Method and system for optimizing wireless network based on tree convolutional network model
Through the tree convolutional network model, the problem of independent management in the wireless network is solved, efficient network optimization and troubleshooting is achieved, and network stability and resource utilization are improved.
Patent Information
- Application Number
- CN202510894479.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-30
- Publication Date
- 2025-09-02
AI Technical Summary
Traffic prediction and fault diagnosis in existing wireless networks are independent of each other and cannot be effectively integrated, resulting in difficulty in management and maintenance. Traditional methods have shortcomings in data processing and computing resources, making it difficult to adapt to complex modes and high-dimensional nonlinear features.
The tree convolution network model is adopted, including the global feature extraction module, the time convolution prediction module and the tree convolution diagnosis module. By constructing graph structure and data processing technology, the fusion optimization of traffic prediction and fault diagnosis is achieved.
It realizes high-precision traffic prediction and fault diagnosis, improves network resource utilization and operation efficiency, and enhances the ability to adapt to complex scenarios and dynamic changes.
Smart Images
Figure CN120583449A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a method and system for optimizing a wireless network based on a tree convolutional network model. Background Art
[0002] With the expansion of wireless network scale and the increase of technical complexity, wireless network management and maintenance face severe challenges. The traffic prediction and fault diagnosis of existing wireless networks are independent of each other, and it is impossible to effectively integrate traffic prediction and fault diagnosis to achieve efficient optimization of wireless networks.
[0003] In terms of traffic prediction, traditional machine learning methods rely on manually designed features and are difficult to adapt to the complex patterns and high-dimensional nonlinear characteristics of encrypted traffic. The classification accuracy continues to decline. Although fault diagnosis methods based on supervised learning have certain effects, they face problems such as difficulty in obtaining labeled data, differences between simulated faults and real situations, high resource consumption, and lack of real-time performance.
[0004] In terms of fault diagnosis, convolutional neural networks (CNNs) have problems such as time-consuming data collection and preprocessing, dependence on large amounts of labeled data, high computing resource requirements, and insufficient generalization capabilities. Graph convolutional networks (GCNs) face limitations such as high computational complexity, easy oversmoothing, difficulty in processing dynamic graphs, and strong dependence on labeled data.
[0005] Therefore, there is an urgent need for a method that can effectively integrate traffic prediction and fault diagnosis to achieve efficient optimization of wireless networks. Summary of the Invention
[0006] In view of this, the present invention proposes a method and system for optimizing wireless networks based on a tree convolutional network model, which can effectively integrate traffic prediction and fault diagnosis to achieve efficient optimization of wireless networks.
[0007] To achieve the above object, the present invention provides the following technical solutions:
[0008] A method for optimizing a wireless network based on a tree convolutional network model, comprising:
[0009] Collect historical wireless network data;
[0010] Constructing a tree convolutional network model, the tree convolutional network model includes a global feature extraction module, a temporal convolution prediction module, and a tree convolution diagnosis module, wherein the global feature extraction module is used to extract global features, the temporal convolution prediction module is used to perform traffic prediction of wireless network data based on the global features to obtain a traffic prediction result, and the tree convolution diagnosis module is used to perform fault diagnosis of the wireless network based on the global features to obtain a fault diagnosis result;
[0011] Inputting the historical wireless network data into the tree convolutional network model for training to obtain an optimal tree convolutional network model;
[0012] Real-time wireless network data is acquired, and the real-time wireless network data is input into the optimal tree convolutional network model to obtain traffic prediction results and fault diagnosis results of the real-time wireless network data, and the wireless network is optimized based on the traffic prediction results and fault diagnosis results.
[0013] On the basis of the above technical solution, the present invention can also be improved as follows:
[0014] Optionally, the global feature extraction module includes a convolutional layer, a spatiotemporal convolutional layer and a fully connected layer;
[0015] The convolutional layer is used to extract local features of wireless network data;
[0016] The spatiotemporal convolution layer is used to extract spatiotemporal features of wireless network data;
[0017] The fully connected layer is used to integrate the local features and the spatiotemporal features to generate global features.
[0018] Optionally, the temporal convolution prediction module includes a first spatiotemporal convolution layer, a second spatiotemporal convolution layer and a fully connected layer;
[0019] The first spatiotemporal convolutional layer is used to capture the spatiotemporal features related to traffic prediction in the global features;
[0020] The second spatiotemporal convolutional layer is used to capture key spatiotemporal features related to traffic prediction;
[0021] The fully connected layer is used to integrate the spatiotemporal features and the key spatiotemporal features to generate a traffic prediction result of the wireless network data.
[0022] Optionally, the method for optimizing a wireless network based on a tree convolutional network model further includes:
[0023] Based on the topological structure and node connection relationship of the wireless network, a graph structure represented by the adjacency matrix is constructed to obtain the connection status between each node in the wireless network.
[0024] Optionally, the tree convolution diagnosis module includes a first tree convolution layer, a spatiotemporal convolution layer, a second tree convolution layer and a fully connected layer;
[0025] The first tree convolution layer is used to convert the adjacency matrix into a tree matrix, and perform preliminary feature extraction on the tree matrix through a convolution operation to obtain preliminary features;
[0026] The spatiotemporal convolution layer is used to capture the spatiotemporal features of the global features;
[0027] The second tree convolution layer is used to perform feature extraction on the tree matrix based on the spatiotemporal features to obtain in-depth features;
[0028] The fully connected layer is used to integrate the preliminary features, spatiotemporal features and the in-depth features to generate a fault diagnosis result of the wireless network.
[0029] Optionally, the first tree convolution layer is used to convert the adjacency matrix into a tree matrix, including:
[0030] Perform a breadth-first traversal on the graph structure represented by the adjacency matrix to obtain a tree structure. When the resulting tree structure contains a loop, use a spatial tree matrix to represent the original tree structure. The tree structure is divided into different subtrees. The maximum number of layers, nodes, and child nodes corresponding to each subtree is calculated. Starting from the specified node of each subtree, all possible paths are found.
[0031] Convert all possible paths into a tree matrix.
[0032] Optionally, before the step of inputting the historical wireless network data into the tree convolutional network model for training, the method further includes:
[0033] Identifying null values and abnormal values in the historical wireless network data, and replacing the null values and abnormal values with 0;
[0034] Identify features of missing values in the historical wireless network data, and fill in the missing values based on average values of time series features corresponding to the features to ensure the integrity and availability of the historical wireless network data;
[0035] Kernel density estimation is used to perform data enhancement on the historical wireless network data.
[0036] A system for optimizing wireless networks based on a tree convolutional network model, comprising:
[0037] Collection module, used to collect historical wireless network data;
[0038] A construction module is used to construct a tree convolutional network model, wherein the tree convolutional network model includes a global feature extraction module, a temporal convolution prediction module, and a tree convolution diagnosis module. The global feature extraction module is used to extract global features. The temporal convolution prediction module is used to perform traffic prediction of wireless network data based on the global features to obtain traffic prediction results. The tree convolution diagnosis module is used to perform fault diagnosis of the wireless network based on the global features to obtain fault diagnosis results.
[0039] A training module, configured to input the historical wireless network data into the tree convolutional network model for training to obtain an optimal tree convolutional network model;
[0040] The optimization module is used to obtain real-time wireless network data, input the real-time wireless network data into the optimal tree convolutional network model to obtain traffic prediction results and fault diagnosis results of the real-time wireless network data, and optimize the wireless network based on the traffic prediction results and fault diagnosis results.
[0041] An electronic device comprises a memory, a processor and a computer program stored in the memory and running on the processor, wherein the steps of the method are implemented when the processor executes the computer program.
[0042] A non-transitory computer-readable storage medium stores a computer program, which implements the steps of the method when executed by a processor.
[0043] The present invention has the following advantages:
[0044] The method for optimizing wireless networks based on a tree convolutional network model in the present invention, by constructing a tree convolutional network model including global feature extraction, traffic prediction, and fault diagnosis modules, can accurately extract features from wireless network data, effectively capture the temporal and spatial characteristics of network data, comprehensively and deeply analyze network conditions, and achieve high-precision traffic prediction and fault diagnosis. Based on accurate traffic prediction and fault diagnosis results, the wireless network can be optimized, resources can be allocated and faults can be eliminated in a targeted manner, network resource utilization can be improved, network operation efficiency and stability can be enhanced, and the ability of wireless networks to cope with complex scenarios and dynamic changes can be enhanced. BRIEF DESCRIPTION OF THE DRAWINGS
[0045] For purposes of illustration and not limitation, the present invention will now be described with reference to embodiments thereof and the accompanying drawings, in which:
[0046] Figure 1 Schematic diagram of a flow chart of a method for optimizing a wireless network based on a tree convolutional network model in an embodiment of the present invention;
[0047] Figure 2 Schematic diagram of the main components of a system for optimizing a wireless network based on a tree convolutional network model in an embodiment of the present invention;
[0048] Figure 3 This is a schematic diagram of the physical structure of the electronic device provided by the present invention. DETAILED DESCRIPTION
[0049] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work should fall within the scope of protection of the present invention.
[0050] It should be noted that the terms "first," "second," and the like in the description of the present invention and the accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a particular order or precedence. It should be understood that the terms used in this manner are interchangeable where appropriate for the embodiments of the present invention described herein. In addition, the terms "including," "having," and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or apparatus comprising a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units that are not explicitly listed or that are inherent to these processes, methods, products, or apparatuses.
[0051] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features thereof can be combined with each other. The embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
[0052] Figure 1 This is a flow chart of a method for optimizing a wireless network based on a tree convolutional network model in an embodiment of the present invention. Figure 1 As shown, the method for optimizing a wireless network based on a tree convolutional network model provided by an embodiment of the present invention includes the following steps S101 to S104.
[0053] S101, collecting historical wireless network data.
[0054] The wireless network data used in the model comes from the ICASSP Communication Network Intelligent Operation and Maintenance Competition. The purpose of this data analysis is to identify the root cause of low feature 1 values. Data has been anonymized for privacy reasons. Each feature represents a key performance indicator (KPI). The training data contains 2,984 samples from different 5G road test segments. Nodes record data using time slice analysis, and a single time slice may contain one or more root causes. The feature data in a node is diverse and heterogeneous, and may be continuous or discrete. The data collection granularity also varies, with some features being recorded every few seconds and others every second. Furthermore, the data for each node is not fixed in length; a single time slice may contain dozens to hundreds of data items. Causal relationships are displayed between feature nodes through connections.
[0055] Since some of the original data have problems such as missing data and small data volume, data processing is required. First, the null values and outliers in the data set are replaced with 0. In the case of missing values, the average value of the time series data of the feature is used to fill in the missing values. In order to solve the problem of too little time series data in the data, kernel density estimation is used for data enhancement. Kernel density estimation is a non-parametric statistical method that uses multiple kernel functions to smoothly estimate the probability density distribution and can be used for a variety of different problems. The kernel density estimation method is used for data enhancement in the tree convolutional network model. Compared with other methods, kernel density estimation can be applied to complex data distributions and the data it generates is smoother than other methods, which can avoid the generation of samples with excessive noise. The kernel density estimation method can be used to expand all data to the same length, making it easier for the model to perform convolution.
[0056] S102, construct a tree convolution network model, which includes a global feature extraction module, a temporal convolution prediction module and a tree convolution diagnosis module. The global feature extraction module is used to extract global features. The temporal convolution prediction module is used to perform traffic prediction of wireless network data based on global features to obtain traffic prediction results. The tree convolution diagnosis module is used to perform fault diagnosis of the wireless network based on global features to obtain fault diagnosis results.
[0057] The global feature extraction module includes a convolutional layer, a spatiotemporal convolutional layer and a fully connected layer;
[0058] The global feature extraction module further extracts global features from local features through convolutional layers and temporal convolutional layers to capture overall trends and patterns in the data. The convolutional layer is used to extract local features and capture spatial patterns or time series patterns in the data. The temporal convolutional layer is implemented based on the temporal convolutional network (TCN) and efficiently captures long-term dependencies in time series through causal convolution, dilated convolution, and residual connections. It is particularly suitable for analyzing dynamic change trends in wireless network data, such as traffic prediction and fault diagnosis. The fully connected layer is used to integrate the local features and the spatiotemporal features to generate global features.
[0059] The temporal convolution module is responsible for predicting traffic. The extracted global features first enter two temporal convolution layers (the first spatiotemporal convolution layer and the second spatiotemporal convolution layer). Causal convolution helps ensure the causal relationship of the time series, that is, the output of the current time step only depends on the input of the current and previous time steps, avoiding future information leakage; dilated convolution increases the receptive field of the convolution kernel and captures long-term dependencies. For example, a convolution kernel with a dilation factor of 2 will skip the intermediate time steps and directly capture information from farther time steps, thereby effectively processing long sequence data. Finally, the shallow features are passed directly to the deep layer through the residual connection, alleviating the gradient disappearance problem and improving the training stability of the model. Compared with recurrent neural networks (RNNs), the convolution operations of TCN can be calculated in parallel, significantly improving the efficiency of training and inference. The information is then passed to the fully connected layer,
[0060] The method for optimizing a wireless network based on a tree convolutional network model further includes:
[0061] Based on the topological structure and node connection relationship of the wireless network, a graph structure represented by the adjacency matrix is constructed to obtain the connection status between each node in the wireless network.
[0062] The tree convolution diagnosis module includes a first tree convolution layer, a spatiotemporal convolution layer, a second tree convolution layer and a fully connected layer;
[0063] The first tree convolution layer is used to convert the adjacency matrix into a tree matrix, and perform preliminary feature extraction on the tree matrix through a convolution operation to obtain preliminary features;
[0064] The first tree convolution layer is used to convert the adjacency matrix into a tree matrix, including:
[0065] Perform a breadth-first traversal on the graph structure represented by the adjacency matrix to obtain a tree structure. When the resulting tree structure contains a loop, use a spatial tree matrix to represent the original tree structure. The tree structure is divided into different subtrees. The maximum number of layers, nodes, and child nodes corresponding to each subtree is calculated. Starting from the specified node of each subtree, all possible paths are found.
[0066] Convert all possible paths into a tree matrix.
[0067] The spatiotemporal convolution layer is used to capture the spatiotemporal features of the global features;
[0068] The second tree convolution layer is used to perform feature extraction on the tree matrix based on the spatiotemporal features to obtain in-depth features;
[0069] The fully connected layer is used to integrate the preliminary features, spatiotemporal features and the in-depth features to generate a fault diagnosis result of the wireless network.
[0070] The construction process of the tree matrix is:
[0071] Based on the existing adjacency matrix, a breadth-first traversal of the graph is performed. Due to the properties of breadth-first traversal, a tree structure can be directly obtained. The result is divided into different subtrees and the maximum number of levels, nodes, and child nodes is calculated. All possible paths starting from the specified node are found and the results are saved to prepare for the tree matrix construction.
[0072] When converting a path to a matrix, the root node must be expanded first. When a graph is converted to a tree, it's common for a node to have multiple children. When representing this in a matrix, each child node should have a parent node represented in the row above it. Nodes of varying lengths should be represented by symbols to ensure a consistent matrix length.
[0073] In some datasets, the graph converted from the adjacency matrix contains cycles. This means that the result of a breadth-first traversal also contains cycles. Therefore, a one-dimensional tree representation of the original graph presents problems. Therefore, we use a spatial tree matrix to represent the original tree. The spatial tree matrix is composed of multiple one-dimensional tree matrices. Setting y as the tree matrix, each node in the original tree can be used as the root node, and the tree representation is achieved by increasing the dimensionality. Relevance is added as weight. At this point, only the node index is obtained. By comparing it with the relevance matrix, the node value can be converted to the magnitude of the relevance. If the weight is 0, the node's relevance value is 1. If the weight is not 0, the node's value is 1 / weight. This process results in a tree matrix. This not only represents more spatial relevance than the adjacency matrix, but also adheres to the requirements of graph regularization in its design. It can handle graph structures of varying complexity and improves stability.
[0074] Then, spatial features are extracted from the data through spatial convolution, which includes path convolution and hierarchical convolution. The process of tree path convolution network is similar to that of graph convolution, both of which perform convolution on the generated graph structure.
[0075] The difference is that due to the hierarchical nature of tree matrices, the tree path convolutional network process includes path convolution and hierarchical convolution. The spatial tree matrix is composed of multiple one-dimensional tree matrices. During the convolution process, each tree matrix is convolved separately and the results are fused.
[0076] The Tree Path Convolutional Network takes a one-dimensional tree matrix as input. Since the generated tree matrix has converted the original tree into a matrix form, each row in the matrix represents a level in the tree. The convolution starts from the last row of the matrix, which is the leaf node in the tree, and moves up layer by layer.
[0077] Path convolution ensures more accurate information is obtained during convolution. A column in the tree matrix can be viewed as a local path, and the convolution process aggregates the local features of each node from the bottom up along this local path. This process allows the model to gradually capture the implicit relationships between nodes, building a high-level representation of the nodes in this path, and reducing network complexity.
[0078] By looping over all columns, we can obtain high-level features for all local paths, which, when concatenated, form the path features of a one-dimensional tree matrix. Furthermore, the convolution process for each local path is performed independently, without interfering with each other, making the features obtained by convolution more accurate.
[0079] Compared to convolving the entire local path at once, this convolution method can better preserve information about deeper nodes. In global convolution, information from all nodes is processed simultaneously, potentially ignoring some of the information about deeper nodes. However, this path convolution is performed step by step, which better preserves information about these nodes, avoiding information loss to a certain extent and allowing for in-depth learning of the characteristics of each path.
[0080] The process of hierarchical convolution differs from path convolution. Path convolution focuses on the relationships between nodes along a path, while hierarchical convolution focuses on the hierarchical relationships of nodes globally. During the tree matrix generation process, the root node is replicated multiple times to ensure that each node has a path to the root node. This allows for multiple accesses to the root node during convolution. Similarly, nodes connected to the root node are also convolved multiple times.
[0081] The convolution kernel performs horizontal translation calculations based on the row-first principle, which allows the computation to capture information about both the node and the level it resides in. Because the root node and its connected nodes are convolved multiple times, they retain more information, minimizing information loss during convolution. This method generates high-level features that not only include features at each level but also include hidden information about the nodes.
[0082] Nodes in wireless networks have varying influences on the root node depending on their level. Using hierarchical convolution can more comprehensively capture the relationship between nodes and the global system, helping to understand the position and characteristics of nodes at each level. The extracted spatial features are input into the temporal convolution layer to extract temporal features. These temporal features are then passed to the tree convolution layer to further extract spatial features. This process enhances the model's understanding of complex spatial relationships in the data. Finally, the processed feature data is fed into the fully connected layer for comprehensive analysis and processing, enabling fault diagnosis. This method effectively integrates temporal and spatial information through staged feature extraction, improving the accuracy and reliability of fault diagnosis.
[0083] S103: Input the historical wireless network data into the tree convolutional network model for training to obtain the optimal tree convolutional network model.
[0084] The four metrics of accuracy, precision, recall, and F1 score are commonly used to evaluate the fault diagnosis performance of a model. Accuracy is the most commonly used evaluation metric and the most intuitive indicator of the quality of a classification model. Accuracy is the ratio of correct samples to the total number of samples, which provides a simple and intuitive overview of the model's overall performance. Precision is the ratio of the number of positive examples correctly classified by the model to the total number of positive examples. This metric measures the proportion of faults detected by the model that are actually faults, allowing us to assess the accuracy of the model's positive predictions and helping improve the model. Recall is the ratio of correct samples to actual correct samples. In wireless networks, this measures the model's fault detection capability. Recall demonstrates the model's ability to identify actual positive examples. The F1 score, which considers both precision and recall, comprehensively reflects the model's capabilities and is a more comprehensive metric.
[0085] In the present invention, in order to evaluate the performance of the prediction model in the traffic prediction task, the mean squared error (MSE) is used as a performance evaluation indicator. The prediction accuracy is quantified by calculating the square average of the difference between the predicted traffic value and the actual traffic observation value. It not only takes into account the deviation between all predicted data points and the corresponding true value, but also enhances the sensitivity to larger errors by squaring the error terms. Therefore, a lower MSE value indicates that the model has higher prediction accuracy, while a higher MSE value indicates that there is a larger prediction deviation, and it may be necessary to further optimize the model structure or adjust the parameters to improve the prediction effect, thereby providing a scientific basis for the iterative improvement of the model, and being able to systematically compare the prediction capabilities of different models or different versions of the same model to ensure that the optimal solution is selected for traffic prediction in actual application environments.
[0086] S104, acquiring real-time wireless network data, inputting the real-time wireless network data into an optimal tree convolutional network model to obtain traffic prediction results and fault diagnosis results of the real-time wireless network data, and optimizing the wireless network based on the traffic prediction results and fault diagnosis results.
[0087] Figure 2 Schematic diagram of the main components of the system for optimizing wireless networks based on the tree convolutional network model in an embodiment of the present invention. Figure 2 As shown, the system 1 for optimizing a wireless network based on a tree convolutional network model provided by an embodiment of the present invention includes a collection module 10, a construction module 20, a training module 30 and an optimization module 40.
[0088] The collection module 10 is used to collect historical wireless network data;
[0089] A construction module 20 is configured to construct a tree convolutional network model, wherein the tree convolutional network model includes a global feature extraction module, a temporal convolution prediction module, and a tree convolution diagnosis module. The global feature extraction module is configured to extract global features. The temporal convolution prediction module is configured to perform traffic prediction of wireless network data based on the global features to obtain traffic prediction results. The tree convolution diagnosis module is configured to perform fault diagnosis of the wireless network based on the global features to obtain fault diagnosis results.
[0090] A training module 30 is configured to input the historical wireless network data into the tree convolutional network model for training to obtain an optimal tree convolutional network model;
[0091] The optimization module 40 is used to obtain real-time wireless network data, input the real-time wireless network data into the optimal tree convolutional network model to obtain traffic prediction results and fault diagnosis results of the real-time wireless network data, and optimize the wireless network based on the traffic prediction results and fault diagnosis results.
[0092] Figure 3 A schematic diagram of the physical structure of an electronic device provided in an embodiment of the present invention, such as Figure 3 As shown, the electronic device 50 includes: a processor 501 (processor), a memory 502 (memory) and a bus 503;
[0093] The processor 501 and the memory 502 communicate with each other via the bus 503.
[0094] The processor 501 is used to call the program instructions in the memory 502 to execute the methods provided by the above-mentioned method embodiments, so as to execute the methods provided by the implementation methods of the present invention.
[0095] This embodiment provides a non-transitory computer-readable storage medium, which stores computer instructions. The computer instructions enable a computer to execute the method provided by the embodiment of the present invention.
[0096] Those skilled in the art will understand that all or part of the steps of implementing the above-mentioned method embodiment can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above-mentioned method embodiment; and the aforementioned storage medium includes: ROM, RAM, disk or optical disk, etc. Various storage media that can store program codes.
[0097] The above specific embodiments do not limit the scope of protection of the present invention. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may occur depending on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention are intended to be included within the scope of protection of the present invention.
Claims
1. A method for optimizing a wireless network based on a tree convolutional network model, characterized in that: include: Collect historical wireless network data; Constructing a tree convolutional network model, the tree convolutional network model includes a global feature extraction module, a temporal convolution prediction module, and a tree convolution diagnosis module, wherein the global feature extraction module is used to extract global features, the temporal convolution prediction module is used to perform traffic prediction of wireless network data based on the global features to obtain a traffic prediction result, and the tree convolution diagnosis module is used to perform fault diagnosis of the wireless network based on the global features to obtain a fault diagnosis result; Inputting the historical wireless network data into the tree convolutional network model for training to obtain an optimal tree convolutional network model; Real-time wireless network data is acquired, and the real-time wireless network data is input into the optimal tree convolutional network model to obtain traffic prediction results and fault diagnosis results of the real-time wireless network data, and the wireless network is optimized based on the traffic prediction results and fault diagnosis results.
2. The method for optimizing a wireless network based on a tree convolutional network model according to claim 1, wherein: The global feature extraction module includes a convolutional layer, a spatiotemporal convolutional layer and a fully connected layer; The convolutional layer is used to extract local features of wireless network data; The spatiotemporal convolution layer is used to extract spatiotemporal features of wireless network data; The fully connected layer is used to integrate the local features and the spatiotemporal features to generate global features.
3. The method for optimizing a wireless network based on a tree convolutional network model according to claim 2, wherein: The temporal convolution prediction module includes a first spatiotemporal convolution layer, a second spatiotemporal convolution layer and a fully connected layer; The first spatiotemporal convolutional layer is used to capture the spatiotemporal features related to traffic prediction in the global features; The second spatiotemporal convolutional layer is used to capture key spatiotemporal features related to traffic prediction; The fully connected layer is used to integrate the spatiotemporal features and the key spatiotemporal features to generate a traffic prediction result of the wireless network data.
4. The method for optimizing a wireless network based on a tree convolutional network model according to claim 2, wherein: The method for optimizing a wireless network based on a tree convolutional network model further includes: Based on the topological structure and node connection relationship of the wireless network, a graph structure represented by the adjacency matrix is constructed to obtain the connection status between each node in the wireless network.
5. The method for optimizing a wireless network based on a tree convolutional network model according to claim 4, wherein: The tree convolution diagnosis module includes a first tree convolution layer, a spatiotemporal convolution layer, a second tree convolution layer and a fully connected layer; The first tree convolution layer is used to convert the adjacency matrix into a tree matrix, and perform preliminary feature extraction on the tree matrix through a convolution operation to obtain preliminary features; The spatiotemporal convolution layer is used to capture the spatiotemporal features of the global features; The second tree convolution layer is used to perform feature extraction on the tree matrix based on the spatiotemporal features to obtain in-depth features; The fully connected layer is used to integrate the preliminary features, spatiotemporal features and the in-depth features to generate a fault diagnosis result of the wireless network.
6. The method for optimizing a wireless network based on a tree convolutional network model according to claim 1, wherein: The first tree convolution layer is used to convert the adjacency matrix into a tree matrix, including: Perform a breadth-first traversal on the graph structure represented by the adjacency matrix to obtain a tree structure. When the resulting tree structure contains a loop, use a spatial tree matrix to represent the original tree structure. The tree structure is divided into different subtrees. The maximum number of layers, nodes, and child nodes corresponding to each subtree is calculated. Starting from the specified node of each subtree, all possible paths are found. Convert all possible paths into a tree matrix.
7. The method for optimizing a wireless network based on a tree convolutional network model according to claim 1, wherein: Before the step of inputting the historical wireless network data into the tree convolutional network model for training, the method further includes: Identifying null values and abnormal values in the historical wireless network data, and replacing the null values and abnormal values with 0; Identify features of missing values in the historical wireless network data, and fill in the missing values based on average values of time series features corresponding to the features to ensure the integrity and availability of the historical wireless network data; Kernel density estimation is used to perform data enhancement on the historical wireless network data.
8. A system for optimizing wireless networks based on a tree convolutional network model, characterized in that: include: Collection module, used to collect historical wireless network data; A construction module is used to construct a tree convolutional network model, wherein the tree convolutional network model includes a global feature extraction module, a temporal convolution prediction module, and a tree convolution diagnosis module. The global feature extraction module is used to extract global features. The temporal convolution prediction module is used to perform traffic prediction of wireless network data based on the global features to obtain traffic prediction results. The tree convolution diagnosis module is used to perform fault diagnosis of the wireless network based on the global features to obtain fault diagnosis results. A training module, configured to input the historical wireless network data into the tree convolutional network model for training to obtain an optimal tree convolutional network model; The optimization module is used to obtain real-time wireless network data, input the real-time wireless network data into the optimal tree convolutional network model to obtain traffic prediction results and fault diagnosis results of the real-time wireless network data, and optimize the wireless network based on the traffic prediction results and fault diagnosis results.
9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.
10. A non-transitory computer readable medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.