Method, device and equipment for identifying equipment with poor network quality, medium and product

Abstract

The invention provides a poor-network-quality equipment identification method and device, equipment, a medium and a product, and the method comprises the steps: obtaining the operation data of a plurality of pieces of network equipment belonging to the same local area network, and constructing a network equipment topological graph of the local area network based on the operation data; acquiring network feature data of the to-be-identified equipment based on the operation data, and inputting the network feature data into a trained poor quality identification model to obtain a poor quality identification result of the to-be-identified equipment output by the poor quality identification model; wherein the poor quality identification model is obtained by training based on multiple groups of training data, each group of training data comprises spliced feature data and poor quality identification tags corresponding to the spliced feature data, and the spliced feature data comprises sample network feature data and embedded features of the network equipment topological graph. According to the invention, accurate positioning of poor-quality equipment in a home indoor network can be realized.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to methods, apparatus, equipment, media and products for identifying devices with poor network quality. Background Technology

[0002] With the increasing number of smart home devices, the number of network-connected devices in homes is constantly growing, and the network architecture in homes has evolved from a single gateway or router to a multi-layered heterogeneous network. In a typical home environment, the indoor network structure is generally divided into three layers from top to bottom: the network entry layer, the network forwarding layer, and the terminal device layer. The network terminal devices involved include various types such as gateways, FTTR (Fiber to the Room), networking routers, smart speakers, security cameras, and set-top boxes, which leads to a significant increase in network load and further exacerbates the risk of network quality degradation.

[0003] Current technology cannot accurately locate devices with poor network quality in home indoor networks, which greatly affects the user's online experience. Summary of the Invention

[0004] This application provides a method, apparatus, equipment, medium, and product for identifying devices with poor network quality, in order to solve the shortcomings of existing technologies that cannot accurately locate devices with poor network quality in home indoor networks, and to achieve accurate location of devices with poor network quality in home indoor networks.

[0005] This application provides a method for identifying devices with poor network quality, including: Obtain operational data from multiple network devices belonging to the same local area network (LAN), and construct a network device topology diagram of the LAN based on the operational data; Based on the running data, network feature data of the device to be identified is obtained, and the network feature data is input into the trained poor quality identification model to obtain the poor quality identification result of the device to be identified output by the poor quality identification model. The poor quality identification model is trained based on multiple sets of training data. Each set of training data includes spliced ​​feature data and a poor quality identification label corresponding to the spliced ​​feature data. The spliced ​​feature data includes sample network feature data and the embedded features of the network device topology map.

[0006] According to the network quality poor device identification method provided in this application, after obtaining the quality poor identification result of the device to be identified output by the quality poor identification model, the method includes: When the poor quality identification result of the device to be identified reflects that the device to be identified is a poor quality device, the connection device to the device to be identified is determined based on the network device topology map; Obtain the quality defect identification result of the connected device and delineate the quality defect layer of the local area network.

[0007] According to the network quality poor device identification method provided in this application, the step of constructing a network device topology map of the local area network based on the operating data includes: The device type of the network device is determined based on the operational data; The connection relationships of the network devices are determined based on their device types. Based on the connection relationships of the network devices, a network device topology diagram is constructed.

[0008] According to the network quality poor device identification method provided in this application, the step of determining the device type of the network device based on the operating data includes: Based on the operational data of the network device, a matching search is performed in the sample device type library to obtain the first device type probability distribution value of the network device. The operating data of the network device is input into the trained device type identification model to obtain the second device type probability distribution value output by the device type identification model. The device type of the network device is determined based on the probability distribution values ​​of the first device type and the probability distribution values ​​of the second device type.

[0009] According to the method for identifying devices with poor network quality provided in this application, the step of obtaining network characteristic data of the device to be identified based on the operating data includes: Based on the importance of each data type in the operating data of the device to be identified, the data types in the operating data of the device to be identified are filtered to obtain the target data type of the device to be identified; The network feature data is determined based on the target data type of the device to be identified.

[0010] According to the network quality poor device identification method provided in this application, the step of filtering the data types in the operating data of the device to be identified based on the importance of each data type in the operating data of the device to be identified, to obtain the target data type of the device to be identified, includes: Based on the feature contribution values ​​of each data type in the operating data of the device to be identified to the identification of poor quality, the first importance score corresponding to each data type is determined; Based on the connection relationships of the network devices in the network device topology diagram, the connection importance of the network devices is determined, and based on the connection importance of each network device, a second importance score is determined for each data type. Based on the first importance score and the second importance score, the data types in the operating data of the device to be identified are filtered to obtain the target data type of the device to be identified.

[0011] This application also provides a device for identifying devices with poor network quality, including: The topology map construction module is used to obtain the operating data of multiple network devices belonging to the same local area network, and construct the network device topology map of the local area network based on the operating data; The poor quality identification module is used to obtain network feature data of the device to be identified based on the running data, input the network feature data into the trained poor quality identification model, and obtain the poor quality identification result of the device to be identified output by the poor quality identification model. The poor quality identification model is trained based on multiple sets of training data. Each set of training data includes spliced ​​feature data and a poor quality identification label corresponding to the spliced ​​feature data. The spliced ​​feature data includes sample network feature data and the embedded features of the network device topology map.

[0012] This application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement any of the network quality poor device identification methods described above.

[0013] This application also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the network quality poor device identification method as described above.

[0014] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the network quality poor device identification method as described above.

[0015] The network quality poor device identification method, apparatus, device, medium, and product provided in this application acquire operational data of multiple network devices belonging to the same local area network (LAN), construct a network device topology map of the LAN based on the operational data, and further acquire network feature data of the device to be identified based on the operational data. This network feature data is then input into a trained quality poor device identification model to obtain a quality poor device identification result output by the model, reflecting whether the device to be identified is a quality poor device. This quality poor device identification model is trained based on multiple sets of training data. Each set of training data includes sample network feature data, spliced ​​feature data of embedded features of the network device topology map, and a quality poor device identification label corresponding to the spliced ​​feature data. During the training process of the quality poor device identification model, the feature-enhanced network device topology map is added as additional information to the model, improving the model's identification ability. Furthermore, the node degree information in the network device topology map can be introduced during the model training phase to guide the learning of the quality poor device identification model, improve its performance, and thus improve the accuracy of network quality poor device identification, achieving accurate location of quality poor devices in home indoor networks. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart illustrating the method for identifying poor network quality devices provided in this application.

[0018] Figure 2 This is a schematic diagram of network device connections in an application scenario of the network quality poor device identification method provided in this application.

[0019] Figure 3 This is a schematic diagram of the device type identification process in the network quality poor device identification method provided in this application.

[0020] Figure 4 This is a schematic diagram of the network device topology in the network poor quality device identification method provided in this application.

[0021] Figure 5 This is a schematic diagram of the topology graph embedding feature acquisition process in the network poor quality device identification method provided in this application.

[0022] Figure 6 This is a schematic diagram of the decision tree sub-model in the network quality poor device identification method provided in this application.

[0023] Figure 7 This is a flowchart illustrating the process of defining poor network quality in the method for identifying poor network quality devices provided in this application.

[0024] Figure 8 This is a schematic diagram of the network quality poor device identification device provided in this application.

[0025] Figure 9 This is a schematic diagram of the structure of the electronic device provided in this application. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0027] It should be understood that, when used in this specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0028] It should also be understood that the terminology used in this application specification is for the purpose of describing particular embodiments only and is not intended to limit the application. As used in this application specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0029] It should also be further understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0030] As used in this specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrases "if determined" or "if detected [the described condition or event]" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once detected [the described condition or event]," or "in response to detection [the described condition or event]."

[0031] The following is combined Figure 1-7 This application describes the method for identifying poor network quality devices. For example... Figure 1 As shown, the method for identifying poor network quality devices includes the following steps: S110. Obtain the operating data of multiple network devices belonging to the same local area network, and construct a network device topology diagram of the local area network based on the operating data; S120. Obtain network feature data of the device to be identified based on the running data, input the network feature data into the trained poor quality identification model, and obtain the poor quality identification result of the device to be identified output by the poor quality identification model. The poor quality identification model is trained based on multiple sets of training data. Each set of training data includes spliced ​​feature data and the corresponding poor quality identification label. The spliced ​​feature data includes sample network feature data and embedded features of network device topology.

[0032] The method for identifying poor-quality network devices provided in this application obtains operational data from multiple network devices belonging to the same local area network (LAN), constructs a network device topology map of the LAN based on the operational data, and further obtains network feature data of the device to be identified based on the operational data. This network feature data is then input into a trained poor-quality identification model to obtain a poor-quality identification result output by the model, reflecting whether the device to be identified is a poor-quality device. This poor-quality identification model is trained based on multiple sets of training data. Each set of training data includes sample network feature data, spliced ​​feature data of embedded features from the network device topology map, and a corresponding poor-quality identification label. During the training process of the poor-quality identification model, the feature-enhanced network device topology map is added as additional information to the model, improving its identification capability. Furthermore, the node degree information in the network device topology map can be introduced during the model training phase to guide the learning of the poor-quality identification model, improve its performance, and thus enhance the accuracy of identifying poor-quality network devices, achieving accurate location of poor-quality devices in home indoor networks.

[0033] In the method provided in this application, the local area network (LAN) can be a home network or other types of networks, such as an office building network or an office network. The following explanation uses a home network as an example.

[0034] Based on the role of home internet access devices in the network, home indoor networks can be divided into a three-layer structure, such as... Figure 2As shown, the network structure consists of three layers: the gateway layer (network entry layer), the router layer (network forwarding layer), and the downstream device layer (network usage layer). The first layer comprises terminal devices such as smart gateways, serving as the unified entry point for the indoor network. The router layer below mainly consists of routers, mesh networking devices, and other equipment responsible for distributing and forwarding network traffic. The downstream device layer represents the actual network usage terminals, typically including wirelessly connected devices such as mobile phones, home cameras, smart speakers, and computers, as well as wired devices such as smart TVs and smart boxes.

[0035] The method provided in this application aims to accurately identify poor-quality devices in a home network. These devices are those with poor network quality. Big data technology can be used to collect operational data from each network device in the home network, analyze this data, construct a network device topology map of the home network, and then use this topology map to identify poor-quality devices.

[0036] Operational data reflects the operational characteristics of network devices. Specifically, operational data can be collected using pre-installed software probe plugins within network devices. Real-time analysis of the operational data reported by these plugins provides a high-quality data source for subsequent model processing. Taking a home indoor network divided into gateway, router, and downstream device layers as an example, different network layers have different devices. Different types of operational data can be collected based on the operational characteristics of the network devices at each layer. Examples are provided below for each layer.

[0037] The gateway layer is the entry point for the entire home indoor network and typically includes devices such as smart networks and FTTR. The key device operation data mainly consists of indicators such as device model, CPU utilization, memory utilization, and signal duty cycle. In one possible implementation, the data types of network devices in the gateway layer are shown in Table 1.

[0038] Table 1

[0039] The router layer, as the middle layer of the indoor network, is a crucial component of the home network. It forwards gateway traffic upstream and distributes this forwarded traffic evenly to various downstream terminal devices downstream. The main device types include routers, network topologies, access points (APs), and switches. The operational data to be collected includes device model, probe plugin version, downstream device information, and operating mode, as shown in Table 2.

[0040] Table 2

[0041] The devices in the downstream device layer are actual network usage examples, representing traffic users in a home network. Common downstream device types are mainly divided into two categories based on connection method: wireless connection and wired connection. Wireless connections include mobile phones, smart speakers, security cameras, smart door locks, and smart TVs, while wired connection devices mainly include set-top boxes, NAS devices, and fixed workstations. The corresponding operational data to be collected is shown in Table 3, mainly including statistical indicators such as device model, maximum transmitted traffic, and average received traffic.

[0042] Table 3

[0043] Because the connections between network devices in a home network are complex and diverse, in order to accurately and comprehensively describe the connections between network devices, this application constructs a network device connection topology diagram based on the operational data after obtaining the operational data. The nodes in the network device connection topology diagram correspond to the network device types, and the edges between the nodes reflect the connection relationships between the network devices.

[0044] In one possible implementation, the connectivity between network devices can be obtained through user input data. However, due to the specialized nature of these connectivity relationships, understanding them presents a significant barrier for users. Therefore, another possible implementation of the method provided in this application involves constructing a network device topology diagram of the local area network based on operational data, including: Determine the device type of network equipment based on operational data; Determine the connection relationships of network devices based on their device types; Based on the connection relationships of network devices, construct a network device topology diagram.

[0045] In practical applications, different device types often have specific connection methods. Therefore, determining the device type first, and then determining the connection relationship between network devices based on the device type, can accurately determine the connection relationship between network devices without increasing the burden on users.

[0046] In one possible implementation, determining the device type of a network device based on operational data can be done using a knowledge-based matching approach. This involves pre-collecting operational data of various device types and storing it in a sample device type library. When it's necessary to determine the device type of a particular network device, a matching search is performed within the sample device type library based on the network device's operational data. The device type of the most matching sample in the library is then taken as the device type of that network device. In another possible implementation, a device type recognition model can be trained, and this model can be used to predict the device type of network devices.

[0047] Furthermore, in one possible implementation, the two methods described above can be combined to obtain a more accurate device type identification result. Specifically, determining the device type of the network device based on operational data includes: Based on the operational data of network devices, a matching search is performed in the sample device type library to obtain the first device type probability distribution value of the network devices. The network device's operating data is input into the trained device type recognition model to obtain the second device type probability distribution value output by the device type recognition model. The device type of the network device is determined based on the probability distribution values ​​of the first device type and the probability distribution values ​​of the second device type.

[0048] The sample device type library includes various sample device types and corresponding operating data. Based on the similarity between the network device's operating data and the operating data of the sample device types in the sample device type library, a matching search is performed to obtain the first device type probability distribution value of the network device. The first device type probability distribution value includes the probability that the network device is each of the various device types.

[0049] The device type identification model is trained based on multiple sets of training data. Each set of training data includes sample running data and the device type label corresponding to the sample running data. Through training, the device type identification model can learn the characteristics of the running data of different device types. Thus, after the running data of the network device is input into the trained device type identification model, the device type identification model can output the probability that the network device is of each device type, that is, output the second device type probability distribution value.

[0050] like Figure 3 As shown, in one possible implementation, the network device's operational data is used to perform a matching search in a sample device type library to obtain the first device type probability distribution value of the network device. This can be specifically implemented using the AC automaton algorithm, while the device type recognition model can be a convolutional neural network model.

[0051] After obtaining the first device type probability distribution value P1 and the second device type probability distribution value P2, the two are weighted and fused to obtain the final device type probability distribution value, which can be expressed by the formula: , where y is the final device type probability distribution value, and α and β are weights. The device type with the highest probability in the final device type probability distribution value is taken as the device type of the network device.

[0052] After determining the device types of network devices, the connection relationships between them can be established using prior knowledge of the connections between different device types. For example, a smart surveillance camera and a router might be connected via a wireless network. This prior knowledge can be obtained by collecting user manuals and technical specifications for different device types. In one possible implementation, web scraping technology can be used to obtain profile information for each model within each device type, such as manufacturing date, model number, connection method, and wireless speed, forming a device fingerprint database. Based on this database, a network device topology map can then be constructed.

[0053] In a network device topology graph, a node V (Vertex) represents the device type, and edges are described using connections and traffic dependencies, forming a directed acyclic graph (DAG) between the devices. For example... Figure 4 As shown, nodes represent various terminal devices such as gateways, routers, smart speakers, and set-top boxes, and edges between nodes represent the connection methods between devices, such as wired, wireless, or virtual. This descriptive method brings together the complex network structure of home indoor terminal devices into a single topology graph, allowing for a quick and intuitive understanding of the relationships between devices. The final network device topology graph is represented as G=(V, E), where V is the set of all nodes and E is the set of all edges.

[0054] Based on the operational data of network devices, network feature data of network devices can be constructed. In one possible implementation, the operational data of network devices can be directly used as network feature data. However, since the operational data of network devices includes various data types, not all of these data types necessarily provide valuable information for identifying network device quality issues. A greater variety of data types increases the dimensionality of the data input to the quality issue identification model, increasing the computational load of the model, and may also lead to poor model training performance due to potential noise features. In another possible implementation of the method provided in this application, the various data types in the operational data are first filtered, retaining the data types that are more important for identifying network device quality issues. Network feature data is then constructed only based on these data types, which can reduce data dimensionality, reduce model computational load, and improve inference efficiency. In other words, in this implementation, the network feature data of the device to be identified is obtained based on the operational data, including: Based on the importance of each data type in the operational data of the device to be identified, the data types in the operational data of the device to be identified are filtered to obtain the target data type of the device to be identified; Network feature data is determined based on the target data type of the device to be identified.

[0055] In one possible implementation, the importance of each data type in the runtime data can be determined by identifying the feature contribution value of each data type through model training, and then using this feature contribution value to determine the target data type. For example, a simple model (such as a random forest model) can be built and trained multiple times in mini-batch. Since the data types in the runtime data collected for each device type are not entirely the same, a training dataset can be built for each device type to train the simple model. The training dataset includes multiple sample runtime data of the same device type and the corresponding quality / poor labels for the sample runtime data. During training, the SHAP (feature contribution value) analysis method is used to obtain the importance of each data type for quality / poor identification. Based on the feature contribution value, the data types in the runtime data of that device type are ranked, and the top k data types with the highest feature contribution values ​​are taken as the target data type for that device type. For example, for gateway devices, the importance ranking of data types is as follows: {main chip temperature, CPU utilization, device model, device runtime, number of connected devices, ...}. Based on the device type of the device to be identified, the pre-determined target data type for that device type can be found.

[0056] If a network device is highly interconnected with other network devices, then the influence of these devices on and the network device itself will be greater, making the operational data of that device more important. In another possible implementation, besides determining the target data type based on feature contribution values, the network device topology map is also used to determine the target data type. In this implementation, the data types in the operational data of the device to be identified are filtered based on their importance to obtain the target data type of the device to be identified, including: Based on the feature contribution values ​​of each data type in the operating data of the device to be identified to the identification of poor quality, the first importance score corresponding to each data type is determined; Based on the connection relationships of network devices in the network device topology diagram, the connection importance of network devices is determined, and based on the connection importance of each network device, the second importance score corresponding to each data type is determined. Based on the first importance score and the second importance score, the data types in the operational data of the device to be identified are filtered to obtain the target data type of the device to be identified.

[0057] The first importance score can be determined based on the ranking of the feature contribution values ​​of the data types, which has been described earlier.

[0058] Based on the connection relationships of network devices in the network device topology diagram, the connection importance of network devices is determined. When the connection importance of a network device is higher, a higher second importance score is assigned to each data type in the operational data of that network device.

[0059] Specifically, the connectivity importance of network devices can be measured based on two metrics: degree centrality and / or proximity centrality. Degree centrality refers to the number of edges directly connected to the network device in the network topology graph, reflecting its local connectivity. Proximity centrality refers to the reciprocal of the average shortest path length from the network device to all other network devices. A linear combination of degree centrality and proximity centrality can then be used to merge these metrics, resulting in a score for each network device in the network topology graph, which is considered its connectivity importance. Afterward, the connectivity importance of each network device in the topology graph is ranked, and network devices with higher connectivity importance are assigned a higher secondary importance score for the data types included in their operational data.

[0060] The first and second importance scores of the data types in the operational data of the device to be identified can be fused, for example, by weighted summation, to obtain the fused importance score of each data type in the operational data of the device to be identified. Based on the fused importance score, multiple target data types are selected as the data types for constructing network feature data.

[0061] By combining the first and second importance scores to select target data types for constructing network feature data, we not only consider the impact of machine learning models on feature scoring, but also integrate the importance of topological relationships between devices on features. This approach can reduce the computational load of the model and improve its performance, thereby obtaining more accurate results for identifying quality defects.

[0062] After obtaining the network feature data of the device to be identified, the network feature data of the device to be identified is input into the trained poor quality identification model to obtain the poor quality identification result of the device to be identified output by the poor quality identification model. In the method provided in this application, structural information of the network device topology is also introduced when training the poor quality identification model to enhance the poor quality identification model.

[0063] In one possible implementation, the poor quality identification model can use an existing classification model to classify whether a device is of poor quality.

[0064] In another possible implementation, to improve the performance of the poor quality identification model and output more accurate results, the poor quality identification model can employ a combined model containing multiple classification decision trees. During the first training iteration, a decision tree model is initialized, and then the information gain metric is used to find the optimal classification point from the numerous input features as the root node of the tree model. The information gain can be calculated using the following formula: ; Where Gain is the information gain. , These are the sum of the gradients of the left and right child nodes after the split, respectively. and These are the sums of the second-order gradients of the left and right child nodes after the split, respectively. It is a regularization parameter used to prevent overfitting. The parameter is an additional penalty term that represents the cost of splitting a node.

[0065] Furthermore, in another possible implementation, in addition to using a combined model containing multiple classification decision trees as the quality defect identification model, the degree information of nodes in the network device topology graph can be incorporated. This allows the network device topology graph to guide the splitting of tree nodes in the quality defect identification model. In this implementation, the method for calculating information gain is further improved as follows: ; Here, D is an adjustment factor based on the node degree (i.e., the number of edges connected to a node) of the network topology device graph. The number of edges connected to each node in the network topology graph truly reflects the importance or activity level of that node in the home network. For example, the most edges connected to the networking device indicate that the device plays a crucial role in the home intranet. This implementation improves upon the node splitting method by introducing a splitting principle guided by the topology graph. Specifically, it adds an adjustment factor D based on the node degree of the topology graph to the information gain formula. By splitting the nodes of the decision tree model using this improved information gain calculation method, the key factor of the device's importance in the home intranet can be integrated into the quality difference identification model, thereby improving the accuracy of the model's predictions.

[0066] In the method provided in this application, during the training process of the poor quality identification model, not only sample network feature data is used, but also the embedded features of the network device topology graph are used as supplementary features. This makes the trained poor quality identification model highly sensitive to device type. Specifically, an adjacency matrix is ​​first constructed based on the network device topology graph. Each value in the matrix represents a connection method or traffic dependency between network devices; for example, 1 indicates a virtual connection between devices, while 0 indicates no network interaction. Then, the Node2Vec method is used to obtain the vectorized representation of the adjacency matrix, resulting in the embedding features of the topology graph. The entire process of embedding network device topology graph features is as follows: Figure 5 As shown.

[0067] Embedded features from the network device topology map are used as supplementary features and fused with sample network feature data to form the complete training data input metric, constructing a poor-quality training dataset. Then, using the node splitting principle, the information gain data is calculated for each feature sequentially, and the feature with the largest gain is selected as the splitting node. This process is repeated until no further splitting is possible or the maximum depth is reached. The trained poor-quality identification model then consists of multiple decision tree sub-models. The decision tree sub-models obtained in the first iteration are visualized, as shown below. Figure 6 As shown in the diagram, the poor quality identification model first judges based on the downlink traffic of the device's LAN port. If it is less than 0.057, it continues to judge the uplink traffic of the 2.4G connection, and so on, until it reaches the leaf node and can no longer split, thus determining whether the current device has experienced poor quality.

[0068] Using a trained poor quality identification model, the network feature data of the device to be identified is input, and the model outputs whether the device to be identified has poor network quality, that is, the output of the poor quality identification result of whether the device to be identified is a poor quality device.

[0069] After obtaining the quality defect identification result of the device to be identified, a quality defect analysis result of the home network can be generated based on this result. This result can include which network devices are considered to have poor quality, allowing technicians to analyze and improve the network. Furthermore, in one possible implementation, after obtaining the quality defect identification result of the device to be identified, the network quality defects can be further delimited to determine the network layer where the root cause of the defects lies. In this implementation, after obtaining the quality defect identification result of the device to be identified output by the quality defect identification model, the following steps are included: When the quality identification result of the device to be identified reflects that the device to be identified is a poor quality device, the connected devices to the device to be identified are determined based on the network device topology map; Obtain the quality defect identification results of connected devices and delineate the quality defect layer of the local area network.

[0070] When the device to be identified is determined to be a poor-quality device, such as Figure 7As shown, based on the network device topology diagram, the network quality status of other devices connected to the device is further queried to determine which network layer the device is experiencing degradation, thus achieving quality degradation. For example, when the model identifies a network device experiencing quality degradation, and all downstream devices connected to it in the topology diagram also experience quality degradation, while the upstream gateway device has a normal network, then it can be determined that the router terminal device in the current indoor network has quality degradation, thereby accurately degrading the poor quality network in the home.

[0071] By using the network device topology structure, after identifying poor-quality devices, it is possible to further implement layered delineation of poor-quality home networks and then proactively intervene to solve the problem of excessive complaints caused by the untimely detection of home broadband network degradation.

[0072] The following describes the network quality poor device identification device provided in this application. The network quality poor device identification device described below can be referred to in correspondence with the network quality poor device identification method described above. For example... Figure 8 As shown, the network quality poor device identification device provided in this application includes: The topology map construction module 810 is used to obtain the operating data of multiple network devices belonging to the same local area network and construct the network device topology map of the local area network based on the operating data. The poor quality identification module 820 is used to obtain network feature data of the device to be identified based on the running data, input the network feature data into the trained poor quality identification model, and obtain the poor quality identification result of the device to be identified output by the poor quality identification model. The poor quality identification model is trained based on multiple sets of training data. Each set of training data includes spliced ​​feature data and the corresponding poor quality identification label. The spliced ​​feature data includes sample network feature data and embedded features of network device topology.

[0073] Figure 9 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 9As shown, the electronic device may include: a processor 910, a communications interface 920, a memory 930, and a communication bus 940. The processor 910, communications interface 920, and memory 930 communicate with each other via the communication bus 940. The processor 910 can call logical instructions in the memory 930 to execute a network quality poor device identification method. This method includes: acquiring operational data of multiple network devices belonging to the same local area network (LAN); constructing a network device topology map of the LAN based on the operational data; acquiring network feature data of the device to be identified based on the operational data; inputting the network feature data into a trained quality poor identification model; and obtaining the quality poor identification result of the device to be identified output by the quality poor identification model. The quality poor identification model is trained based on multiple sets of training data. Each set of training data includes spliced ​​feature data and a quality poor identification label corresponding to the spliced ​​feature data. The spliced ​​feature data includes sample network feature data and embedded features of the network device topology map.

[0074] Furthermore, the logical instructions in the aforementioned memory 930 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0075] On the other hand, this application also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the network quality poor device identification method provided by the above methods. The method includes: acquiring the operating data of multiple network devices belonging to the same local area network, and constructing a network device topology map of the local area network based on the operating data; acquiring network feature data of the device to be identified based on the operating data, inputting the network feature data into a trained quality poor identification model, and obtaining the quality poor identification result of the device to be identified output by the quality poor identification model; wherein, the quality poor identification model is trained based on multiple sets of training data, each set of training data includes spliced ​​feature data and quality poor identification labels corresponding to the spliced ​​feature data, and the spliced ​​feature data includes sample network feature data and embedded features of the network device topology map.

[0076] In another aspect, this application also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program is implemented to perform the network quality poor device identification method provided by the above methods. The method includes: acquiring operating data of multiple network devices belonging to the same local area network (LAN); constructing a network device topology map of the LAN based on the operating data; acquiring network feature data of the device to be identified based on the operating data; inputting the network feature data into a trained quality poor identification model; and obtaining the quality poor identification result of the device to be identified output by the quality poor identification model. The quality poor identification model is trained based on multiple sets of training data. Each set of training data includes spliced ​​feature data and a quality poor identification label corresponding to the spliced ​​feature data. The spliced ​​feature data includes sample network feature data and embedded features of the network device topology map.

[0077] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0078] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0079] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for identifying devices with poor network quality, characterized in that, include: Obtain operational data from multiple network devices belonging to the same local area network (LAN), and construct a network device topology diagram of the LAN based on the operational data; Based on the running data, network feature data of the device to be identified is obtained, and the network feature data is input into the trained poor quality identification model to obtain the poor quality identification result of the device to be identified output by the poor quality identification model. The poor quality identification model is trained based on multiple sets of training data. Each set of training data includes spliced ​​feature data and a poor quality identification label corresponding to the spliced ​​feature data. The spliced ​​feature data includes sample network feature data and the embedded features of the network device topology map.

2. The method for identifying poor network quality devices according to claim 1, characterized in that, After obtaining the quality defect identification result of the device to be identified output by the quality defect identification model, the process includes: When the poor quality identification result of the device to be identified reflects that the device to be identified is a poor quality device, the connection device to the device to be identified is determined based on the network device topology map; Obtain the quality defect identification result of the connected device and delineate the quality defect layer of the local area network.

3. The method for identifying poor network quality devices according to claim 1, characterized in that, The construction of the network device topology diagram of the local area network based on the operational data includes: The device type of the network device is determined based on the operational data; The connection relationships of the network devices are determined based on their device types. Based on the connection relationships of the network devices, a network device topology diagram is constructed.

4. The method for identifying poor network quality devices according to claim 3, characterized in that, Determining the device type of the network device based on the operational data includes: Based on the operational data of the network device, a matching search is performed in the sample device type library to obtain the first device type probability distribution value of the network device. The operating data of the network device is input into the trained device type identification model to obtain the second device type probability distribution value output by the device type identification model. The device type of the network device is determined based on the probability distribution values ​​of the first device type and the probability distribution values ​​of the second device type.

5. The method for identifying poor network quality devices according to claim 1, characterized in that, The process of obtaining network feature data of the device to be identified based on the operational data includes: Based on the importance of each data type in the operating data of the device to be identified, the data types in the operating data of the device to be identified are filtered to obtain the target data type of the device to be identified; The network feature data is determined based on the target data type of the device to be identified.

6. The method for identifying poor network quality devices according to claim 5, characterized in that, The method of filtering the data types in the operational data of the device to be identified based on the importance of each data type to obtain the target data type of the device to be identified includes: Based on the feature contribution values ​​of each data type in the operating data of the device to be identified to the identification of poor quality, the first importance score corresponding to each data type is determined; Based on the connection relationships of the network devices in the network device topology diagram, the connection importance of the network devices is determined, and based on the connection importance of each network device, a second importance score is determined for each data type. Based on the first importance score and the second importance score, the data types in the operating data of the device to be identified are filtered to obtain the target data type of the device to be identified.

7. A device for identifying devices with poor network quality, characterized in that, include: The topology map construction module is used to obtain the operating data of multiple network devices belonging to the same local area network, and construct the network device topology map of the local area network based on the operating data; The poor quality identification module is used to obtain network feature data of the device to be identified based on the running data, input the network feature data into the trained poor quality identification model, and obtain the poor quality identification result of the device to be identified output by the poor quality identification model. The poor quality identification model is trained based on multiple sets of training data. Each set of training data includes spliced ​​feature data and a poor quality identification label corresponding to the spliced ​​feature data. The spliced ​​feature data includes sample network feature data and the embedded features of the network device topology map.

8. 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, it implements the network poor quality device identification method as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the network quality poor device identification method as described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the network quality poor device identification method as described in any one of claims 1 to 6.

Images