Twin network performance prediction method and device based on graph neural network, and storage medium

Abstract

The invention discloses a twin network performance prediction method and device based on a graph neural network and a storage medium, and belongs to the technical field of communication networks, and the method comprises the steps: carrying out the preprocessing of the obtained historical information of a physical communication network, and obtaining a dependency graph data set; initializing the state of each node in the dependency graph data set to obtain the initial hidden state of each node; the initial hidden state is used to train a pre-constructed network performance prediction model to obtain a trained network performance prediction model, and the network performance prediction model obtains iteratively updated hidden state information and adjusts model parameters through cooperative work of an attention mechanism, a message passing neural network and a gating circulation unit; and inputting the obtained real-time information of the physical communication network into the trained network performance prediction model to obtain a performance prediction result, thereby improving the stability, accuracy and efficiency of performance prediction of the digital twin communication network.

Description

Technical Field

[0001] The present invention belongs to the field of communication network technology, and in particular relates to a twin network performance prediction method, device and storage medium based on graph neural network. Background Art

[0002] In recent years, the field of communications has been continuously innovating in technology, and new network paradigms such as software-defined networks (SDN) and digital twin networks (DTN) have emerged, promoting the transformation and upgrading of traditional networks. Among them, DTN is a network system that creates virtual twins of physical network entities in a digital way, and can interact and map with physical network entities in real time to expand the functions of physical networks. At present, the digital twinning of communication networks is in the exploratory stage. However, due to the dynamic and changeable nature of communication networks at the physical level and the great influence of the environment, their internal structure and data are complex, which increases the difficulty of predicting network performance and ultimately makes it difficult to achieve digital twins.

[0003] Graph neural network GNN has a powerful ability to process graph structured data, which further enhances the accuracy of performance prediction and shows great application potential in digital twins of communication networks. At present, many methods have emerged in the research of network performance prediction through graph neural network at home and abroad.

[0004] In the prior art, there is a method that uses graph spatial domain convolution to process the message passing process of graph network nodes, and performs relationship reasoning between network information, which realizes accurate prediction of network performance including latency, jitter and packet loss rate, but the features extracted by the model are relatively one-sided and the generalization performance is poor; in the prior art, there is also a method that trains a new lightweight graph network GLN in a federated learning manner to predict the network performance indicators of the slice network, which improves the efficiency of model training, but there is still room for further improvement in the balance between model performance and training efficiency. In summary, the existing network performance prediction technology has limitations when dealing with dynamic and changing network environments, and it is difficult to meet the high-fidelity and low-latency requirements of digital twins. Summary of the invention

[0005] The present invention provides a twin network performance prediction method, device and storage medium based on graph neural network, which selects the best features by performing feature selection and feature integration on training set data, and improves the stability of the hidden state update process of graph structure data by combining attention mechanism, GRU unit and message passing neural network MPNN, thereby improving the stability, accuracy and efficiency of communication network performance prediction to adapt to rapidly changing communication service needs.

[0006] To achieve the above object, the present invention is implemented by adopting the following technical solutions:

[0007] In a first aspect, the present invention provides a twin network performance prediction method based on a graph neural network, comprising:

[0008] Obtain real-time relevant network information of the physical communication network, input it into the trained network performance prediction model, and obtain performance prediction results;

[0009] The process of determining the trained network performance prediction model includes:

[0010] Preprocessing the historical network information related to the physical communication network obtained by the digital twin communication network to obtain a dependency graph data set, wherein the dependency graph data set includes link, path and queue information;

[0011] Initializing the states of each link, path, and queue in the dependency graph data set to obtain the initial hidden state of each link, path, and queue;

[0012] The pre-built network performance prediction model is trained using the initial hidden state to obtain a trained network performance prediction model, wherein the network performance prediction model obtains iteratively updated hidden state information through the collaborative work of an attention mechanism, a message passing neural network, and a gated recurrent unit, and adjusts model parameters according to the hidden state information.

[0013] Optionally, the physical communication network history-related network information acquired by the digital twin communication network is preprocessed to obtain a dependency graph data set, including:

[0014] Normalize the read original network samples to obtain a standardized data set;

[0015] The original network samples in the standardized data set are converted into a network dependency graph to obtain a network dependency graph data set, wherein the network dependency graph includes a set of links , a set of queues on the output ports of a network device and a set of source-to-destination paths ;

[0016] , where Represents the first link in a group of links Links, Indicates the number of links in a set of links;

[0017] , where Represents the first queues, Indicates the number of queues in a group of queues;

[0018] , where Represents the first Path, Represents the number of paths in a set of paths, represented by a series of tuples , , represent The number of elements in Represents the first queues, Represents the first A link.

[0019] Optionally, the initializing the states of each link, path and queue in the dependency graph data set to obtain the initial hidden state of each link, path and queue includes:

[0020] Path state initialization: Create a first state vector by connecting traffic characteristics, packet number characteristics, and zero padding , where the traffic feature represents the data traffic passing through the path, the packet number feature represents the number of packets generated on the path, and zero padding is used to form a feature vector of a predetermined dimension;

[0021] And / or, Link State Initialization: Create a second state vector by connecting capacity characteristics, policy characteristics, and zero padding , where the capacity feature represents the bandwidth capacity of the link, the policy feature represents the queue scheduling policy on the link, and zero padding is used to meet the dimensionality requirements;

[0022] And / or, queue state initialization: create a third state vector by concatenating the size feature, priority feature, weight feature, and zero padding , where the size feature represents the current number of packets in the queue, the priority feature represents the packet priority, the weight feature represents the queue weight associated with the scheduling policy, and zero padding is used to meet the dimensionality requirement.

[0023] Optionally, the pre-built network performance prediction model includes: a GRUCell gated recurrent unit layer, an Attention layer, a message passing loop module and a Readout layer connected in sequence; the message passing loop module includes an MPNN message passing neural network layer and an RNN recurrent neural network layer connected to each other.

[0024] Optionally, the using the initial hidden state to train a pre-built network performance prediction model to obtain a trained network performance prediction model includes:

[0025] A path cycle unit, a link cycle unit and a queue cycle unit are defined respectively through the GRUCell gated cycle unit layer;

[0026] In combination with the Attention layer, the MPNN message passing neural network layer and the RNN recurrent neural network layer, the hidden state information of the path cycle unit, the link cycle unit and the queue cycle unit is iteratively updated from the initial hidden state through the message passing process to obtain the global state of the physical communication network;

[0027] Outputting the performance prediction value according to the updated path hidden state through the Readout readout layer;

[0028] The network performance prediction model is trained according to the mean square error MSE between the performance prediction value and the true value, and the model hyperparameters are set, wherein the training objective of the network performance prediction model is It is expressed as:

[0029] ,

[0030] Where: is the actual performance value of the i-th dependency graph sample, is the performance prediction value of the i-th dependency graph sample, is the number of samples in the dependency graph dataset, and i represents the sequence number of the sample in the dependency graph dataset.

[0031] Optionally, the method further comprises: selecting a test set from the dependency graph data set, inputting the test set into the network performance prediction model generated by each n rounds of training and the last round of training, and calculating the mean absolute error MAE, mean absolute percentage error MAPE and correlation coefficient between the predicted value and the true value of the network performance prediction model Evaluate the model training effect and find better hyperparameters to obtain the best network performance prediction model, where:

[0032] Calculate the mean absolute error (MAE), mean absolute percentage error (MAPE), and correlation coefficient between the predicted value and the true value of the network performance prediction model , the calculation formula is:

[0033] ,

[0034] ,

[0035] ,

[0036] Where: is the average of the actual performance values ​​of n dependency graph samples.

[0037] Optionally, the iterative updating of the hidden state information encoded by the path cycle unit, the link cycle unit and the queue cycle unit through a message passing process includes: in each iteration, the network performance prediction model performs the following three stages of message passing:

[0038] Queue and link to path: collects information about the queues and links contained in each path, and hides the status of the links at a fixed time. and queue hidden state Compute the temporary state generated by the attention mechanism ,according to and the path hidden state at that moment As input, the GRU unit in the RNN recurrent neural network layer is used to update the hidden state of the new path and the new path temporarily hides the state ;

[0039] Path to Queue: aggregates temporary hidden state information of the queue and all paths containing the queue , temporarily hide all status information After summing, we get the sum of temporary hidden state information of the queue. , based on the sum of the temporarily hidden state information of the queue and queue hidden status The new queue hidden state is obtained by updating the GRU unit in the RNN recurrent neural network and the new queue temporarily hides the state ;

[0040] Queue to link: Temporarily hide state based on new queue and link hidden status , the new link hidden state is obtained by updating the GRU unit in the RNN recurrent neural network .

[0041] In a second aspect, the present invention provides a twin network performance prediction device based on a graph neural network, comprising:

[0042] Performance prediction result acquisition module: used to obtain real-time relevant network information of the physical communication network, input it into the trained network performance prediction model, and obtain the performance prediction result;

[0043] The process of determining the trained network performance prediction model includes:

[0044] Preprocessing the historical network information related to the physical communication network obtained by the digital twin communication network to obtain a dependency graph data set, wherein the dependency graph data set includes link, path and queue information;

[0045] Initializing the features of each link, path and queue state in the dependency graph data set to obtain the initial hidden state of each link, path and queue;

[0046] The pre-built network performance prediction model is trained using the initial hidden state to obtain a trained network performance prediction model, wherein the network performance prediction model obtains iteratively updated hidden state information through the collaborative work of an attention mechanism, a message passing neural network, and a gated recurrent unit, and adjusts model parameters according to the hidden state information.

[0047] In a third aspect, the present invention provides a computer storage medium having a computer program stored thereon, which, when executed by a processor, implements the twin network performance prediction method based on a graph neural network as described in any step of the first aspect.

[0048] In a fourth aspect, the present invention provides an electronic terminal comprising a processor and a memory connected to the processor, wherein a computer program is stored in the memory. When the computer program is executed by the processor, the twin network performance prediction method based on graph neural network as described in any step of the first aspect is executed.

[0049] Compared with the prior art, the beneficial effects achieved by the present invention are as follows: by preprocessing the data collected by the digital twin communication network system, a suitable data set is obtained for the training and evaluation of the network performance prediction model, and the node features of the graph structure data are initialized, mainly feature selection and feature integration are performed, the best features are selected, and then a graph neural network GNN model including an attention mechanism, a GRU unit, and a message passing neural network MPNN is constructed to improve the stability of the hidden state update process of the graph structure data, and at the same time improve the generalization performance of the model. Under the dynamic and changeable conditions of the physical communication network, the stability, accuracy and efficiency of the performance prediction of the digital twin communication network can be improved; in the process of model training, an optimizer is used to efficiently optimize the model parameters and prevent overfitting, and then the hyperparameter tuning technology is used to evaluate the training effect of the model, and the hyperparameters are optimized in time, which can further improve the model prediction accuracy and reduce the running time. The network performance prediction model proposed in the present invention shows higher accuracy than traditional models and more advanced models when processing the same data set, and also shows good generalization performance when processing various types of test sets. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 Shown is a flow chart of a method for predicting performance of a digital twin communication network in Embodiment 1 of the present invention;

[0051] Figure 2 Shown is a flow chart of a method for predicting performance of a digital twin communication network in Embodiment 2 of the present invention;

[0052] Figure 3 Shown is a digital twin communication network system architecture diagram in one embodiment of the present invention;

[0053] Figure 4 The figure shows a schematic diagram of the internal structure of a network performance prediction model in an embodiment of the present invention;

[0054] Figure 5 The figure shows a schematic diagram of the change of loss value in the training and verification process in one embodiment of the present invention;

[0055] Figure 6 The figure is a schematic diagram of the comparison results of the cumulative distribution functions of the relative errors of different test sets in one embodiment of the present invention;

[0056] Figure 7 The figure shows a schematic diagram of the comparison of prediction performance of different models in an embodiment of the present invention;

[0057] Figure 8 Shown is a diagram of the internal structure of a computer device in an embodiment of the present invention. DETAILED DESCRIPTION

[0058] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present invention.

[0059] Example 1

[0060] like Figure 1 As shown, this embodiment provides a twin network performance prediction method based on a graph neural network, including:

[0061] Obtain real-time relevant network information of the physical communication network, input it into the trained network performance prediction model, and obtain performance prediction results;

[0062] The process of determining the trained network performance prediction model includes:

[0063] Preprocessing the historical network information related to the physical communication network obtained by the digital twin communication network to obtain a dependency graph data set, wherein the dependency graph data set includes link, path and queue information;

[0064] Initializing the states of each link, path, and queue in the dependency graph data set to obtain the initial hidden state of each link, path, and queue;

[0065] The pre-built network performance prediction model is trained using the initial hidden state to obtain a trained network performance prediction model, wherein the network performance prediction model obtains iteratively updated hidden state information through the collaborative work of an attention mechanism, a message passing neural network, and a gated recurrent unit, and adjusts model parameters according to the hidden state information.

[0066] By further selecting and integrating features of the graph structure data in the historical data set, the data that has a greater impact on the final result is removed to obtain a more refined training set. The network performance prediction model that includes the attention mechanism, message passing neural network, and gated recurrent unit is trained using the feature-initialized training set. This can improve the stability of the hidden state update process of the graph structure data and make the output prediction results more accurate, thereby improving the stability and accuracy of the performance prediction of the digital twin communication network.

[0067] Example 2

[0068] On the basis of Example 1, this embodiment also makes the following design: Figure 2 As shown, the digital twin communication network performance prediction method is divided into the following steps.

[0069] Step 1: Build the digital twin communication network system architecture, using the historical data obtained by the data acquisition and transmission module as the data set for training the network performance prediction model;

[0070] Step 2: Get the original network data set, process the original network data into a dependency graph structure through Tensorflow and networkx and other dependent libraries, and perform normalization preprocessing on the data;

[0071] Step 3: Initialize node features, define node attributes, select features in the data set that have a greater impact on model performance, and construct a feature matrix through feature integration as the input of the network performance prediction model to improve operating efficiency;

[0072] Step 4: Construct a graph neural network (GNN) model as a network performance prediction model and apply the model to the performance prediction of the digital twin communication network;

[0073] Step 5: Train the graph neural network GNN model, define the loss function and optimizer and related hyperparameters.

[0074] Step 6: Evaluate the trained graph neural network GNN model and select the mean square error (MSE), mean absolute error (MAE), mean absolute percentage error (MAPE), and correlation coefficient between the predicted value and the true value. As a performance indicator for evaluating the prediction performance of the algorithm. Based on the evaluation results of the model, the hyperparameter tuning technology is used to further optimize the parameters of the model, and the optimal model is selected for use in the digital twin communication network performance prediction task;

[0075] Step 7: The data information collected by the digital twin system in real time is transmitted in sequence to the optimal performance prediction module, thereby realizing real-time prediction of intelligent communication network performance.

[0076] In step 1, the present invention first constructs a digital twin communication network system architecture.

[0077] Digital twin communication network system architecture Figure 3 As shown in the figure. The whole system consists of three layers, namely the physical network layer, the twin network layer and the network application layer. The physical network layer is similar to the data plane of the software-defined network SDN, and contains the actual information of the network such as network topology, traffic, device status, etc. The twin network layer contains a data sharing warehouse for storing and managing the data collected from the physical network layer, a twin network model for digital modeling of the physical network, and algorithm models such as prediction algorithms and optimization algorithms, as well as network control components. The network application layer is the top layer of the system, which includes the specific application of the model and is responsible for the overall management and innovation of the network. The system uses sensors, communication protocols, etc. to collect relevant status information of the physical network, properly handles massive redundant data through data transmission modules and data processing modules, and saves the processed data to the shared database of the twin layer. The network performance prediction model obtains historical data from the shared database as a data set for model training, and lays the foundation for further network optimization through performance prediction.

[0078] In step 2, although the collected data set contains relevant network information, it is still not a strict graph structure data and cannot be directly used for model training, so further processing is required. Use tensorflow TensorFlow and network library networkx and other dependent libraries to read the original network sample and convert the original network graph into a directed graph to represent the dependency relationship of the network and form strict graph structure data. At the same time, due to the large difference in the values ​​of different features in the data set, some values ​​need to be normalized. Normalization is a mathematical method to simplify the amount of calculation, that is, to transform a dimensional expression into a dimensionless expression after processing and transformation, and become a scalar. The data after normalization can eliminate the influence of different indicator units and their numerical orders of magnitude, reduce evaluation errors, and speed up model training. Therefore, before training, each indicator of the selected training set data is normalized. The processing method is as follows:

[0079] ,

[0080] Where: is the normalized i-th feature data, is the i-th original feature data, and are the maximum and minimum values ​​in the original data set respectively, and i is.

[0081] At the same time, the network dependency graph is defined as a set of links and queues on output ports of network devices and a set of source-to-destination paths Composition, among which, , , , Represents the first link in a group of links Links, represents the number of links in a set of links, Represents the first queues, Indicates the number of queues in a group of queues, Represents the first Path, Indicates the number of paths in a set of paths. In addition, a path can be viewed as a sequence of multiple tuples, each tuple consisting of a queue on a port and the subsequent links, which is used to represent all queues and links contained in the path. Therefore, the path It can be further defined as: ,in, Indicates the path The number of tuples in Represents the first queues, Represents the first A link.

[0082] Finally, use the tf.data.Dataset.from_generator dependency library of TensorFlow to create a TensorFlow dataset, shuffle the new dataset, and selectively filter topological structures of a specific size.

[0083] In step 3, the sample feature dimensions contained in the network dataset used are too many, including some redundant features that have little impact on the final result. Therefore, node feature initialization is performed first, which helps to reduce the complexity of the feature space, thereby reducing the running time and improving the model efficiency.

[0084] Node feature initialization is the process of setting an initial feature representation for each node in the graph before the graph neural network GNN model starts training. This feature representation is usually a high-dimensional vector containing various attributes and status information related to the node. In step 2, the composition of the newly generated dependency graph dataset has been defined, including link, path and queue information, and these three objects contain many features such as data packets, traffic, priority, etc., based on which node feature initialization is performed.

[0085] The specific node feature initialization method is as follows:

[0086] (1) Path state path_state initialization, create a 32-dimensional state vector by connecting the traffic feature, the number of packets feature, and zero padding tf.zeros , where the traffic feature represents the data flow through the path, the packet number feature represents the number of packets generated on the path, and zero padding is used to form a feature vector of a predetermined dimension.

[0087] (2) Link state link_state initialization, creating a 32-dimensional feature vector by connecting the capacity feature, the policy feature, and the zero padding tf.zeros , where the capacity feature represents the bandwidth capacity of the link, and the policy feature uses one-hot encoding to represent the queue scheduling policy on the link, such as WFQ weighted fair queue, SP strict priority, DRR deficit polling, FIFO first-in first-out, etc., and finally a tf.zeros zero padding is used to meet the dimension requirements.

[0088] (3) Queue state queue_state is initialized, creating a 32-dimensional feature vector by connecting the size feature, priority feature, weight feature, and zero padding tf.zeros , where the size feature represents the current number of packets in the queue, the priority feature is the packet priority represented by one-hot encoding, the weight feature represents the queue weight associated with the scheduling policy, and zero padding satisfies the dimensionality requirement.

[0089] (4) State vector concatenation: Use the connection function tf.concat of TensorFlow to concatenate the above feature vectors to form a complete node state representation and complete the node state initialization.

[0090] In step 4, a complete graph neural network (GNN) model is constructed as a network performance prediction model. Figure 4The complete GNN model consists of a multi-layer neural network, including a gated recurrent unit layer GRUCell, an attention layer Attention, a message passing loop module and a readout layer Readout connected in sequence, wherein the message passing loop module includes a message passing neural network MPNN layer and a recurrent neural network RNN ​​layer connected to each other, which are used to execute and iterate the message passing step.

[0091] In this embodiment, the graph neural network GNN model also includes an Initialization layer. The initialization layer actually includes the node state initialization process in step 3. It is a preprocessing step before the model starts training, that is, the node state initialization function is integrated into the graph neural network GNN model. The initialization layer uses the call method to perform the forward propagation of the model. It accepts the input data of the model and returns the output of the model.

[0092] The gated recurrent unit layer is used to define the gated recurrent unit GRUCell, including three GRUCell instances: custom path update self.path_update, custom link update self.link_update, and custom queue update self.queue_update. GRUCell is the basic unit of the gated recurrent unit GRU, which is equivalent to a GRU with half a time step, but GRUCell provides higher flexibility and allows custom loop logic, so GRUCell is used instead of GRU. GRUCell contains an update gate and a reset gate. These two gating mechanisms allow the model to control the flow of information at each time step. The formulas of its components are expressed as follows:

[0093] Update Gate:

[0094] ,

[0095] Where: is the output of the update gate at time step t, is the update gate weight matrix, is the previous hidden state and the current time step The splicing, is the sigmoid activation function.

[0096] Reset the gate:

[0097] ,

[0098] Where: is the output of the reset gate at time step t, is the reset gate weight matrix.

[0099] Candidate hidden states:

[0100] ,

[0101] Where: is a candidate hidden state, is the weight matrix, is the reset gate output and the previous hidden state The element-wise product of .

[0102] Final hidden state:

[0103] ,

[0104] Where: is the final hidden state at time step t.

[0105] Here we first define the GRUCell instance, which will then be integrated into the message passing process.

[0106] The attention layer is used for custom attention calculation. Through the attention function, a fully connected dense layer Dense is used to calculate the input attention score, and then the softmax function is applied to obtain the standardized attention weight.

[0107] The calculation formula of attention weight is as follows:

[0108] ,

[0109] ,

[0110] Where: represents the input matrix, is the bias term, tanh is the hyperbolic tangent function used to introduce nonlinearity, scores is the obtained attention score, softmax is the normalization function, axis equals 1, which means that the scores are normalized along the second dimension of the sequence, and attention_weights is the final normalized attention weight.

[0111] In the message passing loop module, the MPNN message passing neural network layer and the RNN recurrent neural network layer are fused to process graph structure data and time series data. MPNN aggregates the neighbor information of nodes in the graph through the message passing step, collects features from neighbor nodes during the iterative process of message passing, and continuously updates the status information. At the same time, dynamically changing graph data is also a kind of time series data. The GNN model uses the GRU unit in the RNN to memorize historical information, thereby processing node features with time attributes. In the message passing process of the graph neural network, the following principles are mainly followed:

[0112] (1) The state of a path depends on the state of all queues and links traversed by the path.

[0113] (2) The state of a link depends on the state of all queues that inject traffic into the link.

[0114] (3) The state of a queue depends on the state of all paths that inject traffic into the queue.

[0115] The above principles can be expressed mathematically as follows:

[0116] ,

[0117] Where: , and are some unknown functions related to queue status, link status and path status respectively, Representation Queue The hidden state of Indicates link The hidden state of Indicates the path The hidden state of It is a link The set of queues into which traffic is injected, Indicates the path The following is a more detailed description of the state update function and update process in the message passing process of this model.

[0118] First, the node hidden state is initialized in step 3, and the output includes , and Three hidden states, and then start the message passing phase through the message passing loop module, refer to Figure 4 , in the message passing process, there are T iterations, and in each iteration, the model performs three stages of message passing:

[0119] (1) Queue and link to path: Each path collects information about the queues and links it passes through, and hides the status of the links at the time of passing through. and queue hidden state Compute the temporary state generated by the attention mechanism , and then and the path hidden state at that moment As input, the recurrent gated unit GRUCell is used to update the temporary hidden state of the new path And the new path hidden state .

[0120] (2) Path to queue: Temporary hidden status information of the aggregated queue and all paths through the queue , for all After summing, we get the sum of temporary hidden state information of the queue. , and then according to and queue hidden status The new queue hidden state is updated by the recurrent gating unit GRUCell And the new queue temporarily hidden state .

[0121] (3) Queue to link: Temporarily hide the state according to the new queue and link hidden status , the new link hidden state is updated through the recurrent gating unit GRUCell .

[0122] It can be seen that in the process of message transmission, the state of the path depends on the queues and links it passes through, and the states of the queues and links depend on the paths that contain them. Therefore, there is a circular dependency. Through multiple iterations, the state of each component gradually stabilizes and eventually reflects the global state of the entire network.

[0123] The readout layer of the GNN model is actually a multi-layer perceptron MLP, which consists of multiple fully connected layers. The first layer is the input layer, which is defined using the tensorflow modular input layer tf.keras.layers.Input, a dependency library in tensorflow. The shape of the input data is specified as 32. The second and third layers are two fully connected dense layers with the same number of units. The SELU activation function is used and applied Regularization is used to prevent the model from overfitting. The fourth layer is also a dense layer, but without an activation function, which is used to directly output the predicted value. The entire Readout layer uses the tensor flow modular sequence model tf.keras.Sequential to build a linear stacked neural network, which outputs the predicted value of the performance after a series of dense layers and nonlinear activation functions according to the path state.

[0124] In step 5, train the designed graph neural network GNN model. Define the model's loss function as the mean square error (MSE) between the predicted value and the true value, set the optimizer and hyperparameters, and minimize the mean square error as the training goal. The calculation formula for the mean square error (MSE) is as follows:

[0125] ,

[0126] Where: is the actual performance value of the i-th dependency graph sample, is the performance prediction value of the i-th dependency graph sample, and n is the number of samples in the dependency graph dataset.

[0127] In step 6, the trained graph neural network GNN model is evaluated, and the model hyperparameters are further tuned according to the evaluation results to ensure the best performance of the trained model. The mean square error MSE, mean absolute error MAE, mean absolute percentage error MAPE and correlation coefficient between the predicted value and the true value of the communication network performance parameters are selected. As performance indicators for evaluating model prediction performance. Among them, MAE, MAPE and The calculation formula is as follows:

[0128] ,

[0129] ,

[0130] ,

[0131] Where: is the average of the actual performance values ​​of n dependency graph samples. MAE and MAPE reflect the deviation between the predicted value and the true value. MAE is insensitive to outliers and is suitable for robust error measurement, while MAPE is suitable for comparing data sets of different sizes. It is the goodness of fit of the model. The closer its value is to 1, the stronger the explanatory power of the model. After the model is trained, it is necessary to use the evaluation set or test set to evaluate the training effect of the model, determine whether the training is overfitting, and find better hyperparameters, and then retrain the model to finally obtain the model with the best prediction performance.

[0132] In step 7, the data information collected by the digital twin system in real time is transmitted to the performance prediction module in sequence, thereby realizing efficient and accurate prediction of the performance of the intelligent communication network.

[0133] In this embodiment, the present invention uses tensor flow TensorFlow as a deep learning framework to develop and implement the proposed twin network performance prediction model based on graph neural network. In order to verify the constructed model, a network open source data set generated by a network simulator is selected. This example uses a data set generated by the OMNet++ network simulator. It contains three data sets. The first is an Nsfnet data set of about 20,000 samples consisting of a 14-node topology communication network and a 24-node topology communication network. The second is a Geant data set of about 1,000 samples consisting of a 22-node topology communication network. The third is an Abilene mixed data set of about 1,000 samples consisting of multiple real-world network topologies and real traffic matrices. All data sets contain network topology, routing configuration, partial network configuration (traffic matrix, routing, queue scheduling strategy) and some measurable network performance parameters (such as end-to-end delay) and other related network information.

[0134] In the experiment, the end-to-end delay of the data packet from the source to the destination is mainly used as the network performance prediction target. The Nsfnet dataset is selected for training, and 20,000 samples are divided into the train set, the eval set, and the test set in a ratio of 8:1:1. The training optimizer uses the adaptive moment estimation algorithm Adam and sets the initial learning rate to 0.001, the number of iterations to 400,000 steps, the decay rate of the learning rate to 0.7 every 40,000 steps, the batch size of the training data to 32, and the readout layer The regularization coefficient is set to 0.001, and a checkpoint is saved every 20 minutes during training and validated on the validation set. Figure 5 Describes the change in the mean square error between the true value and the predicted value in the loss function on the training and validation sets during model training, which is given by Figure 5 It can be seen that as the number of training steps increases, the mean prediction error of the model on the validation set becomes smaller and smaller, and gradually converges, and the final mean square error stabilizes at around 0.008. At the same time, the loss function on the training set and the validation set is reduced synchronously, and there is no overfitting phenomenon.

[0135] Figure 6 The results of testing the trained GNN model on different test sets are described, mainly the cumulative distribution function (CDF) graph of the relative error (MRE) between the predicted value and the true value. The calculation formula of MRE is:

[0136] .

[0137] CDF represents the probability that a random variable is less than or equal to a certain value, and the formula is expressed as:

[0138] ,

[0139] Where: is the value of the cumulative distribution function, is a random variable, is a possible value of a random variable. In this experiment, the relative error between the predicted value and the true value is used as the random variable. The cumulative distribution function CDF is the sum of discrete variables and presents a trapezoidal structure. The more the trapezoidal line is concentrated to the center point, that is, the relative error is concentrated near 0, the better the prediction effect is. Figure 6 It can be seen that the model has good test results for each data set. Among them, the Nsfnet data set is a topology type seen during training, and the effect is the best. The Geant data set has a larger network scale than the training set, and the test effect is still good. The Abilene data set contains a variety of real network topologies and real traffic matrices. The test results on this data set are slightly lower than the first two. In general, the model shows good generalization performance. The specific experimental results of other evaluation indicators on the three data sets are shown in Table 1.

[0140] Table 1 Performance of the model on different test sets

[0141] Dataset MAE MSE MAPE r² / % Nsfnet 0.0482 0.0099 0.0251 99.47 Geant 0.0702 0.0126 0.0302 98.32 Abilene 0.0712 0.0093 0.0351 97.12

[0142] For further verification, the same Nsfnet dataset is selected as the training set and the Abilene dataset is selected as the test set. The model proposed in this invention is called the gated recurrent-attention GRU-AT model, and compared with the multi-layer perceptron MLP model and the routing network RouteNet model. Figure 7 The comparison of network performance prediction results of different models is described. Among them, the MLP model consists of a four-layer network (one input layer, one output layer and two hidden layers) with nonlinear activation nodes. The traffic matrix and queue and link configuration information are used as inputs to model the end-to-end delay through the multi-layer perceptron MLP. The RouteNet model is similar to the model of the present invention. It also uses a graph neural network model to perform network modeling by performing two-stage message passing between links and paths in the network.

[0143] The specific experimental results of the three network performance prediction models are shown in Table 2. Since the network-related information of the test set, including topology and traffic matrix, is completely different from that of the training set, the prediction errors of the MLP model and the RouteNet model are large, indicating poor generalization performance, while the GRU-AT model can still maintain high accuracy. In terms of the time spent on each training iteration, the MLP model has a simple internal structure and the shortest training time per step. The GRU-AT model takes less time than the RouteNet model because the extracted features are more representative and the GRU and attention mechanism are used in the message passing process.

[0144] Table 2 Performance comparison of different models

[0145] Model MAE MSE MAPE r² / % Time / s MLP 0.295 16.31 0.712 3.112 0.112 RouteNet 0.351 14.45 0.351 7.268 0.281 GRU-AT 0.071 0.009 0.035 97.12 0.175

[0146] Starting from the prediction of network performance of digital twin communication network, the present invention analyzes the advantages and disadvantages of several network performance prediction methods currently used. On this basis, a twin network performance prediction model based on graph neural network is proposed, which solves some defects of current network performance prediction methods to a certain extent and improves the operation efficiency and prediction accuracy. Therefore, the digital twin communication network performance prediction model proposed by the present invention is feasible.

[0147] Example 3

[0148] This embodiment provides a twin network performance prediction device based on a graph neural network, including:

[0149] Performance prediction result acquisition module: used to obtain real-time relevant network information of the physical communication network, input it into the trained network performance prediction model, and obtain the performance prediction result;

[0150] The process of determining the trained network performance prediction model includes:

[0151] Preprocessing the historical network information related to the physical communication network obtained by the digital twin communication network to obtain a dependency graph data set, wherein the dependency graph data set includes link, path and queue information;

[0152] Initializing the features of each link, path and queue state in the dependency graph data set to obtain the initial hidden state of each link, path and queue;

[0153] The pre-built network performance prediction model is trained using the initial hidden state to obtain a trained network performance prediction model, wherein the network performance prediction model obtains iteratively updated hidden state information through the collaborative work of an attention mechanism, a message passing neural network, and a gated recurrent unit, and adjusts model parameters according to the hidden state information.

[0154] Example 4

[0155] This embodiment provides a computer storage medium on which a computer program is stored. When the computer program is executed by a processor, the twin network performance prediction method based on a graph neural network as described in any step of Example 2 is implemented.

[0156] Example 5

[0157] The embodiment of the present invention further provides a computer device, which may be a server, and its internal structure diagram may be as shown in FIG. Figure 8 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, 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. Among them, the processor of the computer device is used 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, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data obtained and generated in the method of the robot autonomously entering the packaging container. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the method of the aforementioned embodiment 2 is implemented.

[0158] Those skilled in the art will understand that Figure 8 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0159] The computer device provided in the embodiment of the present invention can execute the network performance prediction method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0160] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media including but not limited to disk storage, CD-ROM, optical storage, etc., which contain computer-usable program codes.

[0161] The present application is described with reference to the flowcharts and / or block diagrams of the methods, device systems, and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0162] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0163] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0164] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the enlightenment of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the purpose of the present invention and the claims, which all fall within the protection of the present invention.

Claims

1. A twin network performance prediction method based on graph neural network, characterized in that: include: Obtain real-time relevant network information of the physical communication network, input it into the trained network performance prediction model, and obtain performance prediction results; The process of determining the trained network performance prediction model includes: Preprocessing the historical network information related to the physical communication network obtained by the digital twin communication network to obtain a dependency graph data set, wherein the dependency graph data set includes link, path and queue information; Initializing the features of each link, path and queue state in the dependency graph data set to obtain the initial hidden state of each link, path and queue; The pre-built network performance prediction model is trained using the initial hidden state to obtain a trained network performance prediction model, wherein the network performance prediction model obtains iteratively updated hidden state information through the collaborative work of an attention mechanism, a message passing neural network, and a gated recurrent unit, and adjusts model parameters according to the hidden state information.

2. The twin network performance prediction method based on graph neural network according to claim 1 is characterized in that: The physical communication network history related network information obtained by the digital twin communication network is preprocessed to obtain a dependency graph data set, including: Normalize the read original network samples to obtain a standardized data set; The original network samples in the standardized data set are converted into a network dependency graph to obtain a network dependency graph data set, wherein the network dependency graph includes a set of links , a set of queues on the output ports of a network device and a set of source-to-destination paths ; , where Represents the first link in a group of links Links, represents the number of links in a set of links; , where Represents the first queues, Indicates the number of queues in a group of queues; , where Represents the first Path, Represents the number of paths in a set of paths, represented by a series of tuples , , represent The number of elements in Represents the first queues, Represents the first A link.

3. The twin network performance prediction method based on graph neural network according to claim 2 is characterized in that: The state initialization of each link, path and queue state in the dependency graph data set to obtain the initial hidden state of each link, path and queue includes: Path state initialization: Create a first state vector by connecting traffic characteristics, packet number characteristics, and zero padding , where the traffic feature represents the data traffic passing through the path, the packet number feature represents the number of packets generated on the path, and zero padding is used to form a feature vector of a predetermined dimension; And / or, Link State Initialization: Create a second state vector by connecting capacity characteristics, policy characteristics, and zero padding , where the capacity feature represents the bandwidth capacity of the link, the policy feature represents the queue scheduling policy on the link, and zero padding is used to meet the dimensionality requirements; And / or, queue state initialization: create a third state vector by concatenating the size feature, priority feature, weight feature, and zero padding , where the size feature represents the current number of packets in the queue, the priority feature represents the packet priority, the weight feature represents the queue weight associated with the scheduling policy, and zero padding is used to meet the dimensionality requirement.

4. The twin network performance prediction method based on graph neural network according to claim 3 is characterized in that: The pre-built network performance prediction model includes: a GRUCell gated recurrent unit layer, an Attention layer, a message passing loop module and a Readout layer connected in sequence; the message passing loop module includes an MPNN message passing neural network layer and an RNN recurrent neural network layer connected to each other.

5. The twin network performance prediction method based on graph neural network according to claim 4 is characterized in that: The using the initial hidden state to train the pre-built network performance prediction model to obtain a trained network performance prediction model includes: A path cycle unit, a link cycle unit and a queue cycle unit are defined respectively through the GRUCell gated cycle unit layer; In combination with the Attention layer, the MPNN message passing neural network layer and the RNN recurrent neural network layer, the hidden state information of the path cycle unit, the link cycle unit and the queue cycle unit is iteratively updated starting from the initial hidden state through a message passing process to obtain the global state of the physical communication network; Outputting the performance prediction value according to the updated path hidden state through the Readout readout layer; The network performance prediction model is trained according to the mean square error MSE between the performance prediction value and the true value, and the model hyperparameters are set, wherein the training objective of the network performance prediction model is It is expressed as: ， Where: is the actual performance value of the i-th dependency graph sample, is the performance prediction value of the i-th dependency graph sample, is the number of samples in the dependency graph dataset, and i represents the sequence number of the sample in the dependency graph dataset.

6. The twin network performance prediction method based on graph neural network according to claim 5 is characterized in that: The method further includes: selecting a test set from the dependency graph data set, inputting the test set into the network performance prediction model generated by each n rounds of training and the last round of training, and calculating the mean absolute error (MAE), mean absolute percentage error (MAPE) and correlation coefficient between the predicted value and the true value of the network performance prediction model. Evaluate the model training effect and find better hyperparameters to obtain the best network performance prediction model, where: Calculate the mean absolute error (MAE), mean absolute percentage error (MAPE), and correlation coefficient between the predicted value and the true value of the network performance prediction model , the calculation formula is: ， ， ， Where: is the average of the actual performance values ​​of n dependency graph samples.

7. The twin network performance prediction method based on graph neural network according to claim 6 is characterized in that: The iterative updating of the hidden state information of the path cycle unit, the link cycle unit and the queue cycle unit through the message passing process includes: in each iteration, the network performance prediction model performs the following three stages of message passing: Queue and link to path: collects information about the queues and links contained in each path, and hides the status of the links at a fixed time. and queue hidden state Compute the temporary state generated by the attention mechanism ,according to and the path hidden state at that moment As input, the GRU unit in the RNN recurrent neural network layer is used to update the hidden state of the new path and the new path temporarily hides the state ; Path to Queue: aggregates temporary hidden state information of the queue and all paths containing the queue , temporarily hide all status information After summing, we get the sum of temporary hidden state information of the queue. , based on the sum of the temporarily hidden state information of the queue and queue hidden status The new queue hidden state is obtained by updating the GRU unit in the RNN recurrent neural network and the new queue temporarily hides the state ; Queue to link: Temporarily hide state based on new queue and link hidden status , the new link hidden state is obtained by updating the GRU unit in the RNN recurrent neural network .

8. A twin network performance prediction device based on graph neural network, characterized in that: include: Performance prediction result acquisition module: used to obtain real-time relevant network information of the physical communication network, input it into the trained network performance prediction model, and obtain the performance prediction result; The process of determining the trained network performance prediction model includes: Preprocessing the historical network information related to the physical communication network obtained by the digital twin communication network to obtain a dependency graph data set, wherein the dependency graph data set includes link, path and queue information; Initializing the features of each link, path and queue state in the dependency graph data set to obtain the initial hidden state of each link, path and queue; The pre-built network performance prediction model is trained using the initial hidden state to obtain a trained network performance prediction model, wherein the network performance prediction model obtains iteratively updated hidden state information through the collaborative work of an attention mechanism, a message passing neural network, and a gated recurrent unit, and adjusts model parameters according to the hidden state information.

9. A computer storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed by a processor, the twin network performance prediction method based on graph neural network as described in any one of claims 1 to 7 is implemented.

10. An electronic terminal, characterized in that: It includes a processor and a memory connected to the processor, wherein a computer program is stored in the memory. When the computer program is executed by the processor, the twin network performance prediction method based on graph neural network as described in any one of claims 1 to 7 is executed.

Images