A fault diagnosis method and system of a wireless network

Abstract

The application discloses a kind of fault diagnosis method and system of wireless network, method includes: obtaining the graph data of current wireless network, graph data includes the correlation matrix between nodes and adjacency matrix;According to correlation matrix and adjacency matrix, graph data is converted into tree matrix;Spatial convolution operation is carried out to tree matrix, to obtain its spatial feature matrix;Weight matrix is calculated based on the spatial feature matrix of tree matrix;Time convolution operation is carried out to weight matrix, to obtain time sequence feature matrix;Full connection processing and activation operation are carried out to time sequence feature matrix to obtain the classification result of final node fault.Solved the local path problem existing in graph convolution, apply spatial convolution to wireless network fault detection, can be convolved to different path and level, reduce the possibility of missing information, enhance the ability of spatial correlation acquisition, conducive to more quickly and accurately determine the fault in wireless network.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of network fault diagnosis, and in particular to a wireless network fault diagnosis method and system. BACKGROUND

[0002] Wireless sensor network is composed of a series of interconnected sensor nodes, using distributed technology, wireless communication and computer network and other technologies, which can realize data monitoring and other functions. At present, WSN plays an important role in many fields, from basic data monitoring to dangerous warning and traffic flow analysis in smart city, such as real-time vehicle flow monitoring and pedestrian trajectory recording, which fully shows its wide application prospect. In order to protect the stability of wireless network and user experience, fault diagnosis is a key link. With the wide application of wireless network, its complexity is increasing. The current fault diagnosis methods cover data analysis, machine learning, artificial intelligence, distributed computing and other technologies. Among them, machine learning and artificial intelligence methods are particularly representative.

[0003] In the traditional various fault diagnosis methods, there are certain defects, for example, in the method based on time series analysis, this method focuses on low-dimensional data and cannot be applied to high-dimensional data. The method of time series analysis is more detailed for fault division and has high detection efficiency, but cannot capture the dynamic change characteristics in the process of wireless network information transmission. For example, in the method based on graph data analysis, when processing graph data, its adjacency matrix can only reflect the relationship or no relationship. When the relationship between nodes appears strong and weak, the adjacency matrix cannot accurately represent the relationship between nodes. This will affect the acquisition of features in the data and affect the accuracy of fault diagnosis.

[0004] Therefore, how to overcome the shortcomings of the above-mentioned prior art has high research value. SUMMARY

[0005] Therefore, the present application provides a wireless network fault diagnosis method and system, which can convert the original graph structure into a hierarchical tree structure by constructing a tree matrix, so as to improve the efficiency and accuracy of fault diagnosis between nodes. The present application provides the following technical solutions:

[0006] A wireless network fault diagnosis method, comprising:

[0007] Obtaining graph data of the current wireless network, the graph data including a correlation matrix and an adjacency matrix between nodes;

[0008] Converting the graph data into a tree matrix according to the correlation matrix and the adjacency matrix;

[0009] performing a spatial convolution operation on the tree matrix to obtain a spatial feature matrix of the tree matrix;

[0010] calculating a weight matrix based on the spatial feature matrix of the tree matrix;

[0011] performing a time convolution operation on the weight matrix to obtain a time sequence feature matrix;

[0012] performing a full connection and activation operation on the time sequence feature matrix to obtain a final classification result of the node fault.

[0013] Further, the method of converting the graph data into the tree matrix according to the correlation matrix and the adjacency matrix comprises:

[0014] performing a breadth-first traversal operation on the graph data according to the adjacency matrix to obtain a plurality of tree structures;

[0015] respectively obtaining all path data starting from any specified node in the tree structure, and performing root node expansion according to the path data to ensure the uniformity of the lengths of the plurality of tree structures in the matrix representation;

[0016] respectively taking each node of the tree structure as a root node to generate a corresponding one-dimensional tree matrix, so as to construct a spatial tree matrix composed of a plurality of one-dimensional tree matrices;

[0017] converting the node values of the spatial tree matrix into the sizes of the correlations between the nodes according to the correlation matrix, so as to complete the conversion of the tree matrix.

[0018] Further, the method of performing a spatial convolution operation on the tree matrix to obtain a feature matrix of the tree matrix comprises:

[0019] the spatial convolution comprises path convolution and hierarchical convolution;

[0020] the method of the path convolution comprises:

[0021] each column of the tree matrix is taken as a local path in the path tree convolution;

[0022] performing a convolution operation from the bottom end of each local path of each one-dimensional tree matrix upward, and obtaining a feature matrix in which the local features of each node are aggregated;

[0023] performing a cyclic convolution operation on the feature matrix by column to obtain an advanced feature matrix of all local paths, and connecting a plurality of advanced feature matrices to obtain a path feature matrix of each one-dimensional tree matrix;

[0024] the method of the hierarchical convolution comprises:

[0025] performing a convolution operation laterally from the leaf node of each one-dimensional tree matrix to obtain a feature matrix including the node and the hierarchical information;

[0026] performing a cyclic convolution operation on the feature matrix to obtain a high-level feature matrix of all local paths, connecting a plurality of high-level feature matrices to obtain a hierarchical feature matrix of each one-dimensional tree matrix;

[0027] fusing the path feature matrix and the hierarchical feature matrix to obtain a spatial feature matrix of the tree matrix.

[0028] Further, the method for calculating the weight matrix based on the spatial feature matrix of the tree matrix comprises:

[0029] obtaining timing data of the current wireless network;

[0030] multiplying the spatial feature matrix and the timing data to obtain the weight matrix.

[0031] Further, the method for performing a time convolution operation on the weight matrix to obtain a timing feature matrix comprises:

[0032] performing a Gaussian convolution operation on the weight matrix to obtain time feature data of high latitude in the timing data, thereby obtaining the timing feature matrix.

[0033] The application further discloses a fault diagnosis system of a wireless network, comprising:

[0034] a data acquisition module configured to acquire graph data of the current wireless network, the graph data comprising a correlation matrix and an adjacency matrix between nodes;

[0035] a conversion module configured to convert the graph data into a tree matrix according to the correlation matrix and the adjacency matrix;

[0036] a convolution operation module configured to perform a spatial convolution operation on the tree matrix to obtain a spatial feature matrix thereof, and further configured to,

[0037] calculate a weight matrix based on the spatial feature matrix of the tree matrix;

[0038] perform a time convolution operation on the weight matrix to obtain a timing feature matrix;

[0039] a fault diagnosis module configured to perform a full connection processing and an activation operation on the timing feature matrix to obtain a final classification result of node faults.

[0040] Further, the conversion module is further configured to,

[0041] perform a breadth-first traversal operation on the graph data according to the adjacency matrix to obtain a plurality of tree structures;

[0042] obtain all path data starting from any specified node in the tree structure respectively, and perform root node expansion according to the path data to ensure the uniformity of the lengths of the plurality of tree structures in matrix representation.

[0043] Respectively taking each node of the tree structure as a root node, a corresponding one-dimensional tree matrix is generated to construct a space tree matrix composed of multiple one-dimensional tree matrices;

[0044] According to the correlation matrix, the node value of the space tree matrix is converted into the size of the correlation between nodes to complete the conversion of the tree matrix.

[0045] The application further discloses a computer readable storage medium, the storage medium stores a computer program, and the computer program is executed by a processor to realize the method.

[0046] The application further discloses an electronic device, including a memory, a processor and a computer program stored in the memory and executable on the processor, and the processor realizes the method when executing the program.

[0047] The application further discloses a computer program product, including a computer program, and the computer program realizes the method when being executed by a processor.

[0048] According to the technical scheme of the application, the tree matrix is used to replace the original adjacency matrix, and the graph structure is represented in the form of a tree. The tree structure can more clearly represent all nodes that have an influence on the root node, not only considering the neighbor nodes, and solves the local path problem existing in the graph convolution. Meanwhile, the spatial convolution is applied to the wireless network fault detection, the different paths and levels in the tree matrix can be convolved, the important information is reserved while the different level features are aggregated, the possibility of losing information is reduced, and the hierarchy of the tree matrix facilitates us to set the correlation size of the node, enhances the spatial correlation acquisition ability, and is beneficial to more quickly and accurately determining the fault in the wireless network. BRIEF DESCRIPTION OF DRAWINGS

[0049] For the purpose of illustration and not limitation, the application will now be described in conjunction with embodiments thereof and the accompanying drawings, in which:

[0050] Figure 1 is a flowchart of a fault diagnosis method of a wireless network in an embodiment of the application;

[0051] Figure 2 is a schematic diagram of an implementation architecture of a fault diagnosis method of a wireless network in an embodiment of the application;

[0052] Figure 3 is a schematic diagram of a structure of a fault diagnosis system of a wireless network in an embodiment of the application.

[0053] Figure 4 is a schematic diagram of an electronic device structure in an embodiment of the application.

[0054] Figure 5 is a schematic diagram of path convolution in an embodiment of the present application;

[0055] Figure 6 is a schematic diagram of hierarchical convolution in an embodiment of the present application;

[0056] Figure 7 is a schematic diagram of Gaussian convolution in an embodiment of the present application;

[0057] Figure 8 is a comparison of model accuracy at different numbers of nodes in a comparative experiment in an embodiment of the present application;

[0058] Figure 9 is a comparison of classification accuracy of different comparative models in an embodiment of the present application;

[0059] Figure 10 is a comparison of confusion matrix results of different comparative models in an embodiment of the present application. DETAILED DESCRIPTION

[0060] In order to make the personnel in the art better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work should belong to the scope of protection of the present application.

[0061] It should be noted that the terms "first", "second", and the like in the specification of the present application and the above drawings are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device that includes a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0062] In the present application, the terms "upper", "lower", "left", "right", "front", "back", "top", "bottom", "inner", "outer", "middle", "vertical", "horizontal", "lateral", "longitudinal" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings. These terms are mainly used to better describe the present application and its embodiments, and are not intended to limit the indicated devices, elements or components to have a particular orientation, or to be constructed and operated in a particular orientation.

[0063] In addition, in addition to being used to indicate the orientation or positional relationship, the above-mentioned partial terms can also be used to indicate other meanings, for example, the term "upper" can also be used to indicate a certain dependent relationship or connection relationship in some cases. For those skilled in the art, the specific meaning of these terms in the present application can be understood according to the specific circumstances. In addition, the meaning of the term "a plurality of" should be two and more than two.

[0064] It should be noted that the features in the embodiments and the embodiments of the present application can be combined with each other without conflict. The embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0065] Reference Figure 1 and Figure 2 A fault diagnosis method of a wireless network, comprising:

[0066] S100: Obtain graph data of the current wireless network, the graph data comprising a correlation matrix and an adjacency matrix between nodes.

[0067] The adjacency matrix reflects the direct connection relationship between the nodes on the way, and the correlation matrix represents the relationship strength or similarity between the nodes, reflecting the correlation between the nodes. Since the adjacency matrix can only reflect the relationship or no relationship between the nodes, when the relationship between the nodes appears strong or weak, the relationship between the nodes cannot be accurately represented by the adjacency matrix alone, and the accuracy of fault diagnosis will be greatly affected when the data features are obtained on this basis, therefore, a structure is needed to be constructed to hierarchically represent the original graph structure hierarchy without the spatial correlation between the nodes.

[0068] S200: Convert the graph data into a tree matrix according to the correlation matrix and the adjacency matrix.

[0069] The tree matrix can convert the originally complex graph structure into a hierarchical tree structure to better capture the spatial correlation between the nodes. The tree matrix node contains the correlation size of the node and its parent node. Specifically, the present embodiment discloses a construction method of a tree matrix, which specifically comprises:

[0070] S210: Perform a breadth-first traversal operation on the graph data according to the adjacency matrix to obtain a plurality of tree structures.

[0071] In the graph structure converted from the adjacency matrix, there is a loop, and the result after the breadth-first traversal also has a loop, so the one-dimensional tree matrix has a problem in representing the original graph structure, and the spatial tree matrix is used to represent the original tree structure. Due to the characteristics of the breadth-first traversal, the result of the tree structure can be directly obtained. At the same time, the tree structure result is divided into different sub-trees, and the maximum number of layers, the number of nodes, and the number of sub-nodes are calculated.

[0072] S220: Obtain all path data starting from any specified node in the tree structure, and perform root node expansion according to the path data to ensure the uniformity of the lengths of the multiple tree structures in the matrix representation.

[0073] S230: Generate a one-dimensional tree matrix corresponding to each node of the tree structure as a root node, to construct a spatial tree matrix composed of multiple one-dimensional tree matrices.

[0074] S240: Convert the node values of the spatial tree matrix into the size of the correlation between nodes according to the correlation matrix, to complete the conversion of the tree matrix.

[0075] The fault diagnosis model constructed in the embodiment is a tree path convolutional network, and the model uses a combination of spatial convolution and temporal convolution to complete the fault diagnosis of the wireless network. Specifically:

[0076] S300: Perform a spatial convolution operation on the tree matrix to obtain a spatial feature matrix thereof.

[0077] The specific method includes:

[0078] The spatial convolution includes path convolution and hierarchical convolution.

[0079] Reference Figure 5 The method of path convolution includes:

[0080] S310: Each column of the tree matrix is regarded as a local path in the path tree convolution.

[0081] S320: Perform a convolution operation from the bottom of each local path of each one-dimensional tree matrix to the top, and obtain a feature matrix in which the local features of each node are aggregated.

[0082] In the embodiment, a 3x4 matrix is taken as an example, the first row is the root node of the tree matrix, each column is a local path, and the shaded part in the figure is the convolution kernel of the current convolution operation. The process of path convolution is performed only between the current node and its parent node each time, so that more accurate information can be obtained during convolution, and the implicit relationship between nodes can be captured, a high-level representation of the nodes in the current path is established, and the complexity of the network is reduced.

[0083] S330: Perform a cyclic convolution operation on the feature matrix by column to obtain the advanced feature matrix of all local paths, and connect multiple advanced feature matrices to obtain the path feature matrix of each one-dimensional tree matrix.

[0084] The convolution process of the local path is independent and does not interfere with each other, so that the features obtained by convolution are more accurate. Compared with the way of convolving the entire local path at a time, this convolution method can better preserve the information of deep nodes. When the whole convolution is performed, the information of all nodes is processed at the same time, and the information of deep nodes may be ignored. However, this path convolution is performed step by step, which can better preserve the information of these nodes. To some extent, it avoids information loss and deeply learns the features of each path.

[0085] Reference Figure 6 The hierarchical convolution method includes:

[0086] S340: Starting from the leaf node of each one-dimensional tree matrix, perform a convolution operation transversely to obtain a feature matrix including node and hierarchical information.

[0087] This embodiment takes the tree structure and tree matrix in Figure 6 as an example, where f represents the hierarchical features obtained by the convolution process.

[0088] The tree matrix will copy the root node multiple times during the generation process to ensure that each node has a path to the root node, so that the information of the root node can be obtained multiple times during convolution. Similarly, the nodes connected to the root node will also be convolved multiple times. The convolution kernel performs transverse translation calculation according to the row priority principle, so that the node and the hierarchical information can be obtained during the operation. Because the root node and the nodes connected to the root node are convolved multiple times, they retain more information and reduce the loss of information during convolution. The hierarchical features obtained in this way not only include the features of each level, but also include the hidden information between nodes. The nodes in the wireless network have different effects on the root node according to the level, and the use of hierarchical convolution can more comprehensively capture the relationship between the nodes and the global to obtain the position and features of the nodes in each level.

[0089] S350: Perform a cyclic convolution operation on the feature matrix to obtain the advanced feature matrix of all local paths, and connect multiple advanced feature matrices to obtain the hierarchical feature matrix of each one-dimensional tree matrix;

[0090] S360: Fuse the path feature matrix and the hierarchical feature matrix to obtain the spatial feature matrix of the tree matrix.

[0091] After the spatial convolution operation is completed and the spatial feature matrix is obtained, time convolution needs to be added to obtain the time features in the time series data. Specifically:

[0092] S400: calculating the weight matrix based on the spatial feature matrix of the tree matrix. Specifically, it includes:

[0093] Obtaining timing data of the current wireless network;

[0094] Multiplying the spatial feature matrix and the timing data to obtain the weight matrix.

[0095] In this embodiment, before the weight matrix is calculated, the graph data of the wireless network also needs to be processed. Since each root cause has specific feature data corresponding to it, all data cannot be put into the classifier at once, so the first step after obtaining the graph data is to extract specific features. For example, root cause 1 uses 39 features, root cause 2 and root cause 3 use 18 features, and these three root causes have a few features in common. Then, after dividing the features, data processing is performed according to the input format required by the classifier. For example, it is found through observation that the original data set has problems such as missing sample data and insufficient sample size, so the input data needs to be processed first. Exemplarily, the difference between the feature values of the training set data and the test set data is large, and the feature value range is distributed between 0-700, which is not conducive to model training. Therefore, z-score normalization is used to reduce the feature value range and reduce the data scale. After normalization, the original data is mapped to a standard normal distribution centered at 0, which can speed up the training of the model. Secondly, since the label categories in the data set are imbalanced, this will affect the classification effect, so over-sampling is performed using Borderline SMOTE. Borderline SMOTE can increase the amount of data by synthesizing samples, which can not only solve the problem of data imbalance, but also improve the classification effect of the model. Further, in traditional neural networks, as the number of network layers increases, the performance of the model may degenerate, that is, a deeper network may not perform as well as a shallow network. Therefore, in this embodiment, a residual network (Residual Network, ResNet) is introduced, which is a deep learning network structure that solves the problems of gradient disappearance and degradation in deep neural networks by introducing skip connections. The principle of skip connection is to directly pass the input to the later layer after bypassing several layers, so that the network can learn the residual. Through such design, it allows the gradient to propagate more effectively in the deep network, avoiding the problems of gradient disappearance and gradient explosion.

[0096] Reference Figure 6 , S500: performing time convolution operation on the weight matrix to obtain the timing feature matrix;

[0097] The time convolution process is the process of obtaining the timing feature matrix through Gaussian convolution operation.

[0098] The specific operation is:

[0099] Reference Figure 7 The weight matrix is subjected to a Gaussian convolution operation to obtain time feature data of high latitude in the time series data, thereby obtaining a time feature matrix.

[0100] The time series data in the wireless network is recorded at a certain time interval, and there is continuity between the data, and the occurrence of a fault is usually reflected in the fluctuation of a certain time series data. Therefore, the Gaussian convolution method can extract the high latitude time features in the data.

[0101] The probability density function formula of Gaussian distribution is represented as:

[0102]

[0103] An example is shown in which σ is the standard deviation, the value is 0.7, x is the variable, and μ is the mean.

[0104] After normalizing the Gaussian convolution, the formula of the Gaussian convolution can be obtained:

[0105] M i+1 =G n *f(M i );

[0106] wherein Mi is a spatial feature matrix obtained after spatial convolution, G n is a current convolution kernel of Gaussian convolution, and f(M i ) is an output of an activation function.

[0107] After the convolution is completed, normalization is performed, and the data passes through a ReLu layer and finally through a Dropout regularization. This process is repeated multiple times to capture the time features in the data and form a time series feature matrix.

[0108] S600: Fully connecting and activating the time series feature matrix to obtain a final classification result of the node fault.

[0109] The wireless network data used in the exemplary embodiment shown in this embodiment comes from the icassp communication network intelligent operation and maintenance competition. The purpose of the data research is to find out the root cause of the low value of feature0 in the wireless network data. Each feature represents a key performance indicator (KPI), and the training data contains 2984 samples from different 5G road test segments. In order to simulate the real situation, only 45% of the data is labeled with root cause, and the rest of the data has no label. The node uses time slice analysis to record data, and a time slice may contain one or more root causes. The data of the features in the node are diversified and heterogeneous data, which may be continuous or discrete values, and the data collection granularity is different. Most of the data is recorded once every second, but some of the data interval is up to tens of seconds. In order to simulate the real situation, all the data is not fixed length, and the number of data records varies from a few to several thousand.

[0110] The following is the experimental part:

[0111] In the experimental part, the accuracy, precision, recall and F1 score are used to evaluate the classification effect of the fault diagnosis model in this embodiment. Accuracy is the most commonly used evaluation value index, and is also the most intuitive index to reflect the good or bad of a classification model. The accuracy rate is the proportion of all correct samples to the total samples, which can reflect the overall performance of the model, which is simple and intuitive. The precision is the proportion of the number of correct classification of positive examples to the total number of classification of positive examples. This index can detect the proportion of real faults in the faults detected by the model, and can view the accuracy of the model predicting positive examples, which helps to improve the model. Recall is the proportion of correct samples to actual correct samples, which is to view its fault detection ability in wireless network. Through the recall rate, the recognition ability of the model to the actual positive sample can be seen. F1 score considers both accuracy and recall, and can comprehensively reflect the model ability, which is a more comprehensive index.

[0112] This experiment will be compared with some common machine learning and deep learning algorithms, and the specific comparison model and parameter setting are as follows:

[0113] ACO-GCN: adopts the method of combining ant colony algorithm and GCN and GRU for diagnosis. The initial step is 1, and the compensation decay factor is 0.96.

[0114] GAT+LSTM: For data, first use LSTM to extract time information, and then use GAT to extract spatial information. The dropout rate is 0.2, and the attention unit is 8.

[0115] GCN: Graph convolutional network, which can consider the information of neighbor nodes in the process of convolution. Three layers of GCN are used for wireless network fault diagnosis, and self-loop is added to consider self information. Weight decay is added to prevent overfitting, and the parameter is set to 5e-4

[0116] GraphSAGE: An improved method of graph convolution, which uses random dropout to avoid the problem of long time consumption caused by too many neighbor nodes. Three layers of convolution are used to help learn complex nodes. The number of neighbor samples is set to 10.

[0117] ERT: Similar to random forest, it uses random feature subsets for segmentation and then votes to decide the classification. The number of trees is set to 30, the quality criterion uses Gini coefficient, and the minimum number of samples required for splitting nodes is 2

[0118] Quantitative experiment:

[0119] In this part, tree path convolutional network and the set contrast model are compared. Before the experiment, the tree matrix needs to be set. In order to better compare, all experiments will include ordinary tree path convolutional network and tree path convolutional network with residual network. In the tree path convolutional network model, the size of the tree matrix needs to be defined according to the task requirements. If not, it will affect the processing efficiency of the model. Figure 8 In the figure, the detection accuracy of tree path convolutional network under different node numbers is shown. It can be seen that in the case of small number of nodes, the model cannot get a high accuracy, at this time, the features obtained are too few, which cannot accurately judge the fault. When the number of nodes is more than 50, the accuracy shows a downward trend. The probability of this situation will increase when the number of nodes increases, which will limit the transmission of information and cause the accuracy to decrease. Figure 8 As can be seen in the figure, when the number of nodes is 50, the accuracy reaches the highest 91.83%, so we choose a tree matrix with 50 nodes.

[0120] Figure 9The comparison results of the tree path convolution network and other comparative models in accuracy, precision, and other indicators. In terms of accuracy, path tree convolution is the highest among these models, and the precision and recall of the tree path convolution network have reached a very high value, which shows that compared with other graph-related models, it has fewer misjudgments and can better distinguish positive examples. The GAT+LSTM method uses LSTM to obtain the time features of the nodes, and GAT can set different weights for all neighbor nodes to distinguish the influence size of different nodes. Therefore, the GAT+LSTM method is the highest among the graph-related methods. But in the accuracy, it is 1.83% lower than the path tree convolution. ACO-GCN has improved the accuracy compared with GCN, but since it still uses the graph convolution method in spatial features, it has not exceeded the GAT+LSTM method. SVM is usually used to process structured data, and the precision and recall are not high for this wireless network dataset. Although ERT can capture nonlinear relationships and adapt to a variety of different types of data, it cannot analyze graph-structured data, so the overall F1 score is low. In general, the F1 score of the tree path convolution network is the highest, which is 1.86% and 2.03% higher than the GAT-LSTM and ACO-GCN methods respectively, which are both space-time graph convolution methods. It proves that its comprehensive ability is superior to these comparative models.

[0121] Qualitative experiments and analysis:

[0122] As Figure 10 shown, this is the confusion matrix comparison of the model and other comparative models. Rootcase1 uses R1 to represent, and so on. The confusion matrix can reflect the comparison between the predicted results and the true results, and the deeper the color on the diagonal line represents the closer the predicted value and the true value. Moreover, the confusion matrix can more intuitively show the classification situation of specific root causes, which helps to improve the efficiency of root cause analysis. Figure 7The classification results of the middle path tree convolution are better than other comparative models in general. The case of misclassification occurs more frequently in R2, because there are only four data in the training data for the occurrence of R2 alone, and the rest are the occurrence of R2 and R3 together. This will make the model not enough data to distinguish the case of R2 alone, resulting in misclassification of the results. GAT+LSTM is roughly the same as path tree convolution in single root cause cases, but misclassification increases in multiple root cause cases. In the process of spatial convolution, the GAT+LSTM method cannot consider the influence of more distant nodes, which leads to its overall performance being worse than path tree convolution. Especially when R1 and R3 occur at the same time, the shared features cause the classification effect of R3, an edge node, to weaken, so the classification result on the confusion matrix is 6% worse. Compared with GCN, ACO-GCN adds a time convolution part, so it is better than GCN in overall classification effect. GAT+LSTM can set the correlation size, while ACO-GCN cannot, which leads to its root cause classification accuracy being worse than GAT+LSTM. The classification effect of GCN is excellent in single root cause performance, but it cannot accurately distinguish the root cause in complex multiple root cause cases. This is because the basic graph convolution cannot consider the correlation and time characteristics of the nodes, and the diagnostic ability in complex topological environment is limited. GraphSAGE randomly discards nodes during convolution, further exacerbating the problem of insufficient training data, resulting in inaccurate classification results in multiple root cause cases. That is, random trees cannot utilize the graph structure in wireless networks, so their overall performance in the confusion matrix is lower than graph-related methods.

[0123] Ablation experiment:

[0124] Path tree convolution is a space-time hybrid model that extracts spatial and temporal features from wireless network data for convolution. To further study the influence of temporal and spatial features in the model on the diagnostic ability of the model, we remove the temporal feature extraction part and the spatial feature extraction part of the model respectively and compare the experimental results. As shown in the following table:

[0125] Table 1 Comparison of ablation experiments

[0126]

[0127] From the data in Table 1, when only the spatial feature extraction part is retained, the overall fault diagnosis capability of the model decreases slightly, because in the data of the wireless network, there is a transmission effect of faults between nodes. The path convolution and hierarchical convolution contained in the spatial feature extraction part of the model can capture these faults. Since the faults affect layer by layer, the fault transmission of the leaf node affects the root node in the tree structure, which is the same as the way of extracting features by path convolution, so there is no great influence on the overall performance. However, when only the time feature extraction part is retained, the precision and recall rate and other indicators decrease greatly, because the model only convolves the node time series data and ignores the spatial influence, which increases the probability of incorrectly judging normal nodes as fault nodes, affecting the fault diagnosis capability of the model. Experiments show that for wireless network data with complex topology, using the method of time correlation and space correlation can obtain better diagnosis effect.

[0128] In summary, when using graph convolution for wireless network fault diagnosis, it is found that graph convolution has the limitations of local path and hierarchy. It can only capture the features of the neighbor nodes of a node and cannot consider the influence of more distant nodes on the node. The path between nodes is too complex and has no hierarchy, and the size of the node correlation cannot be studied. To address these limitations, the present embodiment proposes a method of using tree path convolution network for wireless network fault diagnosis. The tree matrix is used to change the original graph structure into a hierarchical tree structure, and the tree path convolution network can also ensure that the spatial relationship between nodes can be captured more comprehensively. The tree matrix replaces the original adjacency matrix and represents the graph structure in the form of a tree. The influence between nodes in the wireless network has a transmission effect, and the tree structure can more clearly represent all the nodes that have an impact on the root node, not just considering the neighbor nodes. This solves the problem of local path in graph convolution. The tree path convolution network is applied to wireless network fault detection. The tree path convolution network can convolve different paths and levels in the tree matrix. This way aggregates different hierarchical features while preserving important information, reducing the possibility of losing information. Moreover, the hierarchy of the tree matrix facilitates setting the size of the node correlation and enhances the ability to obtain spatial correlation. The tree path convolution network is compared with other comparative models on the competition data set. Experiments show that the tree path convolution network improves the accuracy by 2.98% compared with the graph convolution. Compared with other comparative models, it also performs better in precision, recall rate, and other aspects. In the confusion matrix, the path tree convolution also demonstrates its fault diagnosis capability in the multi-cause situation.

[0129] Reference Figure 3 The present embodiment further discloses a wireless network fault diagnosis system, comprising:

[0130] The data acquisition module 21 is configured to acquire graph data of a current wireless network, the graph data comprising a correlation matrix and an adjacency matrix between nodes.

[0131] The conversion module 22 is configured to convert the graph data into a tree matrix according to the correlation matrix and the adjacency matrix, and is further configured to,

[0132] perform a breadth-first traversal operation on the graph data according to the adjacency matrix to obtain a plurality of tree structures;

[0133] acquire all path data starting from any specified node in the tree structure respectively, and perform root node expansion according to the path data to ensure uniformity of lengths of the plurality of tree structures in matrix representation;

[0134] generate a corresponding one-dimensional tree matrix by taking each node of the tree structure as a root node, to construct a spatial tree matrix composed of a plurality of one-dimensional tree matrices;

[0135] convert node values of the spatial tree matrix into sizes of correlations between nodes according to the correlation matrix, to complete conversion of the tree matrix.

[0136] The convolution operation module 23 is configured to perform a spatial convolution operation on the tree matrix to obtain a spatial feature matrix thereof, and is further configured to,

[0137] calculate a weight matrix based on the spatial feature matrix of the tree matrix;

[0138] perform a time convolution operation on the weight matrix to obtain a time sequence feature matrix.

[0139] The fault diagnosis module 24 is configured to perform full connection processing and activation operation on the time sequence feature matrix to obtain a final classification result of node faults.

[0140] Figure 4 An electronic device entity structure schematic diagram provided by the embodiment of the present application is shown in FIG. 5. Figure 4 As shown in FIG. 5, the electronic device 50 comprises a processor 501, a memory 502 and a bus 503.

[0141] The processor 501 and the memory 502 can communicate with each other through the bus 503; the processor 501 is configured to invoke program instructions in the memory 502 to execute the method provided by each method embodiment.

[0142] The embodiment provides a non-transitory computer readable storage medium, which stores computer instructions, and the computer instructions make the computer execute the method provided by each method embodiment.

[0143] The embodiment further provides a computer program product comprising a computer program which, when executed by a processor, implements the method for diagnosing a fault of a wireless network.

[0144] Those skilled in the art can understand that all or part of the steps of the above-mentioned method embodiments can be completed by program instruction related hardware, and the foregoing program can be stored in a computer readable storage medium, and the program performs the steps of the above-mentioned method embodiments when executed; and the foregoing storage medium includes ROM, RAM, magnetic disc or optical disc and various storage media that can store program codes.

[0145] The apparatus embodiments described above are merely illustrative, wherein the units shown as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, i.e., can be located in one place, or can be distributed on a plurality of network units. Part or all of the modules can be selected to achieve the purpose of the embodiment scheme according to actual needs. Those skilled in the art can understand and implement without creative labor.

[0146] From the above description of the embodiments, those skilled in the art can clearly understand that the embodiments can be implemented by means of software and necessary universal hardware platforms, and of course can also be implemented by hardware. Based on such understanding, the above technical solutions, essentially or in other words, the part that contributes to the prior art, can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as ROM / RAM, magnetic disc, optical disc, etc., and includes a plurality of instructions for making a computer device (which can be a personal computer, a server, or a network device, etc.) execute the method of the embodiments or some parts of the embodiments.

[0147] The above specific embodiments do not constitute a limitation on the protection scope of the present application. Those skilled in the art should understand that various modifications, combinations, sub-combinations and substitutions can occur depending on design requirements and other factors. Any modification, equivalent replacement and improvement made within the spirit and principles of the present application shall fall within the scope of the present application.

Claims

1. A method for diagnosing wireless network faults, characterized in that: include: Obtain graph data of the current wireless network, including a correlation matrix and an adjacency matrix between nodes; Convert graph data into a tree matrix based on the correlation matrix and adjacency matrix; Perform spatial convolution on the tree matrix to obtain its spatial feature matrix; Calculate the weight matrix based on the spatial feature matrix of the tree matrix; Perform a time convolution operation on the weight matrix to obtain a time series feature matrix; Perform full connection and activation operations on the time series feature matrix to obtain the final node fault classification result.

2. The wireless network fault diagnosis method according to claim 1, wherein: The method for converting graph data into a tree matrix according to a correlation matrix and an adjacency matrix includes: Perform a breadth-first traversal operation on the graph data according to the adjacency matrix to obtain multiple tree structures; Get all path data starting from any specified node in the tree structure respectively, and expand the root node according to the path data to ensure the uniform length of multiple tree structures when represented in the matrix; Each node of the tree structure is taken as a root node to generate a corresponding one-dimensional tree matrix to construct a spatial tree matrix composed of multiple one-dimensional tree matrices; The node values ​​of the spatial tree matrix are converted into the magnitude of the correlation between nodes according to the correlation matrix to complete the transformation of the tree matrix.

3. The wireless network fault diagnosis method according to claim 1, wherein: The tree matrix is ​​subjected to spatial convolution operation to obtain its feature matrix. Specific methods include: The spatial convolution includes path convolution and layer convolution; The path convolution method includes: Each column of the tree matrix is ​​used as a local path in the path tree convolution; Perform a convolution operation upward from the bottom end of each local path of each one-dimensional tree matrix, and obtain a feature matrix that aggregates the local features of each node; Perform a circular convolution operation on the feature matrix column by column to obtain the high-level feature matrix of all local paths, and connect multiple high-level feature matrices to obtain the path feature matrix of each one-dimensional tree matrix; The hierarchical convolution method includes: Starting from each leaf node of each one-dimensional tree matrix, a convolution operation is performed horizontally to obtain a feature matrix that includes the node and the level information; Perform a circular convolution operation on the feature matrix to obtain high-level feature matrices of all local paths, and connect multiple high-level feature matrices to obtain a hierarchical feature matrix for each one-dimensional tree matrix; The path feature matrix and the hierarchical feature matrix are fused to obtain the spatial feature matrix of the tree matrix.

4. The wireless network fault diagnosis method according to claim 1, wherein: The method for calculating the weight matrix based on the spatial feature matrix of the tree matrix includes: Get the timing data of the current wireless network; Multiply the spatial feature matrix with the temporal data to obtain the weight matrix.

5. The wireless network fault diagnosis method according to claim 1, wherein: The method of performing a time convolution operation on the weight matrix to obtain a time series feature matrix includes: A Gaussian convolution operation is performed on the weight matrix to obtain high-dimensional time feature data in the time series data, thereby obtaining a time series feature matrix.

6. A wireless network fault diagnosis system, characterized in that: include: A data acquisition module is used to obtain graph data of the current wireless network, where the graph data includes a correlation matrix and an adjacency matrix between nodes; A conversion module for converting graph data into a tree matrix based on the correlation matrix and the adjacency matrix; A convolution operation module is used to perform a spatial convolution operation on the tree matrix to obtain its spatial feature matrix; Also used for Calculate the weight matrix based on the spatial feature matrix of the tree matrix; Perform a time convolution operation on the weight matrix to obtain a time series feature matrix; The fault diagnosis module is used to perform full connection processing and activation operations on the time series feature matrix to obtain the final node fault classification result.

7. The wireless network fault diagnosis system according to claim 6, wherein: The conversion module is further configured to: Perform a breadth-first traversal operation on the graph data according to the adjacency matrix to obtain multiple tree structures; Get all path data starting from any specified node in the tree structure respectively, and expand the root node according to the path data to ensure the uniform length of multiple tree structures when represented in the matrix; Each node of the tree structure is taken as a root node to generate a corresponding one-dimensional tree matrix to construct a spatial tree matrix composed of multiple one-dimensional tree matrices; The node values ​​of the spatial tree matrix are converted into the magnitude of the correlation between nodes according to the correlation matrix to complete the transformation of the tree matrix.

8. A computer-readable storage medium, characterized in that The storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 5 is implemented.

9. An electronic device comprising 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 method according to any one of claims 1 to 5 is implemented.

10. A computer program product comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 5.

Images