CN118118420A - Communication network flow optimization method and system based on artificial intelligence - Google Patents

Abstract

The invention relates to the field of artificial intelligence, in particular to a communication network flow optimization method and system based on artificial intelligence. Acquiring a training communication network flow data set; establishing a GRU-GNN communication network flow prediction model based on the GRU double-channel gating circulating unit and the GNN graph neural network in a combined way to obtain an initial GRU-GNN communication network flow prediction model; combining the initial GRU-GNN communication network flow prediction model with a Self-Attention mechanism to obtain a target GRU-GNN communication network flow prediction model; inputting the training communication network flow data set into a target GRU-GNN communication network flow prediction model for prediction; and generating a first network traffic management strategy according to the predicted communication network traffic data, and carrying out configuration management on the communication network traffic by using an openflow protocol on the wireless router flow table based on the first network traffic management strategy. According to the real-time network condition and service demand, dynamically adjusting the flow management strategy, thereby realizing the accurate control of the network flow.

Description

A communication network traffic optimization method and system based on artificial intelligence

Technical Field

The present invention relates to the field of artificial intelligence, and in particular to a communication network traffic optimization method and system based on artificial intelligence.

Background technique

With the development of communication technology, people's demand for the network is increasing, and communication networks have become a key infrastructure to support the operation of society. However, with the continuous expansion of network scale and the increasing complexity of business needs, communication networks are facing increasing traffic pressure. Traditional traffic management methods are often unable to cope with this dynamically changing environment, resulting in decreased network performance. Similarly, computer networks also need to be constantly updated and developed like mobile communication networks in order to adapt to the increasing demand in the future. At the same time, various new applications continue to emerge. In emerging technologies such as big data, cloud services, artificial intelligence, and the metaverse, new technical applications require high-quality networks as a guarantee. In the face of the continuous development of scientific and technological applications, the distribution of traditional communication network traffic is gradually unable to cope with it. Therefore, how to optimize the communication network traffic is a technical problem that needs to be solved at this stage.

Summary of the invention

The purpose of the present invention is to solve the above problems and to design a communication network traffic optimization method and system based on artificial intelligence.

To achieve the above-mentioned purpose, the technical solution of the present invention is that, further, in the above-mentioned communication network traffic optimization method based on artificial intelligence, the communication network traffic optimization method comprises the following steps:

Acquire historical communication network traffic data in the system, perform data preprocessing on the historical communication network traffic data, and obtain a training communication network traffic data set;

Based on the combination of GRU dual-channel gated recurrent unit and GNN graph neural network, a GRU-GNN communication network traffic prediction model is established to obtain the initial GRU-GNN communication network traffic prediction model;

Combining the initial GRU-GNN communication network traffic prediction model with the Self-Attention mechanism to obtain a target GRU-GNN communication network traffic prediction model;

Acquire a training communication network traffic data set in the system, input the training communication network traffic data set into the target GRU-GNN communication network traffic prediction model for prediction, and obtain predicted communication network traffic data;

Generate a first network traffic management strategy according to the predicted communication network traffic data, and based on the first network traffic management strategy, use the openflow protocol to configure and manage the communication network traffic on the wireless router flow table to obtain the target communication network traffic data;

The target communication network traffic data is monitored to obtain the data change range of the target communication network traffic data in real time, and if the data change range is greater than a set threshold, a second network traffic management strategy is generated.

Further, in the above communication network traffic optimization method, the acquisition system acquires historical communication network traffic data, performs data preprocessing on the historical communication network traffic data, and obtains a training communication network traffic data set, including:

Acquire historical communication network traffic data in the system, wherein the historical communication network traffic data at least includes communication data packet size data, communication traffic type data, source and destination address data, bandwidth usage data, communication traffic distribution data, communication traffic anomaly detection data, and user behavior data acquired through the base station;

Performing data enhancement processing on the historical communication network traffic data to obtain enhanced communication network traffic data;

Supplementing the missing values in the enhanced communication network flow data to obtain complete communication network flow data;

After processing the abnormal values in the complete communication network traffic data, the data is normalized to obtain a training communication network traffic data set.

Further, in the above communication network traffic optimization method, the GRU dual-channel gated recurrent unit and the GNN graph neural network are combined to establish a GRU-GNN communication network traffic prediction model to obtain an initial GRU-GNN communication network traffic prediction model, including:

The GRU dual-channel gated recurrent unit includes a reset gate and an update gate, wherein the calculation formula of the reset gate is as follows:

r _t =σ(W _r ·[h _t-1 ,X _t ]+b _r )

h _t = tanh(W _h ·[r _t *h _t-1 ,X _t ]+b _h )

Among them, σ is the sigmoid function, tanh is the activation function of the model, h _t-1 is the output state vector of the previous moment, X _t is the input vector of the current moment; W _r , W _h are the weight parameters of the model; b _r , b _h are the bias parameters of the model;

The calculation formula of the update gate is as follows:

Z _t =σ(W _Z ·[h _t-1 ,X _t ]+b _Z )

Wherein, W _Z , b _Z are weight parameters and bias parameters, and the update gate is used to control the influence of the output state vector h _t-1 at the previous moment and the new input h _t on the output vector h' _t at the current moment; after the reset gate and the update gate, the final output vector at the current moment t is:

h' _t =(1-Z _t )*h _t-1 +Z _t *h _t

In the GNN graph neural network, Fourier transform is used to map the data in the spatial domain to the frequency domain for convolution, and inverse Fourier transform is used to map the result back to the spatial domain;

The GRU dual-channel gated recurrent unit and the GNN graph neural network are combined to establish a GRU-GNN communication network traffic prediction model, and the initial GRU-GNN communication network traffic prediction model is obtained.

Further, in the above communication network traffic optimization method, the initial GRU-GNN communication network traffic prediction model and the Self-Attention mechanism are combined to obtain a target GRU-GNN communication network traffic prediction model, including:

Based on the combination of GRU-GNN communication network traffic prediction model and Self-Attention mechanism;

The calculation formula of the Self-Attention mechanism is as follows:

Among them, _TA is the temporal attention matrix, _VT , _bT , _U1 , _U2 and _U3 are all matrix parameters that need to be learned in the model;

The elements in _{the TA} are normalized, and the calculation formula of the normalization is as follows:

Multiplying the normalized time attention matrix with the communication network traffic data to dynamically adjust the input data of the initial GRU-GNN communication network traffic prediction model; the initial GRU-GNN communication network traffic prediction model includes at least three representation learning components;

The representation learning component is used to perform representation learning on dynamic spatiotemporal information in communication network traffic data, and three representation learning components are adaptively fused to output prediction results;

The representation learning component consists of four spatiotemporal convolution modules, in which the network topology and time weight are adjusted by the self-attention layer;

The convolutional module composed of GNN graph neural network is used to learn the spatiotemporal features and obtain the target GRU-GNN communication network traffic prediction model.

Further, in the above communication network traffic optimization method, the real-time communication network traffic data in the acquisition system is input into the target GRU-GNN communication network traffic prediction model for prediction to obtain the predicted communication network traffic data, including:

Acquire real-time communication network traffic data in the system, and input the real-time communication network traffic data into the target GRU-GNN communication network traffic prediction model for prediction;

The accuracy of the target GRU-GNN communication network traffic prediction model is evaluated using the RMSE root mean square error;

Setting the Adam optimizer as the optimizer for training the target GRU-GNN communication network traffic prediction model;

The MSE loss function is set as the training loss function of the target GRU-GNN communication network traffic prediction model;

The initial learning rate of the target GRU-GNN communication network traffic prediction model is set to 0.006, and the loss value of the validation set is determined every 32 epochs. If the loss value does not decrease, the learning rate is reduced to half of the original value, and the predicted communication network traffic data is obtained through training;

The predicted communication network traffic data at least includes communication traffic size data, communication traffic type data, and communication traffic distribution data.

Further, in the above communication network traffic optimization method, generating a first network traffic management strategy according to the predicted communication network traffic data, and based on the first network traffic management strategy, using the openflow protocol to configure and manage the communication network traffic on the wireless router flow table to obtain the target communication network traffic data, including:

Generate a first network traffic management strategy based on the predicted communication network traffic data, wherein the first network traffic management strategy includes at least a traffic priority transmission strategy, a traffic shaping strategy, a traffic control strategy, and a traffic isolation strategy;

Communicate with the underlying network through the openflow protocol to obtain network device information and connect to the network device. The openflow protocol allows the SDN controller to control the SDN switch remotely.

Use the openflow protocol to configure and manage the flow tables and metering tables of wireless routers for communication network traffic;

The first network traffic management policy is sent to the network device based on the openflow protocol, and the network device database is configured to execute add and delete commands to obtain the target communication network traffic data.

Further, in the above communication network traffic optimization method, the target communication network traffic data is monitored, and the data change range of the target communication network traffic data is obtained in real time. If the data change range is greater than a set threshold, a second network traffic management strategy is generated, including:

Monitor the target communication network traffic data, and obtain the data change range of the target communication network traffic data every 30 seconds;

If the data change range of the target communication network flow data is less than 20% within 30 seconds, the target communication network flow data is recorded;

If the data variation range of the target communication network traffic data within 30 seconds is greater than 20%, then the variation range of the target communication network traffic data within 60 seconds is calculated;

If the target communication network traffic data changes by more than 25% within 60 seconds, a second network traffic management strategy is generated;

Based on the second network traffic management strategy, the openflow protocol is used to configure and manage the communication network traffic on the wireless router flow table.

Furthermore, in the above-mentioned artificial intelligence-based communication network traffic optimization system, the communication network traffic optimization system includes the following modules:

A flow data acquisition module is used to acquire historical communication network flow data in the system, perform data preprocessing on the historical communication network flow data, and obtain a training communication network flow data set;

A prediction model building module is used to combine the GRU dual-channel gated recurrent unit and the GNN graph neural network to establish a GRU-GNN communication network traffic prediction model, and obtain an initial GRU-GNN communication network traffic prediction model;

A prediction model optimization module, used to combine the initial GRU-GNN communication network traffic prediction model with the Self-Attention mechanism to obtain a target GRU-GNN communication network traffic prediction model;

A communication traffic prediction model is used to obtain a training communication network traffic data set in the system, input the training communication network traffic data set into the target GRU-GNN communication network traffic prediction model for prediction, and obtain predicted communication network traffic data;

A communication traffic management module, configured to generate a first network traffic management strategy according to the predicted communication network traffic data, and to configure and manage the communication network traffic of the wireless router flow table using the openflow protocol based on the first network traffic management strategy to obtain target communication network traffic data;

The communication traffic optimization module is used to monitor the target communication network traffic data and obtain the data change range of the target communication network traffic data in real time. If the data change range is greater than a set threshold, a second network traffic management strategy is generated.

Furthermore, in the above-mentioned artificial intelligence-based communication network traffic optimization system, the traffic data acquisition module includes the following submodules:

An acquisition submodule is used to acquire historical communication network traffic data in the system, wherein the historical communication network traffic data at least includes communication data packet size data, communication traffic type data, source and destination address data, bandwidth usage data, communication traffic distribution data, communication traffic anomaly detection data, and user behavior data acquired through a base station;

An enhancement submodule, used to perform data enhancement processing on the historical communication network flow data to obtain enhanced communication network flow data;

A supplement submodule, used to supplement the missing values in the enhanced communication network flow data to obtain complete communication network flow data;

A submodule is obtained, which is used to process the abnormal values in the complete communication network flow data and then perform data normalization to obtain a training communication network flow data set.

Furthermore, in the above-mentioned artificial intelligence-based communication network traffic optimization system, the communication traffic prediction model includes the following submodules:

An input submodule, used for acquiring real-time communication network traffic data in the system, and inputting the real-time communication network traffic data into the target GRU-GNN communication network traffic prediction model for prediction;

An evaluation submodule, used to evaluate the accuracy of the target GRU-GNN communication network traffic prediction model using RMSE root mean square error;

An optimizer submodule, used to set the Adam optimizer as the optimizer for training the target GRU-GNN communication network traffic prediction model;

A loss function submodule, used to set the MSE loss function as the training loss function of the target GRU-GNN communication network traffic prediction model;

A learning rate submodule is used to set the initial learning rate of the target GRU-GNN communication network traffic prediction model to 0.006, and to determine the loss value of the validation set every 32 epochs. If the loss value does not decrease, the learning rate is reduced to half of the original value, and the predicted communication network traffic data is obtained through training;

The type submodule is used to determine that the predicted communication network traffic data at least includes communication traffic size data, communication traffic type data, and communication traffic distribution data.

The beneficial effects are that the historical communication network traffic data in the system is obtained, the historical communication network traffic data is preprocessed to obtain the training communication network traffic data set; the GRU-GNN communication network traffic prediction model is established based on the combination of the GRU dual-channel gated recurrent unit and the GNN graph neural network to obtain the initial GRU-GNN communication network traffic prediction model; the initial GRU-GNN communication network traffic prediction model is combined with the Self-Attention self-attention mechanism to obtain the target GRU-GNN communication network traffic prediction model; the training communication network traffic data set in the system is obtained, and the training communication network traffic data set is input into the target GRU-GNN communication network traffic prediction model for prediction to obtain the predicted communication network traffic data; the first network traffic management strategy is generated according to the predicted communication network traffic data, and the wireless router flow table is configured and managed based on the first network traffic management strategy. The communication network traffic data is configured and managed using the openflow protocol to obtain the target communication network traffic data; the target communication network traffic data is monitored to obtain the data change range of the target communication network traffic data in real time, and if the data change range is greater than the set threshold, the second network traffic management strategy is generated. First, the dynamic management and optimization of network traffic can be realized. Traffic management strategies can be adjusted dynamically according to real-time network conditions and business needs, thereby achieving precise control of network traffic. Through real-time monitoring and prediction of network traffic, the location of network congestion and bottlenecks can be discovered, and then balanced distribution and efficient transmission of traffic can be achieved by adjusting routes and optimizing resource allocation. Secondly, it can improve the utilization efficiency of network resources. Through in-depth analysis of network traffic data, AI-based methods can discover redundant and wasted resources in the network, and achieve rational and maximized utilization of resources by optimizing resource allocation. Finally, it can improve user experience. By accurately controlling and managing network traffic, AI-based methods can ensure the smooth operation of key businesses, reduce network congestion and latency, and thus improve user satisfaction and experience.

BRIEF DESCRIPTION OF THE DRAWINGS

Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the following detailed description of the preferred embodiment.The drawings are only for the purpose of illustrating the preferred embodiments and are not to be construed as limiting the invention.

FIG1 is a schematic diagram of a first embodiment of a communication network traffic optimization method based on artificial intelligence in an embodiment of the present invention;

FIG2 is a schematic diagram of a second embodiment of a communication network traffic optimization method based on artificial intelligence in an embodiment of the present invention;

FIG3 is a schematic diagram of a third embodiment of a communication network traffic optimization method based on artificial intelligence in an embodiment of the present invention;

FIG4 is a schematic diagram of a first embodiment of a communication network traffic optimization system based on artificial intelligence in an embodiment of the present invention.

Detailed ways

In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

Those skilled in the art will appreciate that, unless otherwise stated, the singular forms "a", "an", "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present invention refers to the presence of the features, integers, steps, operations, elements and/or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and/or groups thereof.

The present invention is described in detail below with reference to the accompanying drawings. As shown in FIG1 , a communication network traffic optimization method based on artificial intelligence is provided. The communication network traffic optimization method comprises the following steps:

Step 101: Acquire historical communication network traffic data in the system, perform data preprocessing on the historical communication network traffic data, and obtain a training communication network traffic data set;

Specifically, in this embodiment, historical communication network traffic data in the system is obtained, and the historical communication network traffic data at least includes communication data packet size data, communication traffic type data, source and destination address data, bandwidth usage data, communication traffic distribution data, communication traffic anomaly detection data, and user behavior data obtained through the base station; data enhancement processing is performed on the historical communication network traffic data to obtain enhanced communication network traffic data; missing values in the enhanced communication network traffic data are supplemented to obtain complete communication network traffic data; outliers in the complete communication network traffic data are processed to obtain a training communication network traffic data set.

Specifically, in this embodiment, outlier processing is an important step in data preprocessing, and its purpose is to screen out outliers in the data and correct them to ensure the accuracy and reliability of the data. The outlier screening step includes the following three steps: 1) Visualize the data and use graphical methods such as scatter plots and histograms to find points with obviously abnormal values as preliminary screening results. 2) Use boxplots to find the range of outliers, and use visualizations to judge the data within this range to decide whether to mark it as an outlier. 3) Use professional knowledge and experience in the field where the data is located to determine whether the data contains outliers.

Specifically, the communication network traffic data in this embodiment mainly covers various data volumes and related information generated during network transmission. These data not only help to measure the usage and load of the network, but also provide important basis for network performance analysis, resource monitoring, structure optimization, security assurance and problem diagnosis. The communication network traffic data mainly includes the following aspects: Number and size of data packets: This reflects the number and size of data packets transmitted through the network in a specific time period. Data packets are the basic unit of network transmission, and their number and size are directly related to the load and transmission efficiency of the network. Traffic type: According to the communication protocol and application type, traffic can be divided into multiple types, such as HTTP traffic, FTP traffic, video traffic, etc. Each traffic type has different requirements and performance impacts on the network. Source and destination address: Each data packet has a clear source address and destination address, which helps to track the source and destination of the data, which is crucial for network security and traffic analysis. Bandwidth usage: Network traffic data can reflect the usage of network bandwidth, including bandwidth utilization, peak bandwidth, etc., which helps to evaluate network capacity and plan network expansion. Traffic distribution: that is, the distribution of traffic in different time periods, different locations or different users. This helps to understand the mode and characteristics of network usage and provide a basis for network optimization. Traffic anomaly detection: By analyzing traffic data, abnormal traffic can be detected, such as DDoS attacks, malware propagation, etc., to provide early warning for network security protection. User behavior analysis: By analyzing network traffic data, users' Internet habits, preferences and other information can be understood to provide support for personalized services and precision marketing.

Step 102: Combine the GRU dual-channel gated recurrent unit and the GNN graph neural network to establish a GRU-GNN communication network traffic prediction model, and obtain an initial GRU-GNN communication network traffic prediction model;

Specifically, the GRU dual-channel gated recurrent unit in this embodiment includes a reset gate and an update gate, wherein the calculation formula of the reset gate is as follows:

r _t =σ(W _r ·[h _t-1 ,X _t ]+b _r )

h _t = tanh(W _h ·[r _t *h _t-1 ,X _t ]+b _h )

Among them, σ is the sigmoid function, tanh is the activation function of the model, h _t-1 is the output state vector of the previous moment, X _t is the input vector of the current moment; W _r , W _h are the weight parameters of the model; b _r , b _h are the bias parameters of the model; the calculation formula of the update gate is as follows:

Z _t =σ(W _Z ·[h _t-1 ,X _t ]+b _Z )

Among them, W _Z , b _Z are weight parameters and bias parameters, and the update gate is used to control the influence of the output state vector h _t-1 at the previous moment and the new input h _t on the output vector h' _t at the current moment; after the reset gate and the update gate, the final output vector at the current moment t is:

h' _t =(1-Z _t )*h _t-1 +Z _t *h _t

In the GNN graph neural network, Fourier transform is used to map the spatial domain data to the frequency domain for convolution, and inverse Fourier transform is used to map the result back to the spatial domain; the GRU dual-channel gated recurrent unit and the GNN graph neural network are combined to establish a GRU-GNN communication network traffic prediction model, and the initial GRU-GNN communication network traffic prediction model is obtained.

Step 103: Combine the initial GRU-GNN communication network traffic prediction model and the Self-Attention mechanism to obtain a target GRU-GNN communication network traffic prediction model;

Specifically, this embodiment is based on the combination of the GRU-GNN communication network traffic prediction model and the Self-Attention mechanism; the calculation formula of the Self-Attention mechanism is as follows:

Among them, _TA is the temporal attention matrix, V _T , b _T , U ₁ , U ₂ and U ₃ are all matrix parameters that need to be learned in the model; the elements in _TA are normalized, and the calculation formula for normalization is as follows:

The normalized time attention matrix is multiplied by the communication network traffic data to dynamically adjust the input data of the initial GRU-GNN communication network traffic prediction model; the initial GRU-GNN communication network traffic prediction model includes at least three representation learning components; the representation learning component is used to represent the dynamic spatiotemporal information in the communication network traffic data, and the three representation learning components are adaptively fused to output the prediction results; the representation learning component is composed of 4 spatiotemporal convolution modules, and the network topology and time weight are adjusted by the self-attention layer in the spatiotemporal convolution module shown; the convolution module composed of the GNN graph neural network is used to learn the spatiotemporal features to obtain the target GRU-GNN communication network traffic prediction model.

Step 104: Obtain a training communication network traffic data set in the system, input the training communication network traffic data set into a target GRU-GNN communication network traffic prediction model for prediction, and obtain predicted communication network traffic data;

Specifically, in this embodiment, real-time communication network traffic data in the system is obtained, and the real-time communication network traffic data is input into the target GRU-GNN communication network traffic prediction model for prediction; the accuracy of the target GRU-GNN communication network traffic prediction model is evaluated using the RMSE root mean square error; the Adam optimizer is set as the optimizer for training the target GRU-GNN communication network traffic prediction model; the MSE loss function is set as the training loss function of the target GRU-GNN communication network traffic prediction model; the initial learning rate of the target GRU-GNN communication network traffic prediction model is set to 0.006, and the loss value of the validation set is judged every 32 epochs. If the loss value does not decrease, the learning rate is reduced to half of the original value, and the predicted communication network traffic data is obtained through training; the predicted communication network traffic data at least includes communication traffic size data, communication traffic type data, and communication traffic distribution data.

Step 105: Generate a first network traffic management strategy according to the predicted communication network traffic data, and based on the first network traffic management strategy, use the openflow protocol to configure and manage the communication network traffic on the wireless router flow table to obtain the target communication network traffic data;

Specifically, in this embodiment, a first network traffic management strategy is generated according to the predicted communication network traffic data, and the first network traffic management strategy at least includes a traffic priority transmission strategy, a traffic shaping strategy, a traffic control strategy, and a traffic isolation strategy; the underlying network is communicated with through the openflow protocol to obtain network device information and connect to the network device, and the openflow protocol allows the SDN controller to remotely control the SDN switch; the openflow protocol is used to configure and manage the flow table and metering table of the communication network traffic of the wireless router flow table; based on the openflow protocol, the first network traffic management strategy is sent to the network device, and the network device database is configured to execute add and delete commands to obtain the target communication network traffic data.

Specifically, the first network traffic management strategy in this embodiment includes QoS (Quality of Service) management: QoS management ensures the data transmission quality of key applications by classifying and prioritizing network traffic. Through QoS strategies, different priorities and bandwidth guarantees can be set for different traffic types (such as voice, video, data, etc.), thereby optimizing network performance and improving user experience. Traffic shaping and traffic control: Traffic shaping is a strategy that smoothes network traffic changes by limiting the transmission rate of traffic in the network. It helps to avoid network congestion caused by sudden peak traffic and ensure stable network operation. Traffic control sets traffic thresholds and restrictions based on network load levels to control user network usage rates to maintain network stability and fairness. Intelligent traffic management: Through behavior recognition technology, intelligent traffic management can perceive and analyze users' network usage behaviors, thereby performing personalized bandwidth allocation and traffic management. This strategy can provide targeted services based on user needs and usage patterns, and improve the utilization efficiency of network resources. Security management and strategy: Network security is an important aspect of network traffic management. By implementing security strategies such as firewalls, intrusion detection systems (IDS) and intrusion prevention systems (IPS), network attacks and malicious traffic can be effectively prevented to protect the security and stability of the network. Traffic optimization strategy: Traffic optimization strategies include compression transmission, caching technology, and content distribution network (CDN). These technologies can reduce the amount of data transmission, improve network transmission efficiency, and accelerate the access speed of network data, thereby improving user experience. Differentiated service strategy: Provide differentiated services such as home network optimization and streaming media acceleration based on the needs and usage patterns of different users. This strategy helps to meet the personalized needs of different users and improve user satisfaction. Traffic isolation strategy: In scenarios such as virtual private networks (VPNs), different traffic is isolated in different virtual private networks through traffic isolation strategies to avoid mutual interference and improve network performance and security.

Specifically, the openflow protocol in this embodiment includes: Network controller initiates a message: The network controller first initiates a message to send its flow table update message to the OpenFlow switch. This message encapsulates all information sent from the network controller to the OpenFlow switch, including the rules defined in the network controller and all other messages, such as updates and requests.

OpenFlow switches store flow table information: After receiving a message, the OpenFlow switch stores it in its internal flow table. A flow table is a set of rules set by a network controller that defines how the OpenFlow switch performs corresponding actions based on the results of traffic analysis. Each table entry may include the source IP address, destination IP address, protocol type, and required behaviors defined in the network controller, such as forwarding traffic to a specific port, limiting traffic speed, etc.

OpenFlow switches detect and analyze traffic: OpenFlow switches detect traffic in the network and analyze it based on stored flow table information. The switch matches the source and destination addresses of the data packet with its flow table to find the corresponding forwarding rules.

OpenFlow switch performs actions: Based on the traffic analysis results, the OpenFlow switch performs corresponding actions, such as forwarding traffic to a specific device or port. If the switch finds a match in the flow table, it executes the corresponding instructions, such as forwarding or discarding the packet. If the switch cannot find a match in the flow table, it sends a message to the controller for help. The controller can respond to the message and redefine the flow table rules to meet the needs.

Step 106: Monitor the target communication network traffic data, and obtain the data change range of the target communication network traffic data in real time. If the data change range is greater than a set threshold, generate a second network traffic management strategy.

Specifically, in this embodiment, the target communication network traffic data is monitored, and the data change range of the target communication network traffic data is obtained every 30 seconds; if the data change range of the target communication network traffic data within 30 seconds is less than 20%, the target communication network traffic data is recorded; if the data change range of the target communication network traffic data within 30 seconds is greater than 20%, the change range of the target communication network traffic data within 60 seconds is calculated; if the change range of the target communication network traffic data within 60 seconds is greater than 25%, a second network traffic management strategy is generated; based on the second network traffic management strategy, the openflow protocol is used to configure and manage the communication network traffic on the wireless router flow table.

Its beneficial effects are that, first of all, it can realize dynamic management and optimization of network traffic. According to the real-time network status and business needs, the traffic management strategy can be dynamically adjusted to achieve precise control of network traffic. Through real-time monitoring and prediction of network traffic, the location of network congestion and bottlenecks can be found, and then the balanced distribution and efficient transmission of traffic can be achieved by adjusting routes, optimizing resource allocation, etc. Secondly, it can improve the utilization efficiency of network resources. Through in-depth analysis of network traffic data, artificial intelligence-based methods can find redundant and wasted resources in the network, and realize the rational use and maximum utilization of resources by optimizing resource allocation. Finally, it can improve user experience. By accurately controlling and managing network traffic, artificial intelligence-based methods can ensure the smooth operation of key businesses, reduce network congestion and delays, and thus improve user satisfaction and experience.

In this embodiment, please refer to FIG. 2, a second embodiment of a communication network traffic optimization method based on artificial intelligence in an embodiment of the present invention generates a first network traffic management strategy according to predicted communication network traffic data, and based on the first network traffic management strategy, uses the openflow protocol to configure and manage the communication network traffic on the wireless router flow table, and obtains the target communication network traffic data, including the following steps:

Step 201: Generate a first network traffic management strategy based on predicted communication network traffic data, the first network traffic management strategy at least including a traffic priority transmission strategy, a traffic shaping strategy, a traffic control strategy, and a traffic isolation strategy;

Step 201: Communicate with the underlying network through the openflow protocol to obtain network device information and connect to the network device. The openflow protocol allows the SDN controller to control the SDN switch remotely.

Step 201: Use the openflow protocol to configure and manage the flow table and metering table of the wireless router flow table for communication network traffic;

Step 201: Send a first network traffic management policy to a network device based on the openflow protocol, configure a network device database to execute add and delete commands, and obtain target communication network traffic data.

In this embodiment, please refer to FIG. 3, a third embodiment of a communication network traffic optimization method based on artificial intelligence in an embodiment of the present invention monitors the target communication network traffic data, obtains the data change range of the target communication network traffic data in real time, and if the data change range is greater than the set threshold, generates a second network traffic management strategy including the following steps:

Step 301: Monitor the target communication network traffic data and obtain the data change range of the target communication network traffic data every 30 seconds;

Step 302: If the data variation range of the target communication network flow data is less than 20% within 30 seconds, the target communication network flow data is recorded;

Step 303: If the data variation range of the target communication network flow data within 30 seconds is greater than 20%, then the variation range of the target communication network flow data within 60 seconds is calculated;

Step 304: If the change range of the target communication network traffic data within 60 seconds is greater than 25%, a second network traffic management strategy is generated;

Step 305: Based on the second network traffic management strategy, the openflow protocol is used to configure and manage the communication network traffic on the wireless router flow table.

A communication network traffic optimization method based on artificial intelligence provided by an embodiment of the present invention is described above. A communication network traffic optimization system based on artificial intelligence according to an embodiment of the present invention is described below. Please refer to FIG. 4. An embodiment of the communication network traffic optimization system according to an embodiment of the present invention includes:

A traffic data acquisition module is used to acquire historical communication network traffic data in the system, perform data preprocessing on the historical communication network traffic data, and obtain a training communication network traffic data set;

A prediction model building module is used to combine the GRU dual-channel gated recurrent unit and the GNN graph neural network to establish a GRU-GNN communication network traffic prediction model, and obtain an initial GRU-GNN communication network traffic prediction model;

The prediction model optimization module is used to combine the initial GRU-GNN communication network traffic prediction model with the Self-Attention mechanism to obtain the target GRU-GNN communication network traffic prediction model;

The communication traffic prediction model is used to obtain a training communication network traffic data set in the system, input the training communication network traffic data set into a target GRU-GNN communication network traffic prediction model for prediction, and obtain predicted communication network traffic data;

A communication traffic management module is used to generate a first network traffic management strategy according to the predicted communication network traffic data, and based on the first network traffic management strategy, use the openflow protocol to configure and manage the communication network traffic on the wireless router flow table to obtain the target communication network traffic data;

The communication traffic optimization module is used to monitor the target communication network traffic data and obtain the data change range of the target communication network traffic data in real time. If the data change range is greater than a set threshold, a second network traffic management strategy is generated.

The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and descriptions are only preferred examples of the present invention and are not intended to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims (10)

1. The communication network flow optimization method based on artificial intelligence is characterized by comprising the following steps of:

acquiring historical communication network flow data in a system, and performing data preprocessing on the historical communication network flow data to obtain a training communication network flow data set;

Establishing a GRU-GNN communication network flow prediction model based on the GRU double-channel gating circulating unit and the GNN graph neural network in a combined way to obtain an initial GRU-GNN communication network flow prediction model;

Combining the initial GRU-GNN communication network flow prediction model with a Self-attention mechanism of Self-Attenti on to obtain a target GRU-GNN communication network flow prediction model;

Acquiring a training communication network flow data set in a system, and inputting the training communication network flow data set into the target GRU-GNN communication network flow prediction model for prediction to obtain predicted communication network flow data;

generating a first network flow management strategy according to the predicted communication network flow data, and performing configuration management on communication network flow by using a wireless router flow table by using an openflow protocol based on the first network flow management strategy to obtain target communication network flow data;

And monitoring the target communication network flow data, acquiring the data change range of the target communication network flow data in real time, and generating a second network flow management strategy if the data change range is larger than a set threshold value.

2. The method for optimizing communication network traffic based on artificial intelligence according to claim 1, wherein the acquiring historical communication network traffic data in the system, and performing data preprocessing on the historical communication network traffic data to obtain a training communication network traffic data set, comprises:

Acquiring historical communication network flow data in a system, wherein the historical communication network flow data at least comprises communication data packet size data, communication flow type data, source and target address data, bandwidth use condition data, communication flow distribution data, communication flow abnormality detection data and user behavior data which are acquired through a base station;

Performing data enhancement processing on the historical communication network traffic data to obtain enhanced communication network traffic data;

Supplementing the missing value in the enhanced communication network flow data to obtain complete communication network flow data;

and processing the abnormal value in the complete communication network flow data, and then carrying out data normalization to obtain a training communication network flow data set.

3. The communication network traffic optimization method based on artificial intelligence according to claim 1, wherein the combining and establishing of the GRU-GNN communication network traffic prediction model based on the GRU dual-channel gating circulation unit and the GNN graph neural network to obtain the initial GRU-GNN communication network traffic prediction model comprises:

The GRU double-channel gating circulating unit comprises a reset gate and an update gate, wherein the calculation formula of the reset gate is as follows:

r_t＝σ(W_r·[h_t-1,X_t]+b_r)

h_t＝tanh(W_h·[r_t*h_t-1,X_t]+b_h)

Wherein sigma is a sigmoid function, tan h is an activation function of the model, h _t-1 is an output state vector at the last moment, and X _t is an input vector at the current moment; w _r,W_h is a weight parameter of the model; b _r,b_h is the bias parameter of the model;

the calculation formula of the update gate is as follows:

Z_t＝σ(W_Z·[h_t-1,X_t]+b_Z)

Wherein, W _Z,b_Z is a weight parameter and a bias parameter, and the update gate is used for controlling the influence degree of the output state vector h _t-1 and the new input h _t at the last moment on the output vector h' _t at the current moment; the final output vector at the current time t after passing through the reset gate and the update gate is as follows:

h′_t＝(1-Z_t)*h_t-1+Z_t*h_t

In the GNN graph neural network, mapping data of a spatial domain to a frequency domain by utilizing Fourier transform for convolution, and mapping a result back to the spatial domain by utilizing inverse Fourier transform;

And combining the GRU double-channel gating circulation unit with the GNN graph neural network to establish a GRU-GNN communication network flow prediction model, so as to obtain an initial GRU-GNN communication network flow prediction model.

4. The method for optimizing communication network traffic based on artificial intelligence according to claim 1, wherein the combining the initial GRU-GNN communication network traffic prediction model and Self-Attention mechanism to obtain the target GRU-GNN communication network traffic prediction model comprises:

Based on the combination of GRU-GNN communication network flow prediction model and Self-Attention mechanism;

The Self-Attention mechanism has the following calculation formula:

Wherein T _A is a time attention matrix, and V _T,b_T,U₁,U₂ and U ₃ are matrix parameters to be learned in the model;

and carrying out normalization processing on the elements in the T _A, wherein the calculation formula of the normalization processing is as follows:

Multiplying the normalized time attention matrix by communication network flow data to dynamically adjust the input data of the initial GRU-GNN communication network flow prediction model; the initial GRU-GNN communication network flow prediction model at least comprises three characterization learning components;

the representation learning component is used for carrying out representation learning on dynamic space-time information in the communication network flow data, and three representation learning components output a prediction result after self-adaptive fusion;

The characterization learning component consists of 4 space-time convolution modules, wherein the self-attention layer adjusts the network topology and the time weight in the space-time convolution modules;

And learning the time-space characteristics by using a convolution module formed by the GNN graph neural network to obtain a target GRU-GNN communication network flow prediction model.

5. The communication network traffic optimization method based on artificial intelligence according to claim 1, wherein the acquiring real-time communication network traffic data in the system, inputting the real-time communication network traffic data into the target GRU-GNN communication network traffic prediction model for prediction, and obtaining predicted communication network traffic data, comprises:

Acquiring real-time communication network flow data in a system, and inputting the real-time communication network flow data into the target GRU-GNN communication network flow prediction model for prediction;

Estimating the accuracy of the target GRU-GNN communication network flow prediction model by using RMSE root mean square error;

Setting an Adam optimizer as a trained optimizer of the target GRU-GNN communication network traffic prediction model;

Setting an MSE loss function as a training loss function of the target GRU-GNN communication network flow prediction model;

Setting the initial learning rate of the target GRU-GNN communication network flow prediction model to be 0.006, judging the loss value of a verification set once every 32 epochs, if the loss value does not decrease, reducing the learning rate to one half of the original value, and training to obtain predicted communication network flow data;

the predicted communication network traffic data includes at least communication traffic size data, communication traffic type data, and communication traffic distribution data.

6. The communication network traffic optimization method based on artificial intelligence according to claim 1, wherein the generating a first network traffic management policy according to the predicted communication network traffic data, based on the first network traffic management policy, performing configuration management of communication network traffic on a wireless router flow table by using openflow protocol to obtain target communication network traffic data, comprises:

generating a first network traffic management policy according to the predicted communication network traffic data, wherein the first network traffic management policy at least comprises a traffic priority transmission policy, a traffic shaping policy, a traffic control policy and a traffic isolation policy;

the method comprises the steps that network equipment information is obtained through communication of a bottom layer network through an openflow protocol, the openflow protocol allows an SDN controller to control an SDN switch in a remote mode, and the network equipment information is connected to the network equipment;

Performing configuration management on a flow table and a meter of communication network flow by using an openflow protocol;

And sending the first network flow management strategy to network equipment based on an openflow protocol, and configuring a network equipment database to execute the adding and deleting commands to obtain the target communication network flow data.

7. The method for optimizing communication network traffic based on artificial intelligence according to claim 1, wherein monitoring the target communication network traffic data, acquiring a data change range of the target communication network traffic data in real time, and if the data change range is greater than a set threshold, generating a second network traffic management policy, comprises:

monitoring the target communication network flow data, and acquiring a data change range of the target communication network flow data every 30 seconds;

If the data change range of the target communication network flow data is smaller than 20% within 30 seconds, recording the target communication network flow data;

If the data change range of the target communication network flow data is larger than 20% in 30 seconds, calculating the change range of the target communication network flow data in 60 seconds;

If the change range of the target communication network flow data in 60 seconds is greater than 25%, generating a second network flow management strategy;

And based on the second network flow management strategy, performing configuration management on the communication network flow by using an openflow protocol on a wireless router flow table.

8. An artificial intelligence-based communication network traffic optimization system, characterized in that the communication network traffic optimization system comprises the following modules:

the flow data acquisition module is used for acquiring historical communication network flow data in the system, and carrying out data preprocessing on the historical communication network flow data to obtain a training communication network flow data set;

The prediction model building module is used for combining the GRU double-channel gating circulating unit with the GNN graph neural network to build a GRU-GNN communication network flow prediction model so as to obtain an initial GRU-GNN communication network flow prediction model;

the prediction model optimization module is used for combining the initial GRU-GNN communication network flow prediction model and a Self-attention mechanism of Self-Attenti on to obtain a target GRU-GNN communication network flow prediction model;

The communication flow prediction model is used for acquiring a training communication network flow data set in the system, inputting the training communication network flow data set into the target GRU-GNN communication network flow prediction model for prediction, and obtaining predicted communication network flow data;

The communication flow management module is used for generating a first network flow management strategy according to the predicted communication network flow data, and carrying out configuration management on the communication network flow of the wireless router flow table by utilizing an openflow protocol based on the first network flow management strategy to obtain target communication network flow data;

And the communication flow optimization module is used for monitoring the target communication network flow data, acquiring the data change range of the target communication network flow data in real time, and generating a second network flow management strategy if the data change range is larger than a set threshold value.

9. The artificial intelligence based communication network traffic optimization system of claim 8, wherein the traffic data acquisition module comprises the following sub-modules:

The system comprises an acquisition sub-module, a communication flow detection module and a communication flow detection module, wherein the acquisition sub-module is used for acquiring historical communication network flow data in the system, and the historical communication network flow data at least comprises communication data packet size data, communication flow type data, source and target address data, bandwidth use condition data, communication flow distribution data, communication flow abnormality detection data and user behavior data which are acquired through a base station;

The enhancer module is used for carrying out data enhancement processing on the historical communication network traffic data to obtain enhanced communication network traffic data;

The supplementing sub-module is used for supplementing the missing value in the enhanced communication network flow data to obtain complete communication network flow data;

And the obtaining submodule is used for carrying out data normalization after processing the abnormal value in the complete communication network flow data to obtain a training communication network flow data set.

10. The artificial intelligence based communication network traffic optimization system of claim 8, wherein the communication traffic prediction model comprises the following sub-modules:

The input sub-module is used for acquiring real-time communication network flow data in the system, and inputting the real-time communication network flow data into the target GRU-GNN communication network flow prediction model for prediction;

The evaluation sub-module is used for evaluating the accuracy of the target GRU-GNN communication network flow prediction model by using RMSE root mean square error;

An optimizer sub-module for setting an Adam optimizer as a trained optimizer of the target GRU-GNN communication network traffic prediction model;

the loss function sub-module is used for setting the MSE loss function as a training loss function of the target GRU-GNN communication network flow prediction model;

the learning rate sub-module is used for setting the initial learning rate of the target GRU-GNN communication network flow prediction model to be 0.006, judging the loss value of a verification set once every 32 epochs, and if the loss value is not reduced, reducing the learning rate to one half of the original learning rate, and training to obtain predicted communication network flow data;

And the type sub-module is used for determining that the predicted communication network traffic data at least comprises communication traffic size data, communication traffic type data and communication traffic distribution data.