technology is used as a next generation wireless network standard, has the characteristics of high speed, low delay, large capacity and wide coverage, can meet the requirements of large-scale data transmission and high-density equipment connection, and is currently applied to intelligent home, internet of vehicles, industrial Internet of things, VR (Virtual Reality) and AR (Augmented Reality ) scenes. The technologies such as the internet of things and the cloud computing require high-quality wireless network communication, and QoS (Quality of Service ) is an important index in the wireless network communication, so as to ensure reliability, instantaneity and bandwidth utilization rate of network transmission, therefore, the quality of the wireless network communication can be improved by a proper QoS scheduling method, and the wireless network communication requirements of the technologies such as the internet of things and the cloud computing are met.

The existing QoS scheduling method of the Wi-Fi 6 network mainly divides the data flow into a plurality of categories through the destination address in the Wi-Fi data packet, and distributes different priorities for each category so as to meet different QoS requirements.

However, the related art requires end-to-end configuration and management at each node of the network in the implementation process. And the QoS scheduling efficiency is lower due to the high complexity of the implementation process of end-to-end configuration and management.

Disclosure of Invention

In view of the foregoing, it is desirable to provide a network resource scheduling method, apparatus, computer device, computer readable storage medium, and computer program product.

In a first aspect, the present application provides a network resource scheduling method. The method comprises the following steps:

determining target graph characteristics of a target network based on current space-time diagrams of all target position nodes contained in the target network; the current time-space diagram is determined based on the current time, the position data corresponding to each target position node and the network topology structure data of the target network;

inputting the target graph characteristics into a target QoS prediction model to obtain target QoS prediction parameters respectively corresponding to each target position node at a target time point; the target QoS prediction model is obtained by training based on sample graph features corresponding to each target position node;

and determining a network resource scheduling strategy corresponding to each target position node based on the target QoS prediction parameters corresponding to each target position node at the target time point.

In one embodiment, the inputting the target graph feature into a target QoS prediction model to obtain target QoS prediction parameters corresponding to each target location node at a target time point includes:

Based on the target graph characteristics and a first QoS prediction model, obtaining initial QoS prediction parameters respectively corresponding to each target position node contained in the target network at the current time; the first QoS prediction model is obtained based on sample graph feature training of a sample network; the sample graph characteristics are determined based on network topology data of the sample network, position data of each sample position node contained in the sample network and sample QoS parameters of each sample position node;

and inputting initial QoS prediction parameters corresponding to the target position nodes at the current time respectively into a second QoS prediction model to obtain target prediction QoS parameters corresponding to the target position nodes at the target time point respectively, wherein the second QoS prediction model is trained based on the output result of the first QoS prediction model.

In one embodiment, the obtaining, based on the target graph feature and the first QoS prediction model, each target location node included in the target network, before the initial QoS prediction parameters corresponding to the current time respectively, the method further includes:

acquiring sample data, wherein the sample data comprises network topology data of the sample network at a sample time point, position data of each sample position node contained in the sample network and sample QoS parameters of each sample position node;

Determining a sample graph characteristic corresponding to the sample data based on network topology data of the sample network at the sample time point, position data of each sample position node contained in the sample network and sample QoS parameters of each sample position node;

inputting the characteristics of the sample graph into a first QoS prediction model to be trained, and obtaining predicted QoS parameters corresponding to each sample position node contained in the sample network;

and updating the model parameters of the first QoS prediction model to be trained based on the predicted QoS parameters and the sample QoS parameters respectively corresponding to the sample position nodes to obtain a trained first QoS prediction model.

In one embodiment, the method further comprises:

acquiring predicted QoS parameters corresponding to each sample position node output by the trained first QoS prediction model;

performing dimension reduction processing on the predicted QoS parameters corresponding to the sample position nodes through a one-dimensional convolutional neural network to obtain the predicted QoS parameters corresponding to the dimension reduced sample position nodes;

and training the second QoS prediction model based on the predicted QoS parameters corresponding to the sample position nodes after the dimension reduction to obtain a trained second QoS prediction model.

In one embodiment, the target graph features include node features and edge features; the determining the target graph characteristics of the target network based on the current space-time diagrams of the target position nodes included in the target network comprises the following steps:

determining a current time-space diagram of each target position node based on network topology structure data of each target position node at the current time and position data corresponding to each target position node contained in a target network;

determining node characteristics and edge characteristics corresponding to the current space-time diagram based on a characteristic extraction algorithm; the node characteristics represent position data corresponding to the target position nodes respectively, and the edge characteristics represent network connection relations among the position nodes corresponding to the edge characteristics.

In one embodiment, the determining, based on the target QoS predicted parameters corresponding to the target location nodes at the target time points, a network resource scheduling policy corresponding to the target location nodes includes:

determining QoS error values corresponding to the target position nodes respectively based on the target QoS prediction parameters corresponding to the target position nodes at the target time points and the QoS parameters of the target position nodes at the current time;

And determining network resource scheduling strategies corresponding to the target position nodes respectively according to the QoS error values.

In one embodiment, after determining the network resource scheduling policy corresponding to each target location node based on the target QoS prediction parameters corresponding to each target location node at the target time point, the method further includes:

and aiming at each target position node, scheduling network resources to the network equipment corresponding to the target position node based on the network resource scheduling strategy corresponding to the target position node.

In a second aspect, the application further provides a network resource scheduling device. The device comprises:

the map feature determining module is used for determining target map features of the target network based on the current space-time maps of all target position nodes contained in the target network; the current time-space diagram is determined based on the current time, the position data corresponding to each target position node and the network topology structure data of the target network;

the QoS parameter acquisition module is used for inputting the target graph characteristics into a target QoS prediction model to obtain target QoS prediction parameters respectively corresponding to the target position nodes at target time points; the target QoS prediction model is obtained by training based on sample graph features corresponding to each target position node;

And the scheduling policy determining module is used for determining the network resource scheduling policy corresponding to each target position node based on the target QoS prediction parameters corresponding to each target position node at the target time point.

In one embodiment, the QoS parameter obtaining module is specifically configured to obtain, based on the target graph feature and the first QoS prediction model, initial QoS prediction parameters corresponding to each target location node included in the target network at a current time; the first QoS prediction model is obtained based on sample graph feature training of a sample network; the sample graph characteristics are determined based on network topology data of the sample network, position data of each sample position node contained in the sample network and sample QoS parameters of each sample position node;

In one embodiment, the apparatus further includes a first training module, specifically configured to:

In one embodiment, the apparatus further comprises a second training module, specifically configured to:

In one embodiment, the target graph features include node features and edge features, and the graph feature determining module is specifically configured to:

In one embodiment, the scheduling policy determining module is specifically configured to:

In one embodiment, the apparatus further includes a network resource scheduling module, specifically configured to schedule, for each target location node, network resources to network devices corresponding to the target location node based on a network resource scheduling policy corresponding to the target location node.

In a third aspect, the present application also provides a computer device. The computer device comprises a memory storing a computer program and a processor implementing the steps of the method according to the first aspect when the processor executes the computer program.

In a fourth aspect, the present application also provides a computer-readable storage medium. The computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the method according to the first aspect.

In a fifth aspect, the present application also provides a computer program product. The computer program product comprising a computer program which, when executed by a processor, implements the steps of the method according to the first aspect.

The network resource scheduling method, the network resource scheduling device, the computer equipment, the storage medium and the computer program product are used for determining target graph characteristics of a target network based on the current space-time graphs of all target position nodes contained in the target network; the current time-space diagram is determined based on the current time, the position data corresponding to each target position node and the network topology structure data of the target network; inputting the target graph characteristics into a target QoS prediction model to obtain target QoS prediction parameters respectively corresponding to each target position node at a target time point; the target QoS prediction model is obtained by training based on sample graph features corresponding to each target position node; and determining a network resource scheduling strategy corresponding to each target position node based on the target QoS prediction parameters corresponding to each target position node at the target time point.

The server can obtain a current time-space diagram of the target network according to the network topology structure of the target network and the position data of each target position node in the target network, wherein the current time-space diagram can comprise a plurality of target position nodes and connecting edges of the plurality of target position nodes, and the time-space diagram can represent the connection relation between the target position nodes. Based on this, the server determines target graph features of the current space-time graph and inputs the target graph features into the trained target QoS prediction model. Because the target QoS prediction model is obtained based on sample graph feature training, the target QoS prediction model can analyze implicit features in the target graph features so as to predict a plurality of target position nodes corresponding to the target graph features and target QoS prediction parameters at future time. Therefore, the server can obtain the target QoS prediction parameters corresponding to the plurality of target position nodes in the target network, so that the server determines the network resource scheduling strategies of the plurality of target position nodes according to the target QoS prediction parameters. The server may schedule network resources according to a network resource scheduling policy so that QoS parameters of a plurality of target location nodes in the target network reach the target QoS prediction parameters. That is, the target QoS predicted parameter is a reference value for the server to schedule network resources, and the server can schedule quickly according to the predicted reference value, so that QoS parameters of a plurality of target location nodes at the current time become the target QoS predicted parameter, thereby achieving the effect of improving QoS scheduling efficiency.

Drawings

FIG. 1 is an application environment diagram of a network resource scheduling method in one embodiment;

FIG. 2 is a flow chart of a method for scheduling network resources in one embodiment;

FIG. 3 is a flowchart illustrating steps for obtaining target QoS prediction parameters according to one embodiment;

FIG. 4 is a flowchart illustrating steps for training a first QoS predictive model in one embodiment;

FIG. 5 is a flowchart illustrating steps for training a second QoS predictive model in one embodiment;

FIG. 6 is a flow diagram of steps for determining target graph features in one embodiment;

FIG. 7 is a flowchart illustrating steps for determining a network resource scheduling policy in one embodiment;

FIG. 8 is a flowchart of a method for scheduling network resources according to another embodiment;

FIG. 9 is a block diagram of a network resource scheduling apparatus in one embodiment;

fig. 10 is an internal structural view of a computer device in one embodiment.

Detailed Description

The present application will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present application more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the application.

The network resource scheduling method provided by the embodiment of the application can be applied to an application environment shown in figure 1. The method provided in the embodiment of the present application may be applied to a terminal or a server, and the following description illustrates that the method is applied to the server, where the ue 102 communicates with the server 104 through a communication network, and transmits location information of the ue and network topology data combined by a plurality of ues to the server. The data storage system may store data to be processed by the server 104, where the data may be historical location information generated by the user terminal, historical network topology data, and the like, and the data may also be a pre-stored network resource scheduling policy. The data storage system may be integrated on the server 104 or may be located on a cloud or other network server. The user terminal 102 may be, but not limited to, various personal computers, notebook computers, smart phones, tablet computers, internet of things devices and portable wearable devices, and the internet of things devices may be smart speakers, smart televisions, smart air conditioners, smart vehicle devices, etc. The portable wearable device may be a smart watch, smart bracelet, headset, or the like. The server 104 may be implemented as a stand-alone server or as a server cluster of multiple servers. The application environment is only one application environment in the network resource scheduling method provided by the embodiment of the present application, and not only represents one application environment, but also the network resource scheduling method may include other application environments, for example, the user terminal in fig. 1 may be replaced by a network communication device.

In one embodiment, as shown in fig. 2, a network resource scheduling method is provided, and the method is applied to the server in fig. 1 for illustration, and includes the following steps:

step S202, determining target graph characteristics of the target network based on the current space-time diagrams of the target position nodes contained in the target network.

The target network may be a communication network, and the target network may include a plurality of target location nodes and a plurality of edges, where each edge may be used to connect to a plurality of target location nodes, respectively. The target network may be a WiFi6 network; the target location node may be any network node in the target network, and the target location node may be an Access Point (AP) of the target network, or may be a user terminal accessing the target network through the AP; in one example, the target location node comprises a plurality of APs, and user terminals respectively connected to the respective APs.

The current time-space diagram is determined based on the current time, the location data corresponding to each target location node, and the network topology data of the target network. The location data corresponding to each target location node is a physical location of the communication device or the communication interface corresponding to the target location node, and the location data may be longitude and latitude coordinates, an IP address, a detailed address of a street on which the device is located, and the like. The network topology data of the target network is node information of each target location node corresponding to the target network and connection information between each target location node, and the node information and the connection information can be represented in a graph or a table. The current time space diagram is a time space diagram corresponding to the current time, the time space diagram is used for representing the spatial position relation of each node under different time, and the time space diagram combines two dimensions of time and space and is used for showing the distribution and change of the target position node under different time. Based on the above, the current time-space diagram may include current time information, and location information and connection information corresponding to each target location node at the current time. The target graph features comprise feature information of each target position node in the target network and feature information of the connecting edge corresponding to the target position node, and the target graph features are used for representing implicit features of the whole target network.

Specifically, when the target network is running, each target position node in the target network is changed continuously along with time, so that the target networks at different times are different. Based on the above, the server may acquire, according to a preset time interval, position data corresponding to each target position node of the target network at the current time, and network topology data of the target network. The server establishes a current time-space diagram of the current time based on the position data and the network topology structure data corresponding to each target position node. Based on the current space-time diagram, the server can acquire a plurality of initial diagram features; the server screens the initial graph features to obtain target graph features, wherein the target graph features are the initial graph features which have influence on the target position nodes and the connection relation between the target position nodes or have higher importance. Accordingly, the server may acquire target graph features of the target network existing in the current space-time graph using a preset feature extraction algorithm.

Optionally, the server determines a current time-space diagram of each target location node based on the current time, location data corresponding to each target location node, and network topology data of the target network. The network topology data may include node information, connection information, and a topology map; the node information may include the name and the IP address of the node, and the connection information may include the connection manner between the nodes, the transmission medium of the connection, and the bandwidth of the connection, and the topology graph may be a two-dimensional graph or a three-dimensional graph.

Step S204, inputting the target graph characteristics into the target QoS prediction model to obtain target QoS prediction parameters respectively corresponding to each target position node at the target time point.

The target QoS prediction model is obtained based on sample graph characteristics corresponding to each target position node in a training mode; the target QoS prediction model is used to predict target QoS prediction parameters for each target location node of the target network at a target point in time, which may be any point in time in a future time period. The sample graph feature corresponding to each target position node may be obtained based on a plurality of time-space graphs generated by the target position node in a preset historical time period, so that the sample graph feature includes feature information of each target position node in the target network in the preset historical time period and feature information of a side corresponding to the target position node, and the preset historical time period refers to a certain time period in the past. The target QoS prediction parameters are predicted QoS data of the target location nodes predicted by the target QoS prediction model at time points corresponding to future time periods, one or more target QoS prediction parameters may exist in each target location node in the target network, and the target QoS prediction parameters may include throughput of the target location nodes, link speed, delay and round trip time between the target location nodes.

Specifically, the server inputs the target graph characteristics of the target network into the trained target QoS prediction model, and the target graph characteristics are obtained based on the current time-space graph, so that the target graph characteristics include characteristics corresponding to the current time, characteristics of position data of each target position node of the target network, and characteristics of network topology data of the target network. And the server calls a target QoS prediction model, predicts QoS parameters based on the input target graph characteristics, and obtains target QoS prediction parameters corresponding to each target position node at a target time point.

In one example, the server inputs the target graph feature into a target QoS prediction model to obtain target QoS prediction parameters corresponding to each target location node at a target time point, where the target QoS prediction parameters may include bandwidth, delay, packet loss rate, throughput, and delay.

In one example, the server inputs the target graph feature into the target QoS prediction model to obtain target QoS prediction parameters corresponding to each target location node at the target time point, where the target QoS prediction parameters may be a set of target QoS prediction parameters corresponding to one target location node, or may be a plurality of sets of target QoS prediction parameters corresponding to one target location node and a plurality of connected target location nodes.

Step S206, determining a network resource scheduling strategy corresponding to each target position node based on the target QoS prediction parameters corresponding to each target position node at the target time point.

The target QoS prediction parameter is data of the target position node at a certain moment in the future, which is predicted by a target QoS prediction model, can be used as a reference value for network resource scheduling by a server or a scheduler based on the target QoS prediction parameter, and a preset network resource scheduling strategy is selected based on the reference value, or a network resource scheduling strategy corresponding to the target position node is formulated. The network resource scheduling policy is a policy for performing resource scheduling on target position nodes in the network, and the server can perform network resource scheduling on each target position node in the target network based on the network resource scheduling policy, so that QoS parameters corresponding to each target position node continuously change along with the execution of the network resource scheduling policy.

Specifically, after the server obtains the target QoS prediction parameters corresponding to each target location node in the target network, a preset network resource scheduling policy may be determined according to the target QoS prediction parameters corresponding to each target location node and network topology data of the target network. In one example, the server determines a network resource scheduling policy corresponding to each target location node based on a target QoS prediction parameter corresponding to each target location node at a target time point, where the network resource scheduling policy is a preset multiple network resource scheduling policy, and the server may configure the preset multiple network resource scheduling policy based on the target QoS prediction parameter.

Based on the above network resource scheduling method, it can be known that, according to the network topology structure of the target network and the position data of each target position node in the target network, the server may obtain a current time-space diagram of the target network, where the current time-space diagram may include a plurality of target position nodes and edges of the plurality of target position nodes, and the time-space diagram may represent a connection relationship between the target position nodes. Based on this, the server determines target graph features of the current space-time graph and inputs the target graph features into the trained target QoS prediction model. Because the target QoS prediction model is obtained based on sample graph feature training, the target QoS prediction model can analyze implicit features in the target graph features so as to predict a plurality of target position nodes corresponding to the target graph features and target QoS prediction parameters at future time. Therefore, the server can obtain the target QoS prediction parameters corresponding to the plurality of target position nodes in the target network, so that the server determines the network resource scheduling strategies of the plurality of target position nodes according to the target QoS prediction parameters. The server may schedule network resources according to a network resource scheduling policy so that QoS parameters of a plurality of target location nodes in the target network reach the target QoS prediction parameters. That is, the target QoS predicted parameter is a reference value for the server to schedule network resources, and the server can schedule quickly according to the predicted reference value, so that QoS parameters of a plurality of target location nodes at the current time become the target QoS predicted parameter, thereby achieving the effect of improving QoS scheduling efficiency.

In one embodiment, as shown in fig. 3, the step of inputting the target graph feature into the target QoS prediction model to obtain the specific implementation manner of the target QoS prediction parameters corresponding to each target location node at the target time point, includes:

step S302, based on the target graph characteristics and the first QoS prediction model, obtaining initial QoS prediction parameters respectively corresponding to each target position node contained in the target network at the current time.

The target QoS prediction model comprises a first QoS prediction model and a second QoS prediction model, wherein the first QoS prediction model is obtained based on sample graph feature training of a sample network; the sample graph features are determined based on network topology data of a sample network, position data of each sample position node contained in the sample network and sample QoS parameters of each sample position node; accordingly, the sample network is a network used as a training sample, and the network topology data corresponding to the sample network, the position data of each sample position node, and the sample QoS parameters of the sample position node are all training data for training the first QoS prediction model. The first QoS prediction model analyzes the spatial characteristics of the sample network based on training data during training, and the first QoS prediction model can predict initial QoS prediction parameters of each target position node in the target network based on analysis results after multiple rounds of training.

Based on the above, the trained first QoS prediction model may output an initial QoS prediction parameter in a model application process, where the initial QoS prediction parameter may be a QoS parameter of a current time corresponding to each target location node corresponding to the target network during the model application. In one example, if the server inputs target graph features into the target first QoS prediction model, the server outputs initial QoS prediction parameters for each target location node in the target network.

Specifically, after the server trains to obtain a first QoS prediction model based on the sample graph features of the sample network, the server inputs the target graph features to the trained first QoS prediction model to obtain initial QoS prediction parameters respectively corresponding to each target position node included in the target network at the current time.

Step S304, the initial QoS prediction parameters corresponding to the target position nodes at the current time are input into the second QoS prediction model, and the target prediction QoS parameters corresponding to the target position nodes at the target time points are obtained.

The second QoS prediction model is trained based on the output result of the first QoS prediction model, and when the second QoS prediction model is applied, the output result of the first QoS prediction model may be an initial QoS prediction parameter of each target location node in the target network. The second QoS prediction model is configured to predict a target QoS prediction parameter for each target location node in the target network at a target time point, where the target time point is any one of the future time periods.

Specifically, the server inputs initial QoS prediction parameters corresponding to each target position node at the current time to the second QoS prediction model to obtain target prediction QoS parameters corresponding to each target position node at the target time point.

In one example, the training data of the second QoS prediction model may be an output of the first QoS prediction model, where the output of the first QoS prediction model may be an initial QoS prediction parameter for each sample location node in the sample network.

In this embodiment, the QoS parameters of each target location node in the target network at the current time can be obtained by inputting the target graph feature into the first QoS prediction model to obtain the initial QoS prediction parameters; and inputting the initial QoS prediction parameters into a second QoS prediction model to obtain target prediction QoS parameters, determining the QoS parameters of a certain target time point in the future on the basis of the initial QoS prediction parameters, and improving the accuracy of the model prediction QoS parameters through the operation of the two models.

In one embodiment, as shown in fig. 4, in step "based on the target graph feature and the first QoS prediction model, each target location node included in the target network is obtained, and before the initial QoS prediction parameters corresponding to the current time respectively", a specific implementation manner of the network resource scheduling method further includes:

In step S402, sample data is acquired.

The sample data comprises network topology data of a sample network at a sample time point, position data of each sample position node contained in the sample network and sample QoS parameters of each sample position node. The sample data is used to train a first QoS prediction model.

Based on this, the sample time points are time points at which sample data are acquired, and each sample time point may correspond to a set of sample data, for example, all sample data at a first time point may be regarded as first sample data, all sample data at a second time point may be regarded as second sample data, and the data structures of the first sample data and the second sample data are the same. Accordingly, the sample network includes a plurality of sample location nodes, each sample location node including location data and a sample QoS parameter corresponding to the sample location node.

Specifically, the server may obtain, from the database, network topology data of the sample network at the sample time point, location data of each sample location node included in the sample network, and a sample QoS parameter of each sample location node.

Step S404, determining a sample graph feature corresponding to the sample data based on the network topology data of the sample network at the sample time point, the position data of each sample position node included in the sample network, and the sample QoS parameter of each sample position node.

Wherein the sample network may be one of a communication network, a plurality of sample location nodes may be included in the sample network, and a plurality of edges may be included in the sample network, the plurality of edges may be used to connect the plurality of sample location nodes. The location data corresponding to each sample location node is a physical location of the communication device or the communication interface corresponding to the sample location node, and the location data may be geographical location data. The network topology data of the sample network is node information of each target location node corresponding to the sample network, and connection information between each target location node. The sample graph features comprise feature information of each sample position node in the target network and feature information of a connecting edge corresponding to the sample position node, and the sample graph features are used for representing implicit features of the whole sample network.

Accordingly, the sample QoS parameter is an existing QoS parameter corresponding to a sample location node in the sample network, and the sample QoS parameter may be a tag value of sample data, which is used for training the first QoS prediction model.

Specifically, the server determines a plurality of different sample time-space diagrams based on network topology data of the sample network, position data of each sample position node, and sample QoS parameters of each sample position node at a plurality of different sample time points. The server obtains sample graph features corresponding to each sample time space graph based on a preset feature extraction algorithm, wherein the sample time space graph is a time space graph corresponding to sample time points, the time space graph is used for representing spatial position relations of each node under different time, one sample time space graph corresponds to one sample graph feature, and in one example, the server determines the sample graph features corresponding to each sample time space graph based on the preset graph theory algorithm.

Step S406, inputting the sample graph characteristics into a first QoS prediction model to be trained, and obtaining the predicted QoS parameters corresponding to each sample position node included in the sample network.

Step S408, based on the predicted QoS parameters and the sample QoS parameters corresponding to the sample position nodes, updating the model parameters of the first QoS predicted model to be trained to obtain a trained first QoS predicted model.

The predicted QoS parameters corresponding to the sample position nodes are the results of the untrained first QoS prediction model continuously output in training, and the server updates the predicted QoS parameters and the loss values corresponding to the sample QoS parameters through multiple training, so that the predicted QoS parameters of the trained first QoS prediction model are similar to the sample QoS parameters.

Specifically, the server may calculate, based on the characteristics of the sample map and the first QoS prediction model to be trained, a predicted QoS parameter corresponding to the first QoS prediction model to be trained, calculate a loss value corresponding to a loss function between the predicted QoS parameter and the sample QoS parameter, and update the parameter of the first QoS prediction model to obtain an updated model when it is determined that the loss value does not satisfy the training completion condition. And re-executing the step of obtaining the sample data based on the updated model until a preset training completion condition is met, so as to obtain a trained first QoS prediction model.

In one example, in the case where the server determines that the current number of training iterations has met a preset number of training iterations threshold, the server may determine that a preset training completion condition has currently been met; in another example, in a case where the server determines that the loss value corresponding to the currently calculated loss function has satisfied a preset convergence condition, the server may determine that the preset training completion condition is currently satisfied, the preset convergence condition may be that the loss value corresponding to the loss function has not changed, or that the loss value has reached a minimum loss value threshold value, or the like.

In this embodiment, the first QoS prediction model for accurately predicting the initial QoS prediction parameters in the target network can be obtained by using the sample data having the same data structure as the target network as training data and inputting the training data into the untrained first QoS prediction model for training. Thereby improving the prediction accuracy of the first QoS prediction model.

In one embodiment, as shown in fig. 5, a specific implementation manner of the network resource scheduling method further includes:

step S502, obtaining the predicted QoS parameters corresponding to the position nodes of each sample outputted by the trained first QoS prediction model.

The predicted QoS parameters corresponding to the sample position nodes output by the first QoS prediction model may be predicted QoS parameters corresponding to the sample data, and after the training of the first QoS prediction model is completed, the second QoS prediction model is trained based on the output result of the first QoS prediction model.

Specifically, the server inputs sample graph features of a plurality of sample time points in sample data into a trained first QoS prediction model to obtain predicted QoS parameters of each sample position node corresponding to the plurality of sample time points.

And step S504, performing dimension reduction processing on the predicted QoS parameters corresponding to the sample position nodes through a one-dimensional convolutional neural network to obtain the predicted QoS parameters corresponding to the dimension reduced sample position nodes.

The one-dimensional convolutional neural network is used for reducing the dimension of the output result of the first QoS prediction model.

Specifically, the server acquires dimension reduction dimensions preset by a user, and performs dimension reduction processing on the predicted QoS parameters corresponding to the sample position nodes according to a convolution kernel and an output channel which are arranged in the one-dimensional convolution neural network, so as to obtain the predicted QoS parameters corresponding to the sample position nodes after dimension reduction.

And step S506, training the second QoS prediction model based on the predicted QoS parameters corresponding to the position nodes of each sample after dimension reduction to obtain a trained second QoS prediction model.

The second QoS prediction model is used for predicting a target QoS prediction parameter of each target location node in the target network at a target time point, where the target time point is any time point in a future time period.

Specifically, in one example, the server inputs the predicted QoS parameters corresponding to the position nodes of each sample after the dimension reduction to the untrained second QoS prediction model, extracts time sequence feature information through structures such as a convolution layer, a pooling layer, a full connection layer and the like, combines the spatial features and the time sequence features of the initial QoS prediction parameters to obtain feature representations corresponding to the position nodes of each sample, and finally obtains the trained second QoS prediction model.

In this embodiment, the second QoS prediction model for accurately predicting the target QoS prediction parameters in the target network can be obtained by using the predicted QoS parameters corresponding to the sample location nodes output by the first QoS prediction model as training data and inputting the training data into the untrained second QoS prediction model for training. Thereby improving the prediction accuracy of the QoS prediction model.

In one embodiment, the target graph features include node features and edge features, and as shown in fig. 6, a specific implementation manner of the step of determining the target graph features of the target network based on the current space-time diagrams of the target location nodes included in the target network includes:

Step S602, determining a current time-space diagram of each target location node based on network topology data of each target location node included in the target network at the current time and location data corresponding to each target location node.

The current time-space diagram of each target position node is a current time-space diagram corresponding to the target network, and the current time-space diagram comprises network topology structure data of current time and position data of each target position node. The data structure of the current space-time diagram may be a diagram structure.

Specifically, the server may store the position data of each target position node at the current time to the position node corresponding to the network topology structure data; and the server determines the current time-space diagram corresponding to each target position node according to the position node corresponding to the network topology structure data, the edge corresponding to the network topology structure data and the connection relation between the position node and the edge.

Step S604, determining node features and edge features corresponding to the current space-time diagram based on a feature extraction algorithm.

The node features represent position data corresponding to each target position node, and the edge features represent network connection relations between the position nodes corresponding to the edge features, so that the target graph features can be obtained by combining the node features and the edge features by a server.

Specifically, the server acquires node characteristics and edge characteristics corresponding to the current space-time diagram according to a characteristic extraction algorithm, and combines the node characteristics and the edge characteristics to obtain target diagram characteristics; the feature extraction algorithm may be graph theory algorithm, or word bag model, TF-IDF (Term Frequency-Inverse Document Frequency, word Frequency-inverse document Frequency).

In this embodiment, the current time-space diagram is determined according to the network topology structure data of the target network and the position data of the target position node corresponding to the target network, and the node characteristics and the edge characteristics are obtained based on the characteristic extraction algorithm, so as to obtain more accurate target diagram characteristics.

In one embodiment, as shown in fig. 7, the step of determining a specific implementation manner of the network resource scheduling policy corresponding to each target location node based on the target QoS prediction parameters corresponding to each target location node at the target time point includes:

step S702 determines QoS error values corresponding to the target position nodes based on the target QoS prediction parameters corresponding to the target position nodes at the target time points and the QoS parameters of the target position nodes at the current time.

Step S704, according to the QoS error values, determining the network resource scheduling strategies corresponding to the target position nodes respectively.

The target QoS prediction parameters corresponding to the target position nodes at the target time points are output results of the target QoS prediction model, and the output results are used for predicting QoS parameters corresponding to the future target time points. The QoS parameter of each target position node at the current time is the real QoS parameter of each target position node under the current time of the target network. The error value is a phase difference value between the target QoS predicted parameter and the QoS parameter at the current time. The network resource scheduling strategy is a method for changing QoS parameters of each target position node in the target network, which is stored and configured in advance in a server or a database.

Specifically, the server may determine, according to the target QoS prediction parameters corresponding to the target location nodes at the target time points and the QoS parameters corresponding to all the target location nodes in the target network at the current time, a QoS error value between the QoS parameters at the current time and the target QoS prediction parameters. The server can use the error value as a matching parameter to match and obtain network resource scheduling strategies corresponding to the target position nodes respectively.

As an example, the server may determine, according to each QoS error value, one or more network resource scheduling policies corresponding to the target network; the server may also determine a corresponding network resource scheduling policy for each target location node separately based on the QoS error values.

Optionally, the server may determine a target network resource scheduling policy corresponding to the target error value based on a preset correspondence between the error value and the scheduling policy.

In this embodiment, the server determines the network resource scheduling policy of each target location node according to the QoS error values corresponding to each target location node. Different network resource scheduling strategies can be provided end to end, so that QoS scheduling efficiency can be improved more accurately and rapidly.

In one embodiment, after determining the network resource scheduling policy corresponding to each target location node based on the target QoS prediction parameters corresponding to each target location node at the target time point in the step "the specific implementation process of the network resource scheduling method further includes:

The network device is an entity device corresponding to the target location node, and the number of the entity devices corresponding to the target location node may be one or more. The network device may be an AP, may be a user terminal, and the user terminal may be various terminals using a network.

Specifically, the server may allocate network resources, which may be network traffic, network bandwidth, or the like, to one or more network devices of the target location node according to a network resource scheduling policy corresponding to the target location node.

Optionally, the server may send the network resource scheduling policy of the target location node to the QoS scheduler, so that the QoS scheduler schedules network resources to the target network device corresponding to the target location node according to the network resource scheduling policy.

In this embodiment, the server performs network resource scheduling on the network devices of each target location node according to the network resource scheduling policies corresponding to each target location node, so as to improve flexibility of QoS scheduling of all nodes in the target network.

As shown in fig. 8, the following describes in detail a specific implementation procedure of the above network resource scheduling method in combination with a specific embodiment:

the embodiment of the application provides a dynamic QoS scheduling method based on Wi-Fi 6 technology. The method comprises a graph data acquisition and processing module, a QoS scheduling algorithm module and a QoS scheduler module.

The graph data acquisition and processing module is responsible for collecting network topology structure data, user position data and QoS parameters (indexes such as link speed, throughput, delay and round trip time) in the Wi-Fi 6 network, processing the data, and extracting graph characteristics for training of a QoS scheduling algorithm.

The QoS scheduler can dynamically schedule the Wi-Fi6 network according to the prediction result of the QoS scheduling algorithm, so that the QoS performance is improved.

The embodiment of the application adopts a QoS scheduling algorithm combining a graph rolling network (GCN) model and an Omni-Scale 1D CNN model, can use a QoS scheduler to predict and schedule QoS parameters in a Wi-Fi6 network, can improve the QoS performance of the Wi-Fi6 network, and provides better user experience.

The specific implementation process of the embodiment of the application is as follows:

step S802, information such as QoS parameters, user data, network topology and the like in the network is collected and preprocessed.

Specifically, the server collects network topology data, user location data, qoS parameters (link speed, throughput, delay, round trip time, etc. indexes) in the Wi-Fi6 network through the data collection and processing module, and constructs a space-time diagram by using the location information in the Wi-Fi6 network. And processes the data to extract the graph features for training of QoS scheduling algorithms.

Step S804, the preprocessed data is input into the GCN model for training to predict QoS parameters of different positions of the WiFi6 network.

Specifically, the server inputs the graph features into the GCN model for spatial feature analysis to predict QoS parameters for different locations in the Wi-Fi6 network.

In step S806, the output result of the GCN model is used as the input of the Omni-Scale 1D CNN model to predict the QoS parameters of the Wi-Fi 6 network at the future time.

Specifically, the server takes the output result of the GCN model as input of the Omni-Scale 1D CNN model, and performs timing characteristic analysis to predict QoS parameters at future time in the Wi-Fi 6 network. And optimizing and updating the prediction model according to the real-time feedback information, so that the accuracy and stability of prediction are improved.

In step S808, the dynamic QoS scheduling algorithm is implemented as a set of QoS scheduler software programs to dynamically allocate network resources to improve network performance and user experience.

Specifically, the server implements the dynamic QoS scheduling algorithm as a set of QoS scheduler software programs and is installed on the APs and controllers in the Wi-Fi 6 network. The program may run periodically to dynamically allocate network resources to improve network performance and user experience.

The specific details of the application may be modified or varied depending upon the particular circumstances.

Optionally, the embodiment of the application further comprises a Wi-Fi 6-based QoS scheduling algorithm step:

(1) QoS parameters of different positions are represented as a graph through a graph data collection and processing module, wherein each Wi-Fi 6 node represents a position, and each side represents a connection relation between the positions.

(2) Inputting the node characteristics and the edge characteristics into a GCN model for space characteristic analysis, extracting characteristic vectors of each node in different time periods, and calculating the characteristic vectors of the nodes according to a formula:

wherein, the liquid crystal display device comprises a liquid crystal display device,feature vector representing ith node in the first layer, n represents the number of nodes, sigma represents the activation function, W ^(l) Representing the weight parameter matrix of the first layer.

c _ij Representing weights on the connection edges from node i to node j or values on the adjacency matrix. Dividing the formula by c _ij The method aims at removing the influence of different node degrees on the distribution imbalance.

For each node i, its initial feature vector is the input node feature vector x _i I.e.

The feature vector for each layer is then updated according to the formula:

where N (i) represents a set of nodes adjacent to node i.

(3) And performing dimension reduction processing on the GCN output result so as to effectively extract the time sequence characteristics in the Omni-Scale 1D CNN model. Specifically, the one-dimensional convolutional neural network is used for reducing the dimension of the GCN output result:

specifically, the GCN model outputs the result ofWhere n is the number of nodes and L is the number of layers of the GCN model. And performing dimension reduction processing on the result through a one-dimensional convolutional neural network, wherein a convolutional kernel is k, and an output channel is c. The output of each node is a vector in c dimensions, namely:

Wherein f is an activation function, [; a third party; a third party; .]Representing the stitching of vectors into a matrix, w ^k ，b ^(l) The weight and bias of the kth convolution kernel are represented.

Then, taking the result after dimension reduction as input of an Omni-Scale 1D CNN model, and extracting time sequence characteristic information through structures such as a convolution layer, a pooling layer, a full connection layer and the like:

the output result of the Omni-Scale 1D CNN model is:wherein L' is the number of layers of the Omni-Scale 1D CNN model. />

Wherein f is an activation function, [; a third party; a third party; .]Representing the stitching of vectors into a matrix, z _i ^(k′) Representing the output of the k' layer convolution.

(4) The spatial features and the timing features are combined to derive a feature representation of each Wi-Fi 6 node over different time periods to predict QoS parameters of each node over future time periods.

(5) And inputting the prediction result into a QoS scheduler for dynamic scheduling to optimize the QoS performance of the Wi-Fi 6 network.

The output result of the GCN model is used as the input purpose and benefit of the Omni-Scale 1D CNN model:

the GCN model can effectively capture space feature information, the Omni-Scale 1D CNN model can analyze data from a time sequence angle, more accurate time sequence features are extracted, and QoS parameters at future time can be predicted more accurately by combining the feature information of the two aspects, so that the QoS performance of the WiFi6 network is optimized.

The embodiment of the application has the main creativity that the server adopts a dynamic QoS scheduling method based on Wi-Fi 6 technology, combines a graph convolution network and an Omni-Scale 1D CNN model, realizes real-time scheduling and optimization of each node in the network, improves the throughput and QoS of the network, and has higher practical value and innovation. Effective technical support can be provided for optimization and management of Wi-Fi 6 networks.

In addition, the dynamic QoS scheduling method based on Wi-Fi 6 technology adopted in the embodiment of the application can combine the movement track of the user with the performance parameters of the network equipment to realize real-time scheduling of each node in the network. The method can improve the throughput and QoS of the network and improve the network experience of the user.

Accordingly, the graph rolling network (GCNs) and the Omni-Scale 1D CNN model adopted in the embodiment of the application model and predict the network topology, the performance parameters and the user position information of Wi-Fi 6 equipment, and by the method, the model can consider the space and time change, so that future QoS indexes can be predicted more accurately, network scheduling and optimization can be performed in time, and the throughput and QoS of the network are improved.

Optionally, the space-time graph employs a location-based adjacency matrix, wherein each node represents one location in the Wi-Fi 6 network, and edges between nodes represent reachability between the two locations. In the adjacency matrix, each element represents a weight between nodes, and the larger the weight is, the better the reachability between two positions is. At the same time, the adjacency matrix also considers the time factor, namely the position accessibility in the time window.

Optionally, the prediction model adopts a deep learning model of mixing GCN and Omni-Scale 1D CNN, and the model can simultaneously consider space and time information and learn a complex relationship between QoS parameters and position information in a Wi-Fi 6 network through training data so as to improve the prediction precision of the QoS parameters.

Optionally, the dynamic QoS scheduler may dynamically schedule the Wi-Fi 6 network according to the prediction result, so as to optimize QoS performance of the network. The specific implementation mode is as follows:

(1) And selecting a proper QoS strategy according to the prediction result.

Specifically, the QoS scheduler selects an appropriate QoS policy according to the prediction result and the QoS requirements in a specific scenario. For example, according to the demands of the users on the network at the future time obtained by the prediction model, the appropriate strategy can be selected in advance to meet the QoS requirements of the users

(2) And adjusting the network topology structure and the link bandwidth to meet the QoS requirements in different scenes.

Specifically, the server can guide the QoS requirements under the high-bandwidth scene, the low-delay scene and the zero-packet-loss scene according to the prediction result, and dynamically adjust the link bandwidth according to the service perception controller.

(3) And dynamically adjusting network resource allocation according to the user behavior and the application type so as to improve user experience.

Specifically, for example, the application type is real-time voice, where a lower delay and bandwidth need to be reserved for the real-time voice application, and the bandwidth is allocated to the video streaming media application preferentially, and according to the priorities of the users and the applications, the QoS scheduler can continuously adjust the allocation of network resources in real time so as to adapt to the requirement change of different users and applications.

It should be understood that, although the steps in the flowcharts related to the above embodiments are sequentially shown as indicated by arrows, these steps are not necessarily sequentially performed in the order indicated by the arrows. The steps are not strictly limited to the order of execution unless explicitly recited herein, and the steps may be executed in other orders. Moreover, at least some of the steps in the flowcharts described in the above embodiments may include a plurality of steps or a plurality of stages, which are not necessarily performed at the same time, but may be performed at different times, and the order of the steps or stages is not necessarily performed sequentially, but may be performed alternately or alternately with at least some of the other steps or stages.

Based on the same inventive concept, the embodiment of the application also provides a network resource scheduling device for realizing the above-mentioned network resource scheduling method. The implementation of the solution provided by the apparatus is similar to the implementation described in the above method, so the specific limitation in the embodiments of one or more network resource scheduling apparatuses provided below may refer to the limitation of the network resource scheduling method described above, and will not be repeated here.

In one embodiment, as shown in fig. 9, there is provided a network resource scheduling apparatus 900, including: a graph feature determination module 901, a QoS parameter acquisition module 902, and a scheduling policy determination module 903, wherein:

a graph feature determining module 901, configured to determine a target graph feature of a target network based on a current space-time graph of each target location node included in the target network; the current time-space diagram is determined based on the current time, position data corresponding to each target position node and network topology structure data of the target network;

the QoS parameter obtaining module 902 is configured to input the target graph feature to a target QoS prediction model, so as to obtain target QoS prediction parameters corresponding to each target location node at a target time point; the target QoS prediction model is obtained based on sample graph characteristics corresponding to each target position node in a training mode;

The scheduling policy determining module 903 is configured to determine a network resource scheduling policy corresponding to each target location node based on target QoS prediction parameters corresponding to each target location node at a target time point.

Further, the QoS parameter obtaining module 902 is specifically configured to obtain, based on the target graph feature and the first QoS prediction model, initial QoS prediction parameters corresponding to each target location node included in the target network at the current time; the first QoS prediction model is obtained based on sample graph feature training of a sample network; the sample graph features are determined based on network topology data of a sample network, position data of each sample position node contained in the sample network and sample QoS parameters of each sample position node; and inputting initial QoS prediction parameters corresponding to the target position nodes at the current time respectively into a second QoS prediction model to obtain target prediction QoS parameters corresponding to the target position nodes at the target time point respectively, wherein the second QoS prediction model is trained based on the output result of the first QoS prediction model.

Further, the device further comprises a first training module, specifically configured to: acquiring sample data, wherein the sample data comprises network topology data of a sample network at a sample time point, position data of each sample position node contained in the sample network and sample QoS parameters of each sample position node; determining a sample graph characteristic corresponding to sample data based on network topology data of a sample network at a sample time point, position data of each sample position node contained in the sample network and sample QoS parameters of each sample position node; inputting the characteristics of the sample graph into a first QoS prediction model to be trained to obtain predicted QoS parameters corresponding to each sample position node contained in a sample network; based on the predicted QoS parameters and the sample QoS parameters respectively corresponding to the sample position nodes, updating the model parameters of the first QoS predicted model to be trained to obtain a trained first QoS predicted model.

Further, the device further comprises a second training module, specifically configured to: obtaining the predicted QoS parameters corresponding to the position nodes of each sample output by the trained first QoS prediction model; performing dimension reduction processing on the predicted QoS parameters corresponding to the sample position nodes through a one-dimensional convolutional neural network to obtain the predicted QoS parameters corresponding to the sample position nodes after dimension reduction; and training the second QoS prediction model based on the predicted QoS parameters corresponding to the position nodes of each sample after dimension reduction to obtain a trained second QoS prediction model.

Further, the target graph features include node features and edge features, and the graph feature determining module 901 is specifically configured to: determining a current time-space diagram of each target position node based on network topology structure data of each target position node contained in a target network at the current time and position data corresponding to each target position node respectively; determining node characteristics and edge characteristics corresponding to the current space-time diagram based on a characteristic extraction algorithm; the node characteristics represent position data corresponding to each target position node respectively, and the edge characteristics represent network connection relations among the position nodes corresponding to the edge characteristics.

In one embodiment, the scheduling policy determining module 903 is specifically configured to: determining QoS error values corresponding to the target position nodes based on the target QoS prediction parameters corresponding to the target position nodes at the target time points and the QoS parameters of the target position nodes at the current time; and determining network resource scheduling strategies corresponding to the target position nodes respectively according to the QoS error values.

The above-described respective modules in the network resource scheduling apparatus may be implemented in whole or in part by software, hardware, and a combination thereof. The above modules may be embedded in hardware or may be independent of a processor in the computer device, or may be stored in software in a memory in the computer device, so that the processor may call and execute operations corresponding to the above modules.

In one embodiment, a computer device is provided, which may be a server, and the internal structure of which may be as shown in fig. 10. The computer device includes a processor, a memory, an Input/Output interface (I/O) and a communication interface. The processor, the memory and the input/output interface are connected through a system bus, and the communication interface is connected to the system bus through the input/output interface. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database of the computer equipment is used for storing a current time-space diagram generated by the user terminal and a time-space diagram used for training, wherein the current time-space diagram comprises current time, position data corresponding to a target position node, network topology structure data of a target network and the like; the space-time diagram for training includes sample time points of sample location nodes, location data, network topology data, and the like. The input/output interface of the computer device is used to exchange information between the processor and the external device. The communication interface of the computer device is used for communicating with an external terminal through a network connection. The computer program when executed by a processor implements a network resource scheduling method.

It will be appreciated by those skilled in the art that the structure shown in FIG. 10 is merely a block diagram of some of the structures associated with the present inventive arrangements and is not limiting of the computer device to which the present inventive arrangements may be applied, and that a particular computer device may include more or fewer components than shown, or may combine some of the components, or have a different arrangement of components.

In an embodiment, there is also provided a computer device comprising a memory and a processor, the memory having stored therein a computer program, the processor implementing the steps of the method embodiments described above when the computer program is executed.

In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored which, when executed by a processor, carries out the steps of the method embodiments described above.

In an embodiment, a computer program product is provided, comprising a computer program which, when executed by a processor, implements the steps of the method embodiments described above.

It should be noted that, the user information (including but not limited to user equipment information, user personal information, etc.) and the data (including but not limited to data for analysis, stored data, presented data, etc.) related to the present application are information and data authorized by the user or sufficiently authorized by each party, and the collection, use and processing of the related data need to comply with the related laws and regulations and standards of the related country and region.

Those skilled in the art will appreciate that implementing all or part of the above described methods may be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed, may comprise the steps of the embodiments of the methods described above. Any reference to memory, database, or other medium used in embodiments provided herein may include at least one of non-volatile and volatile memory. The nonvolatile Memory may include Read-Only Memory (ROM), magnetic tape, floppy disk, flash Memory, optical Memory, high density embedded nonvolatile Memory, resistive random access Memory (ReRAM), magnetic random access Memory (Magnetoresistive Random Access Memory, MRAM), ferroelectric Memory (Ferroelectric Random Access Memory, FRAM), phase change Memory (Phase Change Memory, PCM), graphene Memory, and the like. Volatile memory can include random access memory (Random Access Memory, RAM) or external cache memory, and the like. By way of illustration, and not limitation, RAM can be in the form of a variety of forms, such as static random access memory (Static Random Access Memory, SRAM) or dynamic random access memory (Dynamic Random Access Memory, DRAM), and the like. The databases referred to in the embodiments provided herein may include at least one of a relational database and a non-relational database. The non-relational database may include, but is not limited to, a blockchain-based distributed database, and the like. The processor referred to in the embodiments provided in the present application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic unit, a data processing logic unit based on quantum computing, or the like, but is not limited thereto.

The technical features of the above embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.

The foregoing examples illustrate only a few embodiments of the application and are described in detail herein without thereby limiting the scope of the application. It should be noted that it will be apparent to those skilled in the art that several variations and modifications can be made without departing from the spirit of the application, which are all within the scope of the application. Accordingly, the scope of the application should be assessed as that of the appended claims.

Claims

1. A method for scheduling network resources, the method comprising:

2. The method of claim 1, wherein said inputting the target graph feature into a target QoS prediction model to obtain target QoS prediction parameters for each of the target location nodes at a target point in time, respectively, comprises:

3. The method according to claim 2, wherein the obtaining, based on the target graph feature and the first QoS prediction model, each target location node included in the target network, before the initial QoS prediction parameter corresponding to the current time, the method further includes:

4. A method according to claim 3, characterized in that the method further comprises:

5. The method of claim 1, wherein the target graph features include node features and edge features; the determining the target graph characteristics of the target network based on the current space-time diagrams of the target position nodes included in the target network comprises the following steps:

6. The method of claim 1, wherein the determining a network resource scheduling policy corresponding to each target location node based on the target QoS prediction parameters corresponding to each target location node at the target time point, comprises:

7. The method of claim 1, wherein after determining the network resource scheduling policy corresponding to each of the target location nodes based on the target QoS prediction parameters corresponding to each of the target location nodes at the target time point, the method further comprises:

8. A network resource scheduling apparatus, the apparatus comprising:

9. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor implements the steps of the method of any of claims 1 to 7 when the computer program is executed.

10. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 7.