Address configuration method and related equipment

Abstract

The present application provides an address configuration method and related equipment. The method uses a spatiotemporal graph neural network model to obtain the predicted number of user terminals in each area at the next moment according to the number of user terminals at the current moment, and then uses a recursive neural network model to obtain the predicted online time of each user terminal in the corresponding area according to historical mobile sequence information. Finally, the address pool capacity of each area is set according to the predicted number of user terminals at the next moment, and the address lease time is allocated to the corresponding user terminal according to the predicted online time of each user terminal in the corresponding area. The recursive neural network model can accurately predict the user's online time, so that the behavioral characteristics of individual users can be fully utilized to ensure that the address lease time is effectively allocated to the individual user terminal. The spatiotemporal graph neural network model is also used to accurately predict the number of group user terminals in each area, so as to provide effective address lease time and address pool capacity.

Description

Technical Field

[0001] The present application relates to the field of computer network communication technology, and in particular to an address configuration method and related equipment. Background Art

[0002] DHCP (Dynamic Host Configuration Protocol) is widely deployed in local area networks to automatically assign IP (Internet Protocol) addresses to WiFi (Wireless Fidelity) devices. However, the frequent movement of user terminal devices poses a great challenge to DHCP performance. The current IP allocation strategy (including IP lease and IP address pool) relies on traditional models. Due to the different behavioral habits of individual users, the online time distribution of different user terminal devices varies greatly. Therefore, the number of user terminal devices in different locations and time periods will vary in space and time. By modeling the number of group user terminals in each area and configuring the IP address pool size according to the periodic pattern, not only the spatial dependence of other areas is ignored, but also the local time dependence of group users is ignored.

[0003] In addition, even if different users use the same user terminal device type and location, all IP addresses are set with the same lease time, and users are grouped according to specific characteristics (for example, area type and device type), and the same IP lease time is assigned to each group. This coarse-grained IP lease time configuration is far from fully utilizing the behavioral characteristics of individual users and cannot guarantee the performance of DHCP. Summary of the invention

[0004] In view of this, the purpose of this application is to propose an address configuration method and related equipment to solve or partially solve the above technical problems.

[0005] Based on the above purpose, the first aspect of the present application provides an address configuration method, including:

[0006] Obtain the number of user terminals at the current moment in each area covered by the target local area network;

[0007] Inputting the number of user terminals at the current moment into a pre-trained spatiotemporal graph neural network model to obtain the predicted number of user terminals in each area at the next moment;

[0008] Acquire historical movement sequence information of each user terminal, input the historical movement sequence information into a pre-trained recursive neural network model, and obtain a predicted online time of each user terminal in a corresponding area;

[0009] Setting the address pool capacity of each area according to the predicted number of user terminals at the next moment;

[0010] An address lease time is allocated to the corresponding user terminal according to the predicted online time of each user terminal in the corresponding area.

[0011] A second aspect of the present application provides an address configuration device, including:

[0012] A user terminal quantity acquisition module is configured to acquire the number of user terminals at the current moment in each area covered by the target local area network;

[0013] A user terminal quantity prediction module is configured to input the number of user terminals at the current moment into a pre-trained spatiotemporal graph neural network model to obtain a predicted number of user terminals in each area at the next moment;

[0014] An online time prediction module is configured to obtain historical movement sequence information of each user terminal, input the historical movement sequence information into a pre-trained recursive neural network model, and obtain a predicted online time of each user terminal in a corresponding area;

[0015] An address pool capacity setting module, configured to set the address pool capacity of each area according to the predicted number of user terminals at the next moment;

[0016] The address lease time allocation module is configured to allocate an address lease time to a corresponding user terminal according to the predicted online time of each user terminal in the corresponding area.

[0017] A third aspect of the present application provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method described in the first aspect when executing the program.

[0018] A fourth aspect of the present application provides a non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the method described in the first aspect.

[0019] From the above, it can be seen that the address configuration method and related equipment provided by the present application obtain the predicted number of user terminals in each area at the next moment through a pre-trained spatiotemporal graph neural network model according to the number of user terminals at the current moment, and then obtain the predicted online time of each user terminal in the corresponding area through a pre-trained recursive neural network model according to the historical mobile sequence information. Finally, the address pool capacity of each area is set according to the predicted number of user terminals at the next moment, and the address lease time is allocated to the corresponding user terminal according to the predicted online time of each user terminal in the corresponding area. The recursive neural network model is used to accurately predict the user's online time, thereby ensuring that the address lease time is effectively allocated to the individual user terminal, and the behavioral characteristics of the individual user can be fully utilized. The spatiotemporal graph neural network model is also used to capture the spatiotemporal dependence of the movement of group user terminals, and the number of group user terminals in each area is accurately predicted, thereby providing effective address lease time and address pool capacity to ensure the high performance of DHCP. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the present application or related technologies, the drawings required for use in the embodiments or related technical descriptions are briefly introduced below. Obviously, the drawings described below are only embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0021] Figure 1 A flowchart of an address configuration method according to an embodiment of the present application;

[0022] Figure 2 A schematic diagram of the structure of an address configuration device according to an embodiment of the present application;

[0023] Figure 3 A schematic diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0024] In order to make the objectives, technical solutions and advantages of the present application more clearly understood, the present application is further described in detail below in combination with specific embodiments and with reference to the accompanying drawings.

[0025] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the embodiments of the present application should be the usual meanings understood by people with ordinary skills in the field to which the present application belongs. The "first", "second" and similar words used in the embodiments of the present application do not represent any order, quantity or importance, but are only used to distinguish different components. "Including" or "comprising" and similar words mean that the elements or objects appearing in front of the word cover the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0026] In the related art, the address lease time is often adjusted according to the device type and the area type, for example, users are grouped according to the user role type and the area type (such as the work area and the living area), and the same address lease time is assigned to the users in the same group. Alternatively, users are grouped based on the operating system type and the area type of the user device, and the same address lease time is assigned to the users in the same group.

[0027] In addition, the address pool size is often updated periodically according to changes in the number of group user terminals. For example, the number of group user terminals in each area is modeled and the address pool size is configured according to a periodic pattern.

[0028] However, the online time of individual users varies greatly due to personal habits, even if their device type and location are the same. The same lease time is set for all addresses. Existing methods group users according to specific characteristics (e.g., region type and device type) and assign the same address lease time to each group. This coarse-grained address lease time configuration is far from fully utilizing the behavioral characteristics of individual users. In addition, the number of group users varies in space and time at different locations and time periods. The size of the address pool is usually configured the same, and existing methods model the number of group user terminals in each area and configure the address pool size according to a periodic pattern. This linear approach not only ignores the spatial dependencies of other areas, but also ignores the local temporal dependencies of group users.

[0029] The embodiment of the present application provides an address configuration method, which uses a recursive neural network model to learn the sequential pattern of individual user movement and accurately predicts the online time of the user terminal, thereby ensuring that the address lease time is effectively allocated to the individual user terminal. At the same time, a spatiotemporal graph neural network model is introduced to capture the spatiotemporal dependency of group user movement and predict the number of user terminals of group users in each area. By accurately modeling the movement patterns of individual users and group users, an effective configuration of address lease time and address pool capacity is provided to ensure the high performance of DHCP.

[0030] like Figure 1 As shown, the method of this embodiment includes:

[0031] Step 101, obtaining the number of user terminals in each area covered by the target local area network at the current moment.

[0032] In this step, DHCP is used to automatically allocate IP (Internet Protocol) addresses and related configuration information to clients in a local area network (LAN).

[0033] Due to the convenience of DHCP, it is widely used in wireless local area networks (WLAN). The target LAN can be a wired LAN or a wireless LAN. Here, a wireless LAN is preferred. The target LAN is usually divided into multiple areas, each of which is managed by an AC (Access Controller).

[0034] User terminals can be communication terminal devices such as computers, mobile phones, and tablet computers.

[0035] Step 102, input the number of user terminals at the current moment into a pre-trained spatiotemporal graph neural network model to obtain the predicted number of user terminals in each area at the next moment.

[0036] In this step, DHCP allocates an IP address pool to the AC of each area, which consists of multiple subnet segments. The DHCP server uses a space-time graph neural network model (STGNN). By inputting the number of user terminals in each area at the current moment into the space-time graph neural network model, the number of user terminals in each area in the next time period is predicted, and the corresponding number of subnet segments are allocated to each area.

[0037] By utilizing the spatiotemporal graph neural network model, the spatiotemporal dependencies of user terminal movements in each area can be captured, thereby accurately predicting the number of user terminals in each area.

[0038] In some embodiments, in step 102, the process of obtaining the spatiotemporal graphical neural network model includes:

[0039] Step 1021, construct a spatiotemporal graph neural network pre-training model.

[0040] Step 1022: Acquire the actual number of user terminals at each historical moment in each area covered by the wireless local area network, and divide the actual number of user terminals at each historical moment into a user terminal number training set and a user terminal number test set.

[0041] Step 1023, input the training set of the number of user terminals at the target historical moment into the spatiotemporal graph neural network pre-training model, and output the predicted number of user terminals at the next moment of the target historical moment in each area, wherein the target historical moment is any moment in the historical moments.

[0042] Step 1024: determine a first mean absolute error loss function loss according to the predicted number of user terminals at the next moment and the actual number of user terminals at the next moment. The first mean absolute error loss function loss is expressed as:

[0043]

[0044] Where V represents the area covered by the wireless local area network, i represents the i-th area, and Y i Indicates area V i The actual number of user terminals in Indicates area V i The number of user terminals is predicted in , θ2 represents all trainable and adjustable parameters in the spatiotemporal graph neural network pre-training model, and λ2 is expressed as a predefined hyperparameter.

[0045] Step 1025: determining a first root mean square error function RMSE according to the predicted number of user terminals at the next moment and the actual number of user terminals at the next moment. The first root mean square error function RMSE is expressed as:

[0046]

[0047] Among them, Y i Indicates area V i The actual number of user terminals in Indicates area V i The number of user terminals is predicted in , i represents the i-th area, and n represents the number of areas.

[0048] Step 1026: determine a first mean absolute error function MAE according to the predicted number of user terminals at the next moment and the actual number of user terminals at the next moment. The first mean absolute error function MAE is expressed as:

[0049]

[0050] Among them, Y i Indicates area V i The actual number of user terminals in Indicates area V i The number of user terminals is predicted in , i represents the i-th area, and n represents the number of areas.

[0051] Step 1027, based on the minimization processing of the first mean absolute error loss function loss, the spatiotemporal graph neural network pre-trained model is continuously trained and adjusted to obtain a trained spatiotemporal graph neural network pre-trained model, and the trained spatiotemporal graph neural network pre-trained model is used as the spatiotemporal graph neural network model.

[0052] Step 1028, input the test set of the number of user terminals at the target historical moment into the spatiotemporal graphical neural network model, output the number of test user terminals at the next moment of the target historical moment in each area, and calculate the accuracy of the spatiotemporal graphical neural network model based on the number of test user terminals at the next moment and the actual number of user terminals at the next moment using the first root mean square error function RMSE and the first mean absolute error function MAE.

[0053] In the above scheme, use Y i and Respectively represent area V i The actual number of user terminals of group users and the predicted number of user terminals of group users in . The loss function is the first mean absolute error loss function, which is expressed as follows:

[0054]

[0055] Where θ2 represents all trainable parameters in the spatiotemporal graph neural network pre-training model, and λ2 is a predefined hyperparameter introduced for L2 regularization to avoid overfitting problems. L2 regularization refers to the sum of the squares of each element in the weight vector and then taking the square root.

[0056] The training set is then used to train the spatiotemporal graph neural network pre-training model, and the first mean absolute error loss function is used to guide the adjustment of the model parameters. The optimal hyperparameters of the model can also be obtained through the validation set.

[0057] The test set is used to evaluate the trained spatiotemporal graph neural network model, and the values ​​of the first root mean square error function (RMSE) and the first mean absolute error function (MAE) of the spatiotemporal graph neural network model on the validation set are recorded.

[0058]

[0059]

[0060] In some embodiments, in step 1022, obtaining the actual number of user terminals at each historical moment in each area covered by the wireless local area network includes:

[0061] Step 10221, obtaining the media access control address of each user terminal at each historical moment in each area covered by the wireless local area network, the media access control address of the wireless access point in the area, and the name of the corresponding wireless access point.

[0062] Step 10222: Group the media access control address of each user terminal, the media access control address of the wireless access point in the area and the name of the corresponding wireless access point in each area covered by the wireless local area network at each historical moment according to a preset time to obtain multiple first groups.

[0063] Step 10223: Group each of the first groups according to the name of the wireless access point to obtain multiple second groups.

[0064] Step 10224, count the number of media access control addresses of each user terminal in the second group with the same name as each wireless access point, and obtain the actual number of user terminals at each historical moment in each area covered by the wireless local area network.

[0065] In the above scheme, SNMP (Simple Network Management Protocol) logs with a time span of 1 month (can be set according to specific circumstances, and are not specifically limited here) are collected from ACs. In the process of obtaining SNMP logs, a management program is executed to poll all ACs every 5 minutes (can be set according to specific circumstances, and are not specifically limited here), and the following key-value pair information is obtained from the ACs:

[0066] (1) The Media Access Control Address (MAC) of a user terminal and the Media Access Control Address (MAC) of the wireless access point (AP) to which it is associated;

[0067] (2) The Media Access Control Address (MAC) of a wireless access point (AP) and the name of the wireless access point (AP);

[0068] (3) A user terminal's Media Access Control Address (MAC) and the IP address SNMP log obtained from the DHCP server are used to determine the location of each user terminal.

[0069] For the media access control address of each user terminal, the media access control address of the wireless access point in the area and the name of the corresponding wireless access point, SNMP (Simple Network Management Protocol) (timestamp) grouping can be performed according to a time period (for example, every hour, which can be set according to specific circumstances and is not specifically limited here). The records in each group are further grouped according to the area (AP name, i.e., the name of the wireless access point) to obtain multiple second groups, and the number of MAC addresses of user terminals in each area is counted. Finally, a data set of the number of user terminals of users in each area group per hour (which can be set according to specific circumstances and is not specifically limited here) is obtained. The data set is used as the actual number of user terminals at each historical moment in each area covered by the wireless local area network for training and performance evaluation of the spatiotemporal graph neural network pre-training model.

[0070] In some embodiments, in step 102, the spatiotemporal graphical neural network model includes:

[0071] Graph convolutional network layers, recurrent neural network layers, and attention layers;

[0072] The graph convolutional network layer is used to obtain the local topological features of the number of user terminals at each moment;

[0073] The recursive neural network layer is used to learn local time dependencies in the time series of the number of user terminals;

[0074] The attention layer is used to learn periodic information in dynamic networks.

[0075] In the above scheme, the spatiotemporal graph neural network model includes a graph convolutional network layer, a recursive neural network layer and an attention layer. The specific structure is as follows:

[0076] First, the training set extracted from the SNMP log is input into the STGNN model. Each data in the data set is a time series. Given a sequence X consisting of the first n time slices = (X t-n+1 , ..., Xt ) and adjacency matrix A, the goal of group user terminal quantity prediction is to predict the number of group user terminals in each area, where X t ∈R |G| It is a snapshot of the number of user terminals in a group at a specific time slice t (t∈1, 2, …, T).

[0077] Since group users migrate frequently between regions, we use the adjacency matrix A∈R |V|×|V| To describe the corresponding static topology. In particular, when there is an edge between nodes i and j ((i, j) ∈ E), the weight is W(i, j), let (A) ij =W(i, j), where W(i, j) represents the strength of the relationship between node i and node j, where a value of 0 indicates that there is no relationship between the two nodes.

[0078] First, the sequence is passed into the graph convolutional network layer to model the topological structure of the group user terminal quantity relationship graph. t As input, the spectral graph convolution operation is performed according to the adjacency matrix A, and the output is X after a fully connected layer. t The overall operation of the GCN (Graph Convolutional Networks) unit can be simply defined as follows:

[0079]

[0080] Among them, Relu(·) is an activation function, is the convolution operation matrix, is the convolution kernel, I N is the identity matrix; is the degree matrix; W represents the weight matrix; X′ t is the output generated by the GCN unit. Finally, the graph convolutional network layer outputs the sequence X′=(X′ t-n+1 , X′ t-n+2 , ..., X′ t ).

[0081] Second, the output sequence X′ of the GCN unit is passed to a recurrent neural network layer composed of GRU (Gate Recurrent Unit) units to learn local temporal dependencies in the time series of the group user terminal quantity graph. The specific process is as follows:

[0082] (1) In a time series of the input data set, at a given time step t, the number of user terminals X in each regional group in the sequence t and the hidden state H t-1 , apply the graph convolutional network layer operation fg (·). The specific calculation method is as follows:

[0083] X′ t =f g (A,X t );H t-1 =f g (A,H t-1 );

[0084] (2) Taking the output sequence X′ of the graph convolutional network layer as input, at time step t, for the region V in the group user terminal quantity relationship graph i , the operation of the GRU unit in the recurrent neural network layer can be expressed as follows:

[0085] z t =σ(W z ·{X t [i,:],H t-1 [i,:]})

[0086] r t =σ(W r ·{X t [i,:],H t-1 [i,:]})

[0087]

[0088]

[0089] Where ⊙ represents element-wise multiplication, W represents the parameter of the corresponding unit, σ(·) represents the sigmoid (i.e., s-shaped growth curve) activation function, and H t[i，：] Indicates area V i The hidden state of the output at time step t. The final recurrent neural network layer outputs a hidden state sequence {H t-n+1 [i,:],H t-n+2 [i,:],...,H t [i,:]}.

[0090] Third, the output sequence of the recurrent neural network layer is used as input to the attention layer to capture the global time dependency in the time series of the group user terminal quantity graph. The specific process is as follows:

[0091] (1) Transformer is used as the attention layer to capture global dependencies directly after the GRU layer.

[0092] (2) For region V i , the output sequence of the recurrent neural network layer {H t-n+1 [i,:],H t-n+2[i,:],...,H t [i,:]} as the input of the Transformer module. The Transformer module includes an attention layer and a feedforward layer. During the model training process, a residual network (ResNet) is added between the attention layer and the feedforward layer and batch normalization is performed to alleviate the gradient vanishing situation and speed up the convergence speed. Here, the feedforward layer is a fully connected network with two linear layers.

[0093] (3) The input of the attention layer has dimensions d k The query vector Q and key vector K of dimension d v First, we sequentially sort the sequence {H t-n+1 [i,:],H t-n+2 [i,:],...,H t [i,:]} are stacked into a matrix form

[0094] (4) For each region V i , and we get the matrix Then the matrix Linearly transform into query vector, key vector and value vector as follows:

[0095]

[0096] Where W is the parameter shared by all regions of the WLAN, which will be learned during model training.

[0097] (5) For a given query vector, the weight is obtained by calculating the dot product between it and the key vector, and then the weight is normalized using the softmax activation function to obtain the final weight. In order to speed up the calculation, all queries are calculated at the same time. The calculation method is as follows:

[0098]

[0099] Since the attention layer in the Transformer module ignores the relative position in the sequence when calculating the Attention function, it performs the same processing on vectors at different positions. t [i,:] Add a code containing position information t , where s t is defined as follows:

[0100] If t is an odd number, If t is an even number,

[0101] Therefore, for the area V in WLAN i, using position encoding to convert the sequence {H t-n+1 [i,:],H t-n+2 [i,:],...,H t [i,:]} stacking is converted to a sequence with position information It is used as input to the Transformer module.

[0102] (6) After the Attention function is completed, the output state is passed to the feedforward layer and finally output As the next time period area V i The number of user terminals of group users. Here the feed-forward layer is a fully connected network with two linear layers.

[0103] Step 103: Obtain historical movement sequence information of each user terminal, input the historical movement sequence information into a pre-trained recursive neural network model, and obtain the predicted online time of each user terminal in the corresponding area.

[0104] In this step, the historical movement sequence information of the user terminal is transmitted to the trained Attention-Based Recurrent Neural Network (ARNN) model to obtain the predicted online time of each user terminal in the corresponding area.

[0105] By utilizing a recurrent neural network model to learn the sequential patterns of individual user movements and accurately predict the online time of user terminals, the address lease time can be effectively allocated to individual user terminals.

[0106] In some embodiments, in step 103, the process of obtaining the recursive neural network model includes:

[0107] Step 1031, construct a recursive neural network pre-training model.

[0108] Step 1032, obtain all the areas where each user terminal is located at each historical moment and the actual online time in each of the areas, take all the areas where each user terminal is located at each historical moment and the actual online time in each of the areas as historical actual mobile sequence information, and divide the historical actual mobile sequence information into a mobile sequence information training set and a mobile sequence information test set.

[0109] Step 1033, input the mobile sequence information training set of the target historical moment into the recursive neural network pre-training model, and output the predicted online time of each user terminal in the corresponding area at the next moment of the target historical moment, wherein the target historical moment is any moment in the historical moments.

[0110] Step 1034: determine a second mean absolute error loss function loss according to the predicted online time of each user terminal in the corresponding area at the next moment and the actual online time of each user terminal in the corresponding area at the next moment. The second mean absolute error loss function loss is expressed as:

[0111]

[0112] Among them, y i represents the actual online time of the corresponding user terminal in area i, represents the actual online time of the corresponding user terminal in region i, i represents the first region, θ1 represents all trainable and adjustable parameters in the recursive neural network pre-training model, and λ1 is expressed as a predefined hyperparameter.

[0113] Step 1035: determining a second root mean square error function RMSE according to the predicted online time of each user terminal in the corresponding area at the next moment and the actual online time of each user terminal in the corresponding area at the next moment, wherein the second root mean square error function RMSE is expressed as:

[0114]

[0115] Among them, y i represents the actual online time of the corresponding user terminal in area i, represents the actual online time of the corresponding user terminal in area i, i represents the i-th area, and n represents the number of areas.

[0116] Step 1036: Determine a second mean absolute error function MAE according to the predicted online time of each user terminal in the corresponding area at the next moment and the actual online time of each user terminal in the corresponding area at the next moment. The second mean absolute error function MAE is expressed as:

[0117]

[0118] Among them, y i represents the actual online time of the corresponding user terminal in area i, represents the actual online time of the corresponding user terminal in area i, i represents the i-th area, and n represents the number of areas.

[0119] Step 1037, based on the minimization processing of the second mean absolute error loss function loss, the recursive neural network pre-trained model is continuously trained and adjusted to obtain a trained recursive neural network pre-trained model, and the trained recursive neural network pre-trained model is used as the recursive neural network model.

[0120] Step 1038, input the mobile sequence information test set of the target historical moment into the recursive neural network model, output the test online time of each user terminal in the corresponding area at the next moment of the target historical moment, and calculate the accuracy of the recursive neural network model based on the test online time of each user terminal in the corresponding area at the next moment and the actual online time of each user terminal in the corresponding area at the next moment using the second root mean square error function RMSE and the second mean absolute error function MAE.

[0121] In the above scheme, first, use y i and Represent the actual online time and the predicted online time respectively. The loss function is the second mean absolute error loss function, as shown below:

[0122]

[0123] Where θ1 represents all trainable parameters in the recurrent neural network pre-trained model, and λ1 is a predefined hyperparameter for L2 regularization to avoid overfitting problems. L2 regularization refers to the sum of the squares of each element in the weight vector and then taking the square root.

[0124] Second, the training set is used to train the recursive neural network pre-training model, and the second mean absolute error loss function is used to guide the adjustment of model parameters. The optimal hyperparameters of the model can also be obtained through the validation set.

[0125] Third, the test set is used to evaluate the trained recurrent neural network model, and the values ​​of the second root mean square error function (RMSE) and the second mean absolute error function (MAE) of the model on the validation set are recorded.

[0126]

[0127]

[0128] In some embodiments, in step 1032, obtaining all regions where each user terminal is located at each historical moment and the actual online time in each region includes:

[0129] Step 10321, obtaining the media access control address of each user terminal at each historical moment in each area covered by the wireless local area network, the media access control address of the wireless access point in the area, and the name of the corresponding wireless access point.

[0130] Step 10322, sort and group the media access control address of each user terminal at each historical moment, the media access control address of the wireless access point in the area, and the name of the corresponding wireless access point according to the order of each historical moment to obtain multiple groups.

[0131] Step 10323, in response to determining that the name of the corresponding wireless access point in the group at the current historical moment is different from the name of the corresponding wireless access point at the next moment of the current historical moment, calculate the online time of the previous area of ​​the area where the corresponding user terminal is currently located according to the current historical moment and the next moment of the current historical moment, and obtain all the areas where each user terminal is located at each historical moment and the actual online time in each of the areas.

[0132] In the above scheme, SNMP (Simple Network Management Protocol) logs with a time span of 1 month (can be set according to specific circumstances, and are not specifically limited here) are collected from ACs. In the process of obtaining SNMP logs, a management program is executed to poll all ACs every 5 minutes (can be set according to specific circumstances, and are not specifically limited here), and the following key-value pair information is obtained from the ACs:

[0133] (1) The Media Access Control Address (MAC) of a user terminal and the Media Access Control Address (MAC) of the wireless access point (AP) to which it is associated;

[0134] (2) The Media Access Control Address (MAC) of a wireless access point (AP) and the name of the wireless access point (AP);

[0135] (3) A user terminal's Media Access Control Address (MAC) and the IP address SNMP log obtained from the DHCP server are used to determine the location of each user terminal.

[0136] The key-value pair information obtained from the AC is converted into a record in the following format {date, timestamp, MAC address of the user terminal, AP name}.

[0137] The media access control address of each user terminal, the media access control address of the wireless access point in the area and the name of the corresponding wireless access point are grouped according to (user equipment MAC address, date);

[0138] Then, sort the records in each group in timestamp order, traverse all records, and obtain multiple groups. When the area represented by the AP name prefix changes, it means that the user has moved to another area. The online time of the user's device in the previous area is calculated based on the timestamp;

[0139] Finally, a sequence is obtained, which records all the areas visited by individual user terminals every day and the online time in each area. This dataset is used as all the areas where each user terminal is located at each historical moment and the actual online time in each area for the training and performance evaluation of the recursive neural network model.

[0140] In some embodiments, in step 103, the recursive neural network model includes:

[0141] Embedding layer, recurrent neural network layer, and attention layer;

[0142] The embedding layer is used to convert the historical movement sequence information of the movement into a low-dimensional dense vector;

[0143] The recurrent neural network layer is used to learn the dependencies embedded in the encoded historical movement sequence information;

[0144] The attention layer is used to learn long-term dependencies in historical movement sequence information.

[0145] In the above scheme, the specific structure of the recursive neural network model is as follows: it consists of an embedding layer, an RNN (Recurrent Neural Networks) layer and an attention layer.

[0146] First, the training set extracted from the SNMP log is input into the recursive neural network model. Each data in the data set is a time series. Through the embedding layer, each high-dimensional feature vector e represented by one-hot (unique hot encoding) in the sequence is converted into a low-dimensional feature vector for learning more semantic information. The specific process is as follows:

[0147] (1) Define a semantic point q = (v, Δt) to represent the online time Δt of a user in area v. For a given user ID u, the mobile trajectory sequence in the wireless LAN is a series of semantic points S u ={q1, q2, ..., q n}. For each sequence {q1, q2, ..., q n}, where v and Δt in the semantic point q = (v, Δt) are represented by one-hot vectors.

[0148] (2) For an original input node v represented by an ont-hot vector, use the matrix Convert the vector v to a lower-dimensional vector e v , where D v is the region embedding code v The dimension of e v =v·E v .

[0149] (3) Similarly, for an original input node Δt represented by an ont-hot vector, the matrix Convert the vector Δt to a low-dimensional vector e Δt , where D Δt is online time embedding coding Δt The dimension of e Δt =Δt·E Δt .

[0150] (4) Then the embedded code e v and e Δt , connected into a vector e i , that is, e i =(e v ;e Δt ), which represents the semantic point q i Through the above operations, each semantic point of the input sequence is converted into a corresponding vector as the output of this layer.

[0151] Second, use the RNN layer, taking the output of the previous layer as input, to learn the dependencies in the embedded encoded input sequence. The specific process is as follows:

[0152] For a sequence of length 1 {q τ-l+1 ,…,q τ}, the RNN layer consists of 1 recurrent unit. The i-th unit is represented by point e τ-l+i The embedded code is taken as input and the hidden state is calculated as follows:

[0153] h τ-l+i =GRU(h τ-l+(i-1) , e τ-l+i )

[0154] where h τ-l+i Represents the hidden state of unit i. Each GRU unit is constructed by combining the previous hidden state h τ-l+i and the current input to learn the sequence pattern.

[0155] Finally, the last hidden state h τ as the final representation because it utilizes all the information of the input sequence.

[0156] Third, use the attention layer to take the hidden state sequence output by the previous layer as input and calculate the weight of each hidden state. The specific process is as follows:

[0157] (1) The attention layer calculates the weight for each hidden state, here the last hidden state h τ The calculation is explained. The normalized similarity score between each previous hidden state and the last state is calculated, expressed as:

[0158] α τ =Normalize([h τ-l+1 h τ ,…,h τ g τ ])

[0159] Among them, Normalize(·) is a softmax (normalized exponential function) activation function, for α at position k τ The calculation expression is α τ is the similarity score vector between the previous state and the last state.

[0160] (2) The final context feature vector c τ Equivalent to the weighted average of the hidden states, expressed as By context c τ , the feed-forward layer finally produces high-quality online time prediction results, that is, the user's online time Here the feed-forward layer is a fully connected network with two linear layers.

[0161] Step 104: setting the address pool capacity of each area according to the predicted number of user terminals at the next moment.

[0162] In this step, the IP address pool is configured. The wireless LAN is usually divided into multiple areas, each of which is managed by an AC. DHCP allocates an IP address pool to the AC of each area, which consists of multiple subnet segments. The IP pool size (i.e., address pool capacity) of each area is set according to the predicted number of user terminals at the next moment, and the corresponding number of subnet segments is allocated to each area.

[0163] Step 105: Allocate an address lease time for the corresponding user terminal according to the predicted online time of each user terminal in the corresponding area.

[0164] In this step, when the user terminal connects to the AC that manages the current area for the first time through an AP (Access Point), the DHCP server assigns an IP lease time to the wireless access point, and assigns an address lease time to the corresponding user terminal according to the predicted online time of each user terminal in the corresponding area. If the user terminal is still in the LAN managed by the same AC, when the user's lease time is halfway through, the DHCP server still assigns the same lease time to the user.

[0165] Through the above scheme, the predicted number of user terminals at the next moment in each area is obtained according to the number of user terminals at the current moment through a pre-trained spatiotemporal graph neural network model, and then the predicted online time of each user terminal in the corresponding area is obtained according to the historical mobile sequence information through a pre-trained recursive neural network model. Finally, the address pool capacity of each area is set according to the predicted number of user terminals at the next moment, and the address lease time is allocated to the corresponding user terminal according to the predicted online time of each user terminal in the corresponding area. The recursive neural network model is used to accurately predict the user's online time, thereby ensuring that the address lease time is effectively allocated to the individual user terminal, and the behavioral characteristics of the individual user can be fully utilized. The spatiotemporal graph neural network model is also used to capture the spatiotemporal dependence of the movement of group user terminals, and the number of group user terminals in each area is accurately predicted, thereby providing effective address lease time and address pool capacity to ensure the high performance of DHCP.

[0166] It should be noted that the method of the embodiment of the present application can be performed by a single device, such as a computer or server. The method of this embodiment can also be applied to a distributed scenario and completed by multiple devices cooperating with each other. In the case of such a distributed scenario, one of the multiple devices can only perform one or more steps in the method of the embodiment of the present application, and the multiple devices will interact with each other to complete the described method.

[0167] It should be noted that the above describes some embodiments of the present application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the above embodiments and still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0168] Based on the same inventive concept, corresponding to any of the above-mentioned embodiment methods, the present application also provides an address configuration device.

[0169] refer to Figure 2 , the address configuration device comprises:

[0170] The user terminal quantity acquisition module 201 is configured to acquire the number of user terminals at the current moment in each area covered by the target local area network;

[0171] The user terminal quantity prediction module 202 is configured to input the number of user terminals at the current moment into a pre-trained spatiotemporal graph neural network model to obtain the predicted number of user terminals in each area at the next moment;

[0172] The online time prediction module 203 is configured to obtain historical movement sequence information of each user terminal, input the historical movement sequence information into a pre-trained recursive neural network model, and obtain the predicted online time of each user terminal in the corresponding area;

[0173] An address pool capacity setting module 204, configured to set the address pool capacity of each area according to the predicted number of user terminals at the next moment;

[0174] The address lease time allocation module 205 is configured to allocate an address lease time to a corresponding user terminal according to the predicted online time of each user terminal in the corresponding area.

[0175] In some embodiments, the address configuration device further includes a spatiotemporal graph neural network model training module, and the spatiotemporal graph neural network model training module includes:

[0176] A first pre-training model building unit is configured to build a spatiotemporal graph neural network pre-training model;

[0177] A first acquisition unit is configured to acquire the actual number of user terminals at each historical moment in each area covered by the wireless local area network, and divide the actual number of user terminals at each historical moment into a user terminal number training set and a user terminal number test set;

[0178] A first input unit is configured to input a training set of the number of user terminals at a target historical moment into the spatiotemporal graph neural network pre-training model, and output a predicted number of user terminals at the next moment of the target historical moment in each region, wherein the target historical moment is any moment among the historical moments;

[0179] The first loss function determination unit is configured to determine a first mean absolute error loss function loss according to the predicted number of user terminals at the next moment and the actual number of user terminals at the next moment, wherein the first mean absolute error loss function loss is expressed as:

[0180]

[0181] Where V represents the area covered by the wireless local area network, i represents the i-th area, and Y i Indicates area V i The actual number of user terminals in Indicates area V i The number of user terminals is predicted in , θ2 represents all trainable and adjustable parameters in the spatiotemporal graph neural network pre-training model, and λ2 is expressed as a predefined hyperparameter;

[0182] The first root mean square error function determining unit is configured to determine a first root mean square error function RMSE according to the predicted number of user terminals at the next moment and the actual number of user terminals at the next moment, wherein the first root mean square error function RMSE is expressed as:

[0183]

[0184] Among them, Y i Indicates area V i The actual number of user terminals in Indicates area V i The number of predicted user terminals in , i represents the i-th region, and n represents the number of regions;

[0185] The first mean absolute error function determining unit is configured to determine a first mean absolute error function MAE according to the predicted number of user terminals at the next moment and the actual number of user terminals at the next moment, wherein the first mean absolute error function MAE is expressed as:

[0186]

[0187] Among them, Y i Indicates area V i The actual number of user terminals in Indicates area V i The number of predicted user terminals in , i represents the i-th region, and n represents the number of regions;

[0188] A first training adjustment unit is configured to perform minimization processing based on the first mean absolute error loss function loss, continuously perform training adjustment on the spatiotemporal graph neural network pre-training model, obtain a trained spatiotemporal graph neural network pre-training model, and use the trained spatiotemporal graph neural network pre-training model as the spatiotemporal graph neural network model;

[0189] The first test unit is configured to input a test set of the number of user terminals at a target historical moment into the spatiotemporal graphical neural network model, output the number of test user terminals at the next moment of the target historical moment in each area, and calculate the accuracy of the spatiotemporal graphical neural network model based on the number of test user terminals at the next moment and the actual number of user terminals at the next moment using the first root mean square error function RMSE and the first mean absolute error function MAE.

[0190] In some embodiments, the first acquisition unit is specifically configured to:

[0191] Acquire the media access control address of each user terminal in each area covered by the wireless local area network at each historical moment, the media access control address of the wireless access point in the area, and the name of the corresponding wireless access point;

[0192] The media access control address of each user terminal in each area covered by the wireless local area network at each historical moment, the media access control address of the wireless access point in the area, and the name of the corresponding wireless access point are grouped according to a preset time to obtain a plurality of first groups;

[0193] Grouping each of the first groups according to the name of the wireless access point to obtain a plurality of second groups;

[0194] The number of media access control addresses of each user terminal in the second group with the same name as each wireless access point is counted to obtain the actual number of user terminals at each historical moment in each area covered by the wireless local area network.

[0195] In some embodiments, the address configuration device further includes a recursive neural network model training module, and the recursive neural network model training module includes:

[0196] A second pre-training model building unit is configured to build a recursive neural network pre-training model;

[0197] A second acquisition unit is configured to acquire all areas where each user terminal is located at each historical moment and the actual online time in each area, use all areas where each user terminal is located at each historical moment and the actual online time in each area as historical actual movement sequence information, and divide the historical actual movement sequence information into a movement sequence information training set and a movement sequence information test set;

[0198] A second input unit is configured to input a training set of mobile sequence information of a target historical moment into the recursive neural network pre-training model, and output a predicted online time of each user terminal in a corresponding area at a next moment of the target historical moment, wherein the target historical moment is any moment among the historical moments;

[0199] The second loss function determination unit is configured to determine a second mean absolute error loss function loss according to the predicted online time of each user terminal in the corresponding area at the next moment and the actual online time of each user terminal in the corresponding area at the next moment, wherein the second mean absolute error loss function loss is expressed as:

[0200]

[0201] Among them, y i represents the actual online time of the corresponding user terminal in area i, represents the actual online time of the corresponding user terminal in region i, i represents the i-th region, θ1 represents all trainable and adjustable parameters in the recursive neural network pre-training model, and λ2 is expressed as a predefined hyperparameter;

[0202] The second root mean square error function determining unit is configured to determine a second root mean square error function RMSE according to the predicted online time of each user terminal in the corresponding area at the next moment and the actual online time of each user terminal in the corresponding area at the next moment, wherein the second root mean square error function RMSE is expressed as:

[0203]

[0204] Among them, y i represents the actual online time of the corresponding user terminal in area i, represents the actual online time of the corresponding user terminal in area i, i represents the i-th area, and n represents the number of areas;

[0205] The second mean absolute error function determining unit is configured to determine a second mean absolute error function MAE according to the predicted online time of each user terminal in the corresponding area at the next moment and the actual online time of each user terminal in the corresponding area at the next moment, wherein the second mean absolute error function MAE is expressed as:

[0206]

[0207] Among them, y i represents the actual online time of the corresponding user terminal in area i, represents the actual online time of the corresponding user terminal in area i, i represents the i-th area, and n represents the number of areas;

[0208] A second training adjustment unit is configured to perform minimization processing based on the second mean absolute error loss function loss, continuously perform training adjustment on the recursive neural network pre-trained model, obtain a trained recursive neural network pre-trained model, and use the trained recursive neural network pre-trained model as the recursive neural network model;

[0209] The second testing unit is configured to input a mobile sequence information test set of a target historical moment into the recursive neural network model, output a test online time of each user terminal in a corresponding area at a next moment of the target historical moment, and calculate the accuracy of the recursive neural network model using the second root mean square error function RMSE and the second mean absolute error function MAE based on the test online time of each user terminal in the corresponding area at the next moment and the actual online time of each user terminal in the corresponding area at the next moment.

[0210] In some embodiments, the second acquisition unit is specifically configured to:

[0211] Acquire the media access control address of each user terminal in each area covered by the wireless local area network at each historical moment, the media access control address of the wireless access point in the area, and the name of the corresponding wireless access point;

[0212] Sorting and grouping the media access control address of each user terminal at each historical moment, the media access control address of the wireless access point in the area where the user terminal is located, and the name of the corresponding wireless access point according to the order of each historical moment to obtain multiple groups;

[0213] In response to determining that the name of the corresponding wireless access point in the group at the current historical moment is different from the name of the corresponding wireless access point at a moment next to the current historical moment, the online time of the previous area of ​​the area where the corresponding user terminal is currently located is calculated according to the current historical moment and the moment next to the current historical moment, and all areas where each user terminal is located at each historical moment and the actual online time in each of the areas are obtained.

[0214] In some embodiments, the spatiotemporal graphical neural network model includes:

[0215] Graph convolutional network layers, recurrent neural network layers, and attention layers;

[0216] The graph convolutional network layer is used to obtain the local topological features of the number of user terminals at each moment;

[0217] The recursive neural network layer is used to learn local time dependencies in the time series of the number of user terminals;

[0218] The attention layer is used to learn periodic information in dynamic networks.

[0219] In some embodiments, the recurrent neural network model includes:

[0220] Embedding layer, recurrent neural network layer, and attention layer;

[0221] The embedding layer is used to convert the historical movement sequence information of the movement into a low-dimensional dense vector;

[0222] The recurrent neural network layer is used to learn the dependencies embedded in the encoded historical movement sequence information;

[0223] The attention layer is used to learn long-term dependencies in historical movement sequence information.

[0224] For the convenience of description, the above device is described in terms of functions divided into various modules. Of course, when implementing the present application, the functions of each module can be implemented in the same or multiple software and / or hardware.

[0225] The device of the above embodiment is used to implement the corresponding address configuration method in any of the above embodiments, and has the beneficial effects of the corresponding method embodiment, which will not be described in detail here.

[0226] Based on the same inventive concept, corresponding to any of the above-mentioned embodiments and methods, the present application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the address configuration method described in any of the above embodiments is implemented.

[0227] Figure 3 A more specific schematic diagram of the hardware structure of an electronic device provided in this embodiment is shown, and the device may include: a processor 301, a memory 302, an input / output interface 303, a communication interface 304, and a bus 305. The processor 301, the memory 302, the input / output interface 303, and the communication interface 304 are connected to each other in communication within the device through the bus 305.

[0228] The processor 301 can be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.

[0229] The memory 302 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 302 can store an operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented by software or firmware, the relevant program codes are stored in the memory 302 and called and executed by the processor 301.

[0230] The input / output interface 303 is used to connect the input / output module to realize information input and output. The input / output module can be configured in the device as a component (not shown in the figure), or it can be externally connected to the device to provide corresponding functions. The input device can include a keyboard, a mouse, a touch screen, a microphone, various sensors, etc., and the output device can include a display, a speaker, a vibrator, an indicator light, etc.

[0231] The communication interface 304 is used to connect a communication module (not shown) to realize communication interaction between the device and other devices. The communication module can realize communication through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0232] The bus 305 comprises a pathway for transmitting information between the various components of the device (eg, the processor 301 , the memory 302 , the input / output interface 303 , and the communication interface 304 ).

[0233] It should be noted that, although the above device only shows the processor 301, the memory 302, the input / output interface 303, the communication interface 304 and the bus 305, in the specific implementation process, the device may also include other components necessary for normal operation. In addition, it can be understood by those skilled in the art that the above device may also only include the components necessary for implementing the embodiments of the present specification, and does not necessarily include all the components shown in the figure.

[0234] The electronic device of the above embodiment is used to implement the corresponding address configuration method in any of the above embodiments, and has the beneficial effects of the corresponding method embodiment, which will not be described in detail here.

[0235] Based on the same inventive concept, corresponding to any of the above-mentioned embodiment methods, the present application also provides a non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores computer instructions, and the computer instructions are used to enable the computer to execute the address configuration method described in any of the above embodiments.

[0236] The computer-readable medium of this embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, read-only compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device.

[0237] The computer instructions stored in the storage medium of the above embodiments are used to enable the computer to execute the address configuration method described in any of the above embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0238] Those skilled in the art should understand that the discussion of any of the above embodiments is merely illustrative and is not intended to imply that the scope of the present application (including the claims) is limited to these examples. In line with the concept of the present application, the technical features in the above embodiments or different embodiments may be combined, the steps may be implemented in any order, and there are many other variations of the different aspects of the embodiments of the present application as described above, which are not provided in detail for the sake of simplicity.

[0239] In addition, to simplify the description and discussion, and in order not to make the embodiments of the present application difficult to understand, the known power supply / ground connection with the integrated circuit (IC) chip and other components may or may not be shown in the provided drawings. In addition, the device can be shown in the form of a block diagram to avoid making the embodiments of the present application difficult to understand, and this also takes into account the fact that the details of the implementation of these block diagram devices are highly dependent on the platform to be implemented in the embodiments of the present application (that is, these details should be fully within the scope of understanding of those skilled in the art). In the case of elaborating specific details (e.g., circuits) to describe exemplary embodiments of the present application, it is obvious to those skilled in the art that the embodiments of the present application can be implemented without these specific details or when these specific details are changed. Therefore, these descriptions should be considered to be illustrative rather than restrictive.

[0240] Although the present application has been described in conjunction with specific embodiments of the present application, many replacements, modifications and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may use the embodiments discussed.

[0241] The embodiments of the present application are intended to cover all such substitutions, modifications and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the embodiments of the present application should be included in the scope of protection of the present application.

Claims

1. An address configuration method, characterized in that: include: Obtain the number of user terminals at the current moment in each area covered by the target local area network; Inputting the number of user terminals at the current moment into a pre-trained spatiotemporal graph neural network model to obtain the predicted number of user terminals in each area at the next moment; Acquire historical movement sequence information of each user terminal, input the historical movement sequence information into a pre-trained recursive neural network model, and obtain a predicted online time of each user terminal in a corresponding area; Setting the address pool capacity of each area according to the predicted number of user terminals at the next moment; Allocating an address lease time for a corresponding user terminal according to the predicted online time of each user terminal in the corresponding area; The process of obtaining the recursive neural network model includes: Build a recursive neural network pre-training model; Acquire all areas where each user terminal is located at each historical moment and the actual online time in each area, use all areas where each user terminal is located at each historical moment and the actual online time in each area as historical mobile sequence information, and divide the historical mobile sequence information into a mobile sequence information training set and a mobile sequence information test set; Inputting the mobile sequence information training set of the target historical moment into the recursive neural network pre-training model, and outputting the predicted online time of each user terminal in the corresponding area at the next moment of the target historical moment, wherein the target historical moment is any moment among the historical moments; Determine a second mean absolute error loss function according to the predicted online time of each user terminal in the corresponding area at the next moment and the actual online time of each user terminal in the corresponding area at the next moment , the second mean absolute error loss function Expressed as: in, Indicates that the corresponding user terminal is in area The actual online time in Indicates that the corresponding user terminal is in area The actual online time in Indicates Regions, Represents all trainable and adjustable parameters in the recursive neural network pre-training model, expressed as predefined hyperparameters; Determine a second root mean square error function according to the predicted online time of each user terminal in the corresponding area at the next moment and the actual online time of each user terminal in the corresponding area at the next moment , the second root mean square error function Expressed as: in, Indicates that the corresponding user terminal is in area The actual online time in Indicates that the corresponding user terminal is in area The actual online time in Indicates Regions, Indicates the number of regions; Determine a second mean absolute error function according to the predicted online time of each user terminal in the corresponding area at the next moment and the actual online time of each user terminal in the corresponding area at the next moment , the second mean absolute error function Expressed as: in, Indicates that the corresponding user terminal is in area The actual online time in Indicates that the corresponding user terminal is in area The actual online time in Indicates Regions, Indicates the number of regions; Based on the second mean absolute error loss function Performing minimization processing, continuously training and adjusting the recursive neural network pre-training model to obtain a trained recursive neural network pre-training model, and using the trained recursive neural network pre-training model as the recursive neural network model; The mobile sequence information test set of the target historical moment is input into the recursive neural network model, the test online time of each user terminal in the corresponding area at the next moment of the target historical moment is output, and the second root mean square error function is used to calculate the test online time of each user terminal in the corresponding area at the next moment and the actual online time of each user terminal in the corresponding area at the next moment. and the second mean absolute error function Calculate the accuracy of the recursive neural network model; The obtaining of all regions where each user terminal is located at each historical moment and the actual online time in each region includes: Obtaining the media access control address of each user terminal in each area covered by the wireless local area network at each historical moment, the media access control address of the wireless access point in the area, and the name of the corresponding wireless access point; Sorting and grouping the media access control address of each user terminal at each historical moment, the media access control address of the wireless access point in the area where the user terminal is located, and the name of the corresponding wireless access point according to the order of each historical moment to obtain multiple groups; In response to determining that the name of the corresponding wireless access point in the group at the current historical moment is different from the name of the corresponding wireless access point at a moment next to the current historical moment, the online time of the previous area of ​​the area where the corresponding user terminal is currently located is calculated according to the current historical moment and the moment next to the current historical moment, so as to obtain all the areas where each user terminal is located at each historical moment and the actual online time in each of the areas.

2. The method according to claim 1, characterized in that The process of obtaining the spatiotemporal graphical neural network model includes: Build a spatiotemporal graph neural network pre-training model; Acquire the actual number of user terminals at each historical moment in each area covered by the wireless local area network, and divide the actual number of user terminals at each historical moment into a user terminal number training set and a user terminal number test set; Inputting a training set of the number of user terminals at a target historical moment into the spatiotemporal graph neural network pre-training model, and outputting a predicted number of user terminals at the next moment of the target historical moment in each region, wherein the target historical moment is any moment among the historical moments; Determine a first mean absolute error loss function according to the predicted number of user terminals at the next moment and the actual number of user terminals at the next moment , the first mean absolute error loss function Expressed as: in, Indicates the area covered by the wireless LAN. Indicates Regions, Indicates area The actual number of user terminals in Indicates area The number of user terminals is predicted in Represents all trainable and adjustable parameters in the spatiotemporal graph neural network pre-training model, expressed as predefined hyperparameters; Determine a first root mean square error function according to the predicted number of user terminals at the next moment and the actual number of user terminals at the next moment , the first root mean square error function Expressed as: in, Indicates area The actual number of user terminals in Indicates area The number of user terminals is predicted in Indicates Regions, Indicates the number of regions; Determine a first mean absolute error function according to the predicted number of user terminals at the next moment and the actual number of user terminals at the next moment , the first mean absolute error function Expressed as: in, Indicates area The actual number of user terminals in Indicates area The number of user terminals is predicted in Indicates Regions, Indicates the number of regions; Based on the first mean absolute error loss function Performing minimization processing, continuously training and adjusting the spatiotemporal graph neural network pre-training model to obtain a trained spatiotemporal graph neural network pre-training model, and using the trained spatiotemporal graph neural network pre-training model as the spatiotemporal graph neural network model; The test set of the number of user terminals at the target historical moment is input into the spatiotemporal graphical neural network model, the number of test user terminals at the next moment of the target historical moment in each region is output, and the first root mean square error function is used to respectively calculate the number of test user terminals at the next moment and the actual number of user terminals at the next moment. and the first mean absolute error function Calculate the accuracy of the spatiotemporal graph neural network model.

3. The method according to claim 2, characterized in that The obtaining of the actual number of user terminals at each historical moment in each area covered by the wireless local area network includes: Acquire the media access control address of each user terminal in each area covered by the wireless local area network at each historical moment, the media access control address of the wireless access point in the area, and the name of the corresponding wireless access point; The media access control address of each user terminal in each area covered by the wireless local area network at each historical moment, the media access control address of the wireless access point in the area, and the name of the corresponding wireless access point are grouped according to a preset time to obtain a plurality of first groups; Grouping each of the first groups according to the name of the wireless access point to obtain a plurality of second groups; The number of media access control addresses of each user terminal in the second group with the same name as each wireless access point is counted to obtain the actual number of user terminals at each historical moment in each area covered by the wireless local area network.

4. The method according to claim 1, characterized in that: The spatiotemporal graph neural network model includes: Graph convolutional network layers, recurrent neural network layers, and attention layers; The graph convolutional network layer is used to obtain the local topological features of the number of user terminals at each moment; The recursive neural network layer is used to learn local time dependencies in the time series of the number of user terminals; The attention layer is used to learn periodic information in dynamic networks.

5. The method according to claim 1, characterized in that: The recursive neural network model includes: Embedding layer, recurrent neural network layer, and attention layer; The embedding layer is used to convert the historical movement sequence information of the movement into a low-dimensional dense vector; The recurrent neural network layer is used to learn the dependencies embedded in the encoded historical movement sequence information; The attention layer is used to learn long-term dependencies in historical movement sequence information.

6. An address configuration device, characterized in that: include: A user terminal quantity acquisition module is configured to acquire the number of user terminals at the current moment in each area covered by the target local area network; A user terminal quantity prediction module is configured to input the number of user terminals at the current moment into a pre-trained spatiotemporal graph neural network model to obtain a predicted number of user terminals in each area at the next moment; An online time prediction module is configured to obtain historical movement sequence information of each user terminal, input the historical movement sequence information into a pre-trained recursive neural network model, and obtain a predicted online time of each user terminal in a corresponding area; The process of obtaining the recursive neural network model includes: constructing a recursive neural network pre-training model; obtaining all the areas where each user terminal is located at each historical moment and the actual online time in each of the areas, taking all the areas where the terminal is located at each historical moment and the actual online time in each of the areas as historical mobile sequence information, and dividing the historical mobile sequence information into a mobile sequence information training set and a mobile sequence information test set; inputting the mobile sequence information training set of the target historical moment into the recursive neural network pre-training model, and outputting the predicted online time of each user terminal in the corresponding area at the next moment of the target historical moment, wherein the target historical moment is any moment in the historical moments; determining a second mean absolute error loss function according to the predicted online time of each user terminal in the corresponding area at the next moment and the actual online time of each user terminal in the corresponding area at the next moment. , the second mean absolute error loss function Expressed as: in, Indicates that the corresponding user terminal is in area The actual online time in Indicates that the corresponding user terminal is in area The actual online time in Indicates Regions, Represents all trainable and adjustable parameters in the recursive neural network pre-training model, expressed as predefined hyperparameters; Determine a second root mean square error function according to the predicted online time of each user terminal in the corresponding area at the next moment and the actual online time of each user terminal in the corresponding area at the next moment , the second root mean square error function Expressed as: in, Indicates that the corresponding user terminal is in area The actual online time in Indicates that the corresponding user terminal is in area The actual online time in Indicates Regions, Indicates the number of regions; Determine a second mean absolute error function according to the predicted online time of each user terminal in the corresponding area at the next moment and the actual online time of each user terminal in the corresponding area at the next moment , the second mean absolute error function Expressed as: in, Indicates that the corresponding user terminal is in area The actual online time in Indicates that the corresponding user terminal is in area The actual online time in Indicates Regions, Indicates the number of regions; Based on the second mean absolute error loss function Performing minimization processing, continuously training and adjusting the recursive neural network pre-training model to obtain a trained recursive neural network pre-training model, and using the trained recursive neural network pre-training model as the recursive neural network model; The mobile sequence information test set of the target historical moment is input into the recursive neural network model, the test online time of each user terminal in the corresponding area at the next moment of the target historical moment is output, and the second root mean square error function is used to calculate the test online time of each user terminal in the corresponding area at the next moment and the actual online time of each user terminal in the corresponding area at the next moment. and the second mean absolute error function Calculating the accuracy of the recursive neural network model; obtaining the entire area where each user terminal is located at each historical moment and the actual online time in each of the areas, specifically for obtaining the media access control address of each user terminal at each historical moment, the media access control address of the wireless access point in the area, and the name of the corresponding wireless access point in each area covered by the wireless local area network; sorting and grouping the media access control address of each user terminal at each historical moment, the media access control address of the wireless access point in the area, and the name of the corresponding wireless access point in the order of each historical moment to obtain a plurality of groups; in response to determining that the name of the corresponding wireless access point at the current historical moment in the group is different from the name of the corresponding wireless access point at the next moment of the current historical moment, calculating the online time of the previous area of ​​the area where the corresponding user terminal is currently located according to the current historical moment and the next moment of the current historical moment, and obtaining the entire area where each user terminal is located at each historical moment and the actual online time in each of the areas; An address pool capacity setting module, configured to set the address pool capacity of each area according to the predicted number of user terminals at the next moment; The address lease time allocation module is configured to allocate an address lease time to a corresponding user terminal according to the predicted online time of each user terminal in the corresponding area.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the method according to any one of claims 1 to 5 is implemented.

8. A non-transitory computer-readable storage medium storing computer instructions, characterized in that: The computer instructions are used to enable a computer to execute the method according to any one of claims 1 to 5.

Images