Data-driven communication network performance prediction method and device

Abstract

The invention discloses a data-driven communication network performance prediction method and device. The method comprises the following steps: acquiring a network information set; constructing network graph information based on the network information set; and processing the network graph information by using a trained network performance estimation model to obtain a prediction result value of the network performance. According to the method, the static network diagram and the dynamic network diagram are established based on the network structure of the communication network, comprehensive description of a communication network result is realized, corresponding data variables and structures are specially established from the aspects of data queues, streams and the like, and the accuracy of an evaluation result is ensured.

Description

Technical Field

[0001] The present invention relates to the fields of communications and artificial intelligence, and in particular to a data-driven communication network performance prediction method and device. Background Art

[0002] Communication operators such as mobile networks and data center networks often need to predict the impact of various changes on network performance, such as new workloads, changes in topology, or network configuration. The most common way to predict network performance is to use discrete-time simulators at the packet level, such as ns-3, omnet++, etc. However, the operating efficiency of this type of network simulator is very low. It takes 10 hours to simulate a topology with 6,000 hosts, while the simulation time is only 5 seconds. Any packet-level network simulator has inherent scalability limitations in high-speed, large-scale networks. This problem will become more serious over time as networks become larger and faster. To this end, it is necessary to study more efficient methods to achieve accurate estimates of network performance. Summary of the invention

[0003] The present invention mainly solves the problem of how to improve the efficiency of network simulation and fault prediction, and discloses a data-driven communication network performance prediction method and device.

[0004] In a first aspect of an embodiment of the present invention, a data-driven communication network performance prediction method is disclosed, comprising: S1, obtain network information set; S2, constructing network graph information based on the network information set; S3, using the trained network performance estimation model to process the network graph information to obtain a predicted result value of the network performance.

[0005] The network information set includes communication network information at each moment; The communication network information at a certain moment is defined as , where the router information set is represented as , is the number of routers, represents the i-th router, and the link information set is expressed as , represents the i-th link information, Represents the total number of links, and the queue information set is represented as , Indicates the i-th queue information, is the total number of queues, and the routing strategy set of the router at time t is expressed as , It is a router exist Routing strategy at the moment; The flow information set at time is , Indicates the number of flows, represents the i-th flow information at time t; the flow information at time t is the set of queues and links through which the data sequence of the communication network at time t flows; the data sequence of the communication network at time t, according to the routing strategy Determined; the i-th flow information at time t is expressed as , and They represent the starting queue and starting link through which the flow flows. and They respectively represent the termination queue and termination link through which the flow flows.

[0006] The step of constructing network diagram information based on the network information set includes: S21, constructing a static influence diagram set based on the network information set; the static influence diagram set includes a plurality of static influence diagrams; The expression of the static influence diagram set is , is the number of time steps, represents the static influence diagram at the first moment, Represents the static influence diagram at the Tth moment. Each static influence diagram contains a set of nodes and feature set ; The node set includes node information and feature information; the node information includes flow information, link information and queue information; the feature information includes flow node features, link node features and queue node features; S22, mapping the node information in the static influence diagram into node type information using a node type mapping function; S23, mapping the characteristic information in the static influence graph into influence relationship information using a relationship type mapping function; S24, constructing a dynamic influence diagram using the node type information and the influence relationship information; S25, constructing network diagram information using the dynamic influence diagram and the static influence diagram set.

[0007] The node type mapping function The expression is: , in, Indicates node information. is the node type information, the value range of the node type information is , They represent flow type, link type and queue type respectively; The relationship type mapping function The expression is: , in, Represents feature information, Indicates the influence relationship information. The value range of the influence relationship information is , Link Type Convection Type The impact relationship information, For queue type Convection Type The impact relationship information.

[0008] The network performance estimation model includes a heterogeneous graph multi-head attention layer, a temporal multi-head attention layer and a multi-indicator output layer; The heterogeneous graph multi-head attention layer is used to process the network graph information to obtain an output vector; the input end of the heterogeneous graph multi-head attention layer is used to receive the network graph information; the output end of the heterogeneous graph multi-head attention layer is connected to the input end of the temporal multi-head attention layer; The temporal multi-head attention layer is used to process the output vector of the heterogeneous graph multi-head attention layer to obtain a node embedding feature vector; the output end of the temporal multi-head attention layer is connected to the input end of the multi-index output layer; The multi-indicator output layer is used to process the node embedded feature vector to obtain the predicted result value of the network performance; the output end of the multi-indicator output layer is used to output the predicted result value of the network performance; the predicted result value of the network performance includes the predicted result values ​​of the flow throughput, flow delay, link utilization and queuing delay of the communication network.

[0009] The heterogeneous graph multi-head attention layer includes a first processing model, a second processing model and a third processing model; the second processing model is connected to the first processing model and the third processing model respectively; The calculation expression of the first processing model is: , in, Node type information The corresponding space transformation matrix is, Represents the output vector of the i-th node information of the static influence diagram set after the k-th feature mapping, represents the characteristics of the i-th node of the j-th static influence diagram in the static influence diagram set, , Indicates that the information of the ith node in the static influence diagram set is input into the node type mapping function The node type information obtained later; The expression of the second processing model is: , , in, Represents the influence value of the j-th node on the i-th node through the edge of type r in the static influence graph, represents the vector concatenation operation, Represents the output vector of the j-th node information of the static influence diagram set after the k-th feature mapping, Is of type The attention parameter vector of the edge, Represents the position encoding value of the edge from the jth node to the ith node in the flow information, Indicates The weight coefficient of the jth node corresponds to the node. Representation Node The pass type is The neighbor nodes connected by the edge, represents the activation function; The expression of the third processing model is: , in, , For Node The type value of , The node type is The set of types of edges related to the nodes, represents the edge type variable, for The value range of It represents the output vector of the i-th node information of the static influence diagram set after the k+1-th feature mapping. It represents the output vector of the jth node information of the static influence graph set after the kth feature mapping; head represents the sequence number of the attention head, is the number of attention heads in the heterogeneous graph multi-head attention layer, It is an injective function, implemented using a multi-layer perceptron, which is used to map different types of node influence relationship information and multi-head attention information into a feature space of the same dimension.

[0010] The input sequence of the temporal multi-head attention layer is ,in is the type value of the heterogeneous graph multi-head attention layer at time t The output result of the input sequence length is the time window ; The temporal multi-head attention layer includes a fourth processing model and a fifth processing model; the fourth processing model and the fifth processing model are connected; The processing process of the fourth processing model includes: Perform an update operation on the input sequence, and the expression of the update operation is: , in, is the encoding function for the value at the current time t, is the sin encoding function; right Each of Both perform update operations to obtain updated input sequences; The processing process of the fifth processing model includes: , , , , in, , is the attention weight matrix determined by the scaled dot-product attention function, for The attention weights at time t and time p, is the time influence value between time t and time k, k is the time value of the time window, , is the time influence value between time t and time p, The vector dimension of the node embedding feature vector, type value The query matrix is , type value The bond matrix is , type value The value matrix of , is the number of temporal attention heads in the temporal multi-head attention layer, is the preset mapping matrix, represents the node embedding feature vector obtained after integrating historical time features, Concat represents the vector concatenation operation, Indicates the type value at time t The output of the first temporal attention layer, and so on. Indicates the type value at time t No. The output of the temporal attention layer is Representation Matrix The element in the tth row and pth column of .

[0011] The position encoding value ,pass Calculated; Since links and queues have different performance characteristics at different positions in the flow, it is necessary to serialize the positions of links and queues in the flow Encoding, when the node and nodes Type is not hour In this work, the position is encoded using the commonly used sine and cosine functions of different frequencies as follows , , in is the sequence number of the link or queue's position in the flow, Indicates the position vector elements, It is the position of the vector subscript. Find the quotient of 2 and round it down.

[0012] The sin encoding function can be implemented using a sin function. is the current time t in the time window The position in.

[0013] According to a second aspect of the present invention, a data-driven communication network performance prediction device is disclosed, the device comprising: A memory storing executable program code; a processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the data-driven communication network performance prediction method.

[0014] According to a third aspect of the present invention, a computer storable medium is disclosed, wherein the computer storable medium stores computer instructions, and when the computer instructions are called by a computer, they are used to execute the data-driven communication network performance prediction method.

[0015] According to a fourth aspect of the present invention, an information data processing terminal is disclosed. The information data processing terminal is used to implement the data-driven communication network performance prediction method.

[0016] The beneficial effects of the present invention are: The present invention discloses a data-driven communication network performance prediction method, which first extracts key elements (flows, links, queues) in the communication network, extracts the influence relationship between them and constructs an influence graph. Then, the designed heterogeneous graph multi-head attention layer is used to extract the space between the elements in the influence graph. On the basis of obtaining spatial information, the designed temporal multi-head self-attention mechanism is used to extract the time-dependent features of the influence graph, forming the final embedding of key element features, and finally using the readout function to achieve multi-granular network performance estimation, providing support for communication network performance analysis and network optimization.

[0017] Based on the network structure of the communication network, the present invention establishes static network diagrams and dynamic network diagrams, realizes a comprehensive description of the communication network results, and specifically establishes corresponding data variables and structures from the perspectives of data queues, flows, etc., to ensure the accuracy of the evaluation results.

[0018] The present invention achieves a comprehensive evaluation of communication network performance by fusing spatial features and temporal features, thereby improving the comprehensiveness of performance prediction.

[0019] Other features and advantages of the present invention will be described in the following description, and part of them will become obvious from the description, or will be understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description, claims, and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 1 is a flow chart for implementing the method of the present invention; Figure 2 A schematic diagram of the composition of the heterogeneous graph multi-head attention layer of the network performance estimation model of the present invention; Figure 3 A schematic diagram of the composition of the temporal multi-head attention layer of the network performance estimation model of the present invention; Figure 4 It is a schematic diagram of the composition of the multi-index output layer of the network performance estimation model of the present invention. DETAILED DESCRIPTION

[0021] In order to better understand the content of the present invention, an embodiment is given here.

[0022] Figure 1 1 is a flow chart for implementing the method of the present invention; Figure 2 A schematic diagram of the composition of the heterogeneous graph multi-head attention layer of the network performance estimation model of the present invention; Figure 3 A schematic diagram of the composition of the temporal multi-head attention layer of the network performance estimation model of the present invention; Figure 4 It is a schematic diagram of the composition of the multi-index output layer of the network performance estimation model of the present invention.

[0023] In a first aspect of an embodiment of the present invention, a data-driven communication network performance prediction method is disclosed, comprising: S1, obtain network information set; S2, constructing network graph information based on the network information set; S3, using the trained network performance estimation model to process the network graph information to obtain a predicted result value of the network performance.

[0024] The network information set includes communication network information at each moment; The communication network information at a certain moment is defined as , where the router information set is represented as , is the number of routers, represents the i-th router, and the link information set is expressed as , represents the i-th link information, Represents the total number of links, and the queue information set is represented as , Indicates the i-th queue information, is the total number of queues, and the routing strategy set of the router at time t is expressed as , It is a router exist Routing strategy at the moment; The flow information set at time is , Indicates the number of flows, represents the i-th flow information at time t; the flow information at time t is the set of queues and links through which the data sequence of the communication network at time t flows; the data sequence of the communication network at time t, according to the routing strategy Determined; the i-th flow information at time t is expressed as , and They represent the starting queue and starting link through which the flow flows. and They respectively represent the termination queue and the termination link through which the flow flows; The queue is used to store data on the router port. In a communication network, a flow usually refers to a sequence of data transmitted from a source node to one or more destination nodes in the network.

[0025] The step of constructing network diagram information based on the network information set includes: S21, constructing a static influence diagram set based on the network information set; the static influence diagram set includes a plurality of static influence diagrams; the expression of the static influence diagram set is , is the number of time steps, represents the static influence diagram at the first moment, Represents the static influence diagram at the Tth moment. Each static influence diagram contains a set of nodes and feature set ; The node set includes node information and feature information; the node information includes flow information, link information and queue information; the feature information includes flow node features, link node features and queue node features; The corresponding flow node characteristics are ,in, yes The i-th flow size at time, is the one-hot encoding of the priority type of the i-th stream, including EF, AFx, BE, etc., represents the i-th flow at time Whether the route can be correctly routed to the destination router. A value of 1 indicates no, and a value of 0 indicates yes. The corresponding link node characteristics are ,in, represents the bandwidth of the link connecting router i and router j, is a one-hot encoding of the queue scheduling policy on the port connecting router i and router j, wherein the value of the one-hot encoding is first-in-first-out (FIFO) or differentiated services code point (DSCP); The corresponding queue node characteristics are , . represents the queue size on the port connecting router i and router j, is the one-hot encoding of the queue priority type on the port connecting router i and router j, and its value is EF, AFx, BE, etc. Represents the scheduling weight of the queue on the port connecting router i and router j.

[0026] S22, mapping the node information in the static influence diagram into node type information using a node type mapping function; S23, mapping the characteristic information in the network graph information into influence relationship information using a relationship type mapping function; S24, constructing a dynamic influence diagram using the node type information and the influence relationship information; S25, constructing network diagram information using the dynamic influence diagram and the static influence diagram set.

[0027] The static influence diagram and the dynamic influence diagram both contain nodes and edges.

[0028] The node type mapping function The expression is: , in, Indicates node information. is the node type information, the value range of the node type information is , They represent flow type, link type and queue type respectively; The relationship type mapping function The expression is: , in, Represents feature information, Indicates the influence relationship information. The value range of the influence relationship information is , Link Type Convection Type The impact relationship information, For queue type Convection Type The influence relationship information of other influence relationship information can be deduced by analogy.

[0029] The influence relationship information is a collection of the corresponding two node features. Specifically, the directly connected links and queues form The influence relationship can be expressed by a matrix; the influence relationship of the queues passing through the link of the router is , the impact relationship of the queue not passed is , a flow node is formed between the link and the queue. and Affects relationships.

[0030] The network performance estimation model includes a heterogeneous graph multi-head attention layer, a temporal multi-head attention layer and a multi-indicator output layer; The heterogeneous graph multi-head attention layer is used to process the network graph information to obtain an output vector; the input end of the heterogeneous graph multi-head attention layer is used to receive the network graph information; the output end of the heterogeneous graph multi-head attention layer is connected to the input end of the temporal multi-head attention layer; The temporal multi-head attention layer is used to process the output vector of the heterogeneous graph multi-head attention layer to obtain a node embedding feature vector; the output end of the temporal multi-head attention layer is connected to the input end of the multi-index output layer; The multi-indicator output layer is used to process the node embedded feature vector to obtain the predicted result value of the network performance; the output end of the multi-indicator output layer is used to output the predicted result value of the network performance; the predicted result value of the network performance includes the predicted result values ​​of the flow throughput, flow delay, link utilization and queuing delay of the communication network.

[0031] The heterogeneous graph multi-head attention layer includes a first processing model, a second processing model and a third processing model; The expression of the first processing model is: , in, For type information The corresponding spatial transformation matrix is ​​the parameter of the heterogeneous graph multi-head attention layer, which can be changed during the training process of the network performance estimation model. Represents the output vector of the i-th node information of the static influence diagram set after the k-th feature mapping, represents the characteristics of the i-th node of the j-th static influence diagram in the static influence diagram set, , Indicates that the information of the ith node in the static influence diagram set is input into the node type mapping function The node type information obtained later; The expression of the second processing model is: , , in, Represents the influence value of the j-th node on the i-th node through the edge of type r in the static influence graph, represents the vector concatenation operation, Is of type The attention parameter vector of the edge, which can be learned and changed during the training process of the network performance estimation model, Represents the position encoding value of the edge from the jth node to the ith node in the flow information, Indicates The weight coefficient of the jth node corresponds to the node. This parameter can be learned and changed during the training process of the network performance estimation model. Representation Node The pass type is The neighbor nodes connected by the edge, represents the activation function; The expression of the third processing model is: , in, , For Node The type value of , The node type is The set of types of edges related to the nodes, represents the edge type variable, for The value range of ,but , It represents the output vector of the i-th node information of the static influence diagram set after the k+1-th feature mapping. It represents the output vector of the jth node information of the static influence graph set after the kth feature mapping; head represents the sequence number of the attention head, is the number of attention heads in the heterogeneous graph multi-head attention layer, It is an injective function, implemented using a multi-layer perceptron, which is used to map different types of node influence relationship information and multi-head attention information into a feature space of the same dimension; The dimension transformation process is .

[0032] The relevant edge of a node refers to the influence relationship information corresponding to the feature information related to the node information. For example, the relevant feature information of node F is , .

[0033] The input sequence of the temporal multi-head attention layer is , ,in is the type value of the heterogeneous graph multi-head attention layer at time t The output result of the input sequence length is the time window , The node type value is The number of nodes, is the dimension of node features.

[0034] The temporal multi-head attention layer The output at the moment is , To integrate historical time features The new node feature embedding has a dimension of , which is the default value.

[0035] The temporal multi-head attention layer includes a fourth processing model and a fifth processing model; the processing process of the fourth processing model includes: Perform an update operation on the input sequence, and the expression of the update operation is: , in, is the encoding function for the value at the current time t, is the sin encoding function; Each of Both perform update operations to obtain updated input sequences; The processing process of the fifth processing model includes: , , , , in, , is the attention weight matrix determined by the scaled dot-product attention function, for The attention weights at time t and time p, is the time influence value between time t and time k, k is the time value of the time window, , is the time influence value between time t and time p, is the vector dimension of the node embedding feature vector; It is the MatMul network parameter, type value The query matrix is , which can be learned and changed during the training process of the network performance estimation model. The key matrix is , this parameter can be learned and changed during the training process of the network performance estimation model, and the value matrix is , which can be learned and changed during the training process of the network performance estimation model. is the number of temporal attention heads in the temporal multi-head attention layer, is a preset mapping matrix, which can be learned and changed during the training process of the network performance estimation model. represents the node embedding feature vector obtained after integrating historical time features, Concat represents the vector concatenation operation, Indicates the type value at time t The output result of the first temporal attention layer is: Representation Matrix The element in the tth row and pth column of .

[0036] The fifth processing model includes several temporal attention layers.

[0037] The multi-index output layer includes a first multi-layer perceptron module, a second multi-layer perceptron module, a third multi-layer perceptron module, and a fourth multi-layer perceptron module; The calculation expressions of the first multi-layer perceptron module, the second multi-layer perceptron module, the third multi-layer perceptron module, and the fourth multi-layer perceptron module are respectively: , , , , in represents the predicted result value of flow throughput, flow delay, link utilization and queuing delay of the communication network at time t, , , , Respectively represent the multilayer perceptron operations of the first multilayer perceptron module, the second multilayer perceptron module, the third multilayer perceptron module, and the fourth multilayer perceptron module; The i-th node information of the static influence diagram set may be the i-th node information of the j-th static influence diagram in the static influence diagram set; The sin encoding function can be implemented by a sin function; The scaled dot product attention function can be implemented by a scaled dot product attention module.

[0038] The position encoding value ,pass Calculated.

[0039] The position of the vector is subscripted to 2 and the integer is rounded down. ,For example: is 32-dimensional, The values ​​of are shown in Table 1.

[0040] Table 1 Value table The network performance estimation model can be implemented based on PYTHON 3.7.10 and above, and trained on Huawei Ascend Atlas300 graphics card. The training samples can be obtained by annotating the communication network structure information and the measured communication network performance indicators. During training, the sample batch size is not less than 10, the iteration round is not less than 400 times, the optimizer is Adam, and the learning rate is not more than 1×10 -4The loss function may be a cross entropy loss function, and the trained model may be evaluated using accuracy, precision, and recall, which is not limited in the embodiment of the present invention.

[0041] A second aspect of an embodiment of the present invention discloses a data-driven communication network performance estimation method, which specifically includes the following steps: S1: Capture the mutual influence relationship between various attributes in the communication network and construct an influence graph model; specifically, S11: Characterize the communication network and build a communication network model; specifically, this method will The time communication network is defined as , The router set , Link Collection , Moment Stream Collection , queue set ,Router routing strategy set , It is a router exist This method defines a flow as its routing strategy. A sequence of tuples consisting of queues and links determined by the start and end queue links , representing the queues and links through which the flow flows.

[0042] S12: Abstract the key elements that affect the performance of the communication network into nodes and extract their key features; specifically, The temporal impact diagram to be generated by this method is defined as a series of static impact diagram snapshots. , is the number of time steps, each snapshot is a static heterogeneous graph, defined as , containing a collection of nodes and influence relationship set . Nodes include multiple types, and the node type mapping function is defined as , node type Including flows, links and queues. Influence relationships include multiple types, and the influence relationship type mapping function is defined as ,Since the influence relationships between different node types are different, for example, link bandwidth has a greater impact on flow performance, flow size and burstiness affect queuing delay, and queue scheduling strategy affects link utilization. Therefore, this method defines the influence relationships between six different types of nodes , For Link convection The impact of other impact relationship types is similar.

[0043] This method extracts the key features that have the greatest impact on network performance for different types of nodes and models them. It is worth noting that when users pay attention to other features, they can modify and expand them as needed.

[0044] The flow node characteristics are .in, yes The flow size at the time, It is the one-hot encoding of the stream priority type, including EF, AFx, BE, etc. Represents flow at time Whether the packet can be correctly routed to its destination. 1 indicates an error and 0 indicates successful delivery.

[0045] The link node characteristics are Q represents the link bandwidth, and Is a connection node and nodes The one-hot encoding of the queue scheduling policy on the port, including first-in-first-out (FIFO) or differentiated services code point (DSCP).

[0046] The queue node characteristics are . Modeled as a feature vector . Indicates the queue size, is the one-hot encoding of the queue priority type (e.g. EF, AFx, BE, etc.), and Indicates the queue scheduling weight.

[0047] S13: Construct the influence relationship between flow, link, and queue elements to form an influence diagram. Specifically, the influence relationship refers to the influence relationship between flow, link, and queue nodes, including , and Three types of bidirectional influence relationships. First, obtain the source node and destination node of the flow, and then based on the topology structure and routing strategy Get the links and queues that the flow passes through. Directly connected links and queues form The flow nodes, links, and queues form and If the flow is not connected to the destination node, the flow is unreachable, which is recorded as ;otherwise, It is worth noting that this method considers the network topology as implicit information and only constructs the direct links, queue relationships, and relationships between links and queues related to the flow. This eliminates nodes that are not related to flow performance, thereby reducing the size of the influence graph and improving training speed and prediction accuracy.

[0048] S2: In order to learn the influence relationship between different types of nodes in the heterogeneous influence graph, a heterogeneous graph multi-head attention mechanism (HGMA) is designed. It first uses the edge attention mechanism to learn the influence of different types of nodes. Then the influence results of different types of nodes are aggregated to form the final node embedding. Subsampling extracts indirect effects between elements; specifically, S21: Map the features of different types of nodes into the same feature space; specifically, first, for any influence diagram This method extracts its node features as the initial embedding . Representation Node Type, yes The feature dimension of the type node. Due to the heterogeneity of nodes, different types of nodes have different feature spaces. This method uses a specific transformation matrix Mapping the features of different types of nodes into the same feature space, the calculation expression of S21 is: S22: Use the attention mechanism to learn the weights between different edge types; specifically, given a Connected node pairs , the edge attention is , which represents the node For Node By edge type The calculation expression of S22 is: in, Indicates a connection operation. , is the edge type It should be noted that the links and queues have different performance characteristics due to their different positions in the flow. Therefore, it is necessary to consider the relative positions of the links and queues to the flow. Encoding, when the node and nodes Type is not hour In this work, the position is encoded using the commonly used sine and cosine functions of different frequencies as follows , , in is the position of the link or queue in the flow, is the dimension. Then, after obtaining the influence degree between node pairs based on edge type, this method normalizes them through the softmax function to obtain the weight coefficient , , in Representation Node Based on edge type Neighbors, represents the activation function. Therefore, The weight coefficients of nodes depend on their characteristics. Then, The edge-type-based embeddings can be aggregated by the projected features of the neighbors.

[0049] S23: Design a multi-head attention mechanism to obtain spatial features from multiple angles; specifically, since the heterogeneous influence graph has a scale-free property, too high graph data variance leads to unstable training. Therefore, this method introduces a multi-head attention mechanism and embeds the learned features as fusion features. Each attention head has an independent parameter weight. The expression of step S23 is: , in For Node Type, , Is the node type The relevant edge type, for example, the node type is ,but . is the number of attention heads in the heterogeneous graph multi-head attention layer, It is an injective function, implemented using a layer of MLP, which maps the influence results of different types of nodes and multi-head attention results to the same feature space .

[0050] S3: In order to further capture the temporal evolution characteristics of the dynamic influence graph, this method designs a temporal multi-head attention layer, which embeds the node features at the current moment by fusing the historical time series features; specifically, S31: Aggregation of temporal position features and node initial features; specifically, since there are multiple types in the network, this method uses different attention parameters to learn the temporal evolution features of different types of nodes. The input of this layer is Time-specific node type The historical time series feature of the node type. This feature has fully captured the spatial information of the influence graph, making the structural modeling and time modeling separate. , this method defines the input as ,in is the final embedding vector after capturing the spatial information of the influence graph, and the sequence length is the time window , Node Type The number of nodes, is the node feature dimension. The output is time Integrating historical time features The new node feature embedding has a dimension of The update expression of the embedding vector is: .

[0051] The key goal of the temporal multi-head self-attention layer is to capture the historical time window Internal features for the current The impact of moment-to-moment network embedding. yes The representation after the spatial features are fused at all times. This method uses As a query, we focus on its historical representation. Different from spatial attention that acts on the representations of neighboring nodes, temporal attention relies entirely on the temporal window features of each node, thus promoting effective parallelism between nodes.

[0052] S32: Design a multi-head self-attention strategy to integrate historical time series features to embed time features; specifically, this method designs a Scaled Dot-Product Attention and Multi-HeadAttention strategy, where the query, key, and value are set as node input representations. The query vector is , key is The sum is ,The temporal self-attention function is defined as: , , , in is the attention weight matrix obtained by the multiplication attention function, for The time vector and The attention weight of the moment, In order to capture different types of temporal dependencies between different features, this method uses a multi-head attention mechanism to learn the temporal dependencies of the input sequence from multiple perspectives. Each "head" can learn different patterns and dependencies, thereby providing richer information. The process expression is: , in It's time to pay attention to the number of heads, is the mapping matrix, which maps it to a unified feature space.

[0053] S32: Finally, different readout functions are used to achieve multi-granularity network performance estimation. The readout function is a multi-layer perceptron.

[0054] A communication network is a set of devices and protocols used to transmit information. These devices include but are not limited to routers, switches, servers, terminal devices (such as mobile phones, computers), etc., while protocols are the standards that specify how data is transmitted between these devices. Communication networks can be classified according to different standards, such as coverage, transmission media, network topology, etc.

[0055] Those skilled in the art will appreciate that all or part of the processes of the above-mentioned embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, wherein the computer-readable storage medium is a disk, an optical disk, a read-only storage memory, or a random access memory, etc.

[0056] The above description is only an embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent substitution, improvement, etc. made within the spirit and principle of the present invention shall be included in the scope of the claims of the present invention.

Claims

1. A data-driven communication network performance prediction method, characterized in that: include: S1, obtain network information set; S2, constructing network graph information based on the network information set; S3, using the trained network performance estimation model to process the network graph information to obtain a predicted result value of the network performance.

2. The data-driven communication network performance prediction method according to claim 1, characterized in that: The network information set includes communication network information at each moment; The communication network information at a certain moment is defined as , where the router information set is represented as , is the number of routers, represents the i-th router, and the link information set is expressed as , represents the i-th link information, Represents the total number of links, and the queue information set is represented as , Indicates the i-th queue information, is the total number of queues, and the routing strategy set of the router at time t is expressed as , It is a router exist Routing strategy at the moment; The flow information set at time is , Indicates the number of flows, represents the i-th flow information at time t; the flow information at time t is the set of queues and links through which the data sequence of the communication network at time t flows; the data sequence of the communication network at time t, according to the routing strategy Determined; the i-th flow information at time t is expressed as , and They represent the starting queue and starting link through which the flow flows. and They respectively represent the termination queue and termination link through which the flow flows.

3. The data-driven communication network performance prediction method according to claim 2, characterized in that: The step of constructing network diagram information based on the network information set includes: S21, constructing a static influence diagram set based on the network information set; the static influence diagram set includes a plurality of static influence diagrams; The expression of the static influence diagram set is , is the number of time steps, represents the static influence diagram at the first moment, Represents the static influence diagram at the Tth moment. Each static influence diagram contains a set of nodes and feature set ; The node set includes node information and feature information; the node information includes flow information, link information and queue information; the feature information includes flow node features, link node features and queue node features; S22, mapping the node information in the static influence diagram into node type information using a node type mapping function; S23, mapping the characteristic information in the static influence graph into influence relationship information using a relationship type mapping function; S24, constructing a dynamic influence diagram using the node type information and the influence relationship information; S25, constructing network diagram information using the dynamic influence diagram and the static influence diagram set.

4. The data-driven communication network performance prediction method according to claim 3, characterized in that: The node type mapping function The expression is: ， in, Indicates node information. is the node type information, the value range of the node type information is , They represent flow type, link type and queue type respectively; The relationship type mapping function The expression is: ， in, Represents feature information, Indicates the influence relationship information. The value range of the influence relationship information is , Link Type Convection Type The impact relationship information, For queue type Convection Type The impact relationship information.

5. The data-driven communication network performance prediction method according to claim 4, characterized in that: The network performance estimation model includes a heterogeneous graph multi-head attention layer, a temporal multi-head attention layer and a multi-indicator output layer; The heterogeneous graph multi-head attention layer is used to process the network graph information to obtain an output vector; the input end of the heterogeneous graph multi-head attention layer is used to receive the obtained network graph information; The output end of the heterogeneous graph multi-head attention layer is connected to the input end of the temporal multi-head attention layer; The temporal multi-head attention layer is used to process the output vector of the heterogeneous graph multi-head attention layer to obtain a node embedding feature vector; the output end of the temporal multi-head attention layer is connected to the input end of the multi-index output layer; The multi-indicator output layer is used to process the node embedded feature vector to obtain the predicted result value of the network performance; the output end of the multi-indicator output layer is used to output the predicted result value of the network performance; the predicted result value of the network performance includes the predicted result values ​​of the flow throughput, flow delay, link utilization and queuing delay of the communication network.

6. The data-driven communication network performance prediction method according to claim 5, characterized in that: The heterogeneous graph multi-head attention layer includes a first processing model, a second processing model and a third processing model; the second processing model is connected to the first processing model and the third processing model respectively; The calculation expression of the first processing model is: ， in, Node type information The corresponding space transformation matrix is, Represents the output vector of the i-th node information of the static influence diagram set after the k-th feature mapping, represents the characteristics of the i-th node of the j-th static influence diagram in the static influence diagram set, , Indicates that the information of the ith node in the static influence diagram set is input into the node type mapping function The node type information obtained later; The expression of the second processing model is: ， ， in, Represents the influence value of the j-th node on the i-th node through the edge of type r in the static influence graph, represents the vector concatenation operation, Represents the output vector of the j-th node information of the static influence diagram set after the k-th feature mapping, Is of type The attention parameter vector of the edge, Represents the position encoding value of the edge from the jth node to the ith node in the flow information, Indicates The weight coefficient of the jth node corresponds to the node. Representation Node The pass type is The neighbor nodes connected by the edge, represents the activation function; The expression of the third processing model is: ， in, , For Node The type value of , The node type is The set of types of edges related to the nodes, represents the edge type variable, for The value range of It represents the output vector of the i-th node information of the static influence diagram set after the k+1-th feature mapping. It represents the output vector of the jth node information of the static influence graph set after the kth feature mapping; head represents the sequence number of the attention head, is the number of attention heads in the heterogeneous graph multi-head attention layer, It is an injective function, implemented using a multi-layer perceptron, which is used to map different types of node influence relationship information and multi-head attention information into a feature space of the same dimension.

7. The data-driven communication network performance prediction method according to claim 6, characterized in that: The input sequence of the temporal multi-head attention layer is ,in is the type value of the heterogeneous graph multi-head attention layer at time t The output result of the input sequence length is the time window ; The temporal multi-head attention layer includes a fourth processing model and a fifth processing model; The fourth processing model is connected to the fifth processing model; The processing process of the fourth processing model includes: Perform an update operation on the input sequence, the expression of the update operation is: ， in, is the encoding function for the value at the current time t, is the sin encoding function; right Each of Both perform update operations to obtain updated input sequences; The processing process of the fifth processing model includes: ， ， ， ， in, is the attention weight matrix determined by the scaled dot-product attention function, for The attention weights at time t and time p, is the time influence value between time t and time k, k is the time value of the time window, , is the time influence value between time t and time p, The vector dimension of the node embedding feature vector, type value The query matrix is , type value The bond matrix is , type value The value matrix of , is the number of temporal attention heads of the temporal multi-head attention layer, is the preset mapping matrix, represents the node embedding feature vector obtained after integrating historical time features, Concat represents the vector concatenation operation, Indicates the type value at time t The output of the first temporal attention layer, and so on. Indicates the type value at time t No. The output of the temporal attention layer is Representation Matrix The element in the tth row and pth column of .

8. A data-driven communication network performance prediction device, characterized in that: The device comprises: A memory storing executable program code; a processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the data-driven communication network performance prediction method according to any one of claims 1 to 7.

9. A computer storable medium, characterized in that: The computer storable medium stores computer instructions, and when the computer instructions are called by a computer, they are used to execute the data-driven communication network performance prediction method according to any one of claims 1 to 7.

10. An information data processing terminal, characterized in that: The information data processing terminal is used to implement the data-driven communication network performance prediction method as described in any one of claims 1 to 7.

Images