Home network monitoring method, system and home network monitoring device

Abstract

The present invention discloses a home network monitoring method, system and home network monitoring device. The home network monitoring method of the present invention simulates different network quality scenarios through TC instructions, and performs operations of different applications in different application scenarios through different types of terminal devices in different network quality scenarios, collects and processes corresponding data during the operation, and generates adversarial network training through the processed data to generate prediction data reflecting future network quality, and obtains multiple different network quality scoring models through different data training, so that the network quality of the whole family, different application scenarios and different applications can be predicted according to the prediction data, which is convenient for determining whether the overall family network is poor or the network situation is poor in a specific application scenario or when using a certain application, and timely feedback to users when the network quality is poor, providing more accurate data support for improving user experience.

Description

Technical Field

[0001] The present invention relates to the field of network technology, and in particular to a home network monitoring method, system and home network monitoring device. Background Art

[0002] The quality of the network has a great impact on the user's Internet experience. Therefore, it is necessary to detect and predict the quality of the network used by users.

[0003] The main method of current network quality detection is to directly evaluate the collected network traffic data. The evaluation results of this method have a certain lag and cannot predict subsequent changes in user network quality, which is not conducive to timely following up on user experience.

[0004] There are also some methods to predict network quality in advance, which are based on mobile communication (3G, 4G, 5G) data to predict the network quality score. However, the results of this method cannot determine the corresponding network quality prediction in what application scenario, and lack targeted analysis. Summary of the invention

[0005] The purpose of the present invention is to solve the above problems existing in the prior art and to provide a home network monitoring method, system and home network monitoring device.

[0006] The purpose of the present invention is achieved through the following technical solutions: The home network monitoring method comprises the following steps: S10, obtaining network traffic data of a home gateway running for a period of time; S20, processing the network traffic data for the period of time to obtain data to be evaluated; S30, determining an application scenario corresponding to the data to be evaluated according to the data to be evaluated; S40, when it is determined that the application scenario is an air interface competition scenario, inputting the data to be evaluated into a generative adversarial network to generate prediction data, and inputting the prediction data into a network quality scoring model corresponding to the air interface competition scenario to determine a network quality score in the air interface competition scenario; S50, when it is determined that the application scenario is a video scenario, a game scenario, or an office scenario, inputting the data to be evaluated into a generative adversarial network to generate prediction data, and inputting the prediction data into a network quality scoring model corresponding to the video scenario, the game scenario, or the office scenario to determine the network quality score in the video scenario, the game scenario, or the office scenario; S50, when it is determined that the network quality score in the air interface competition scenario, video scenario, game scenario, or office scenario is lower than the set value, the network quality prediction result in the air interface competition scenario, video scenario, game scenario, or office scenario is fed back to the terminal device under the home gateway.

[0007] Preferably, in S30, whether it is an air interface contention scenario is determined according to the total number of downstream devices.

[0008] Preferably, in S30, the application scenario is determined to be a video scenario, a game scenario, or an office scenario according to the application identified by the data to be evaluated; In the S40, the predicted data is also input into a network quality scoring model corresponding to the identified application to determine a network quality score when the application is used; In the S50, when it is determined that the network quality score when using the application is lower than the set value, the network quality prediction result when using the application is sent to the terminal device connected to the home gateway.

[0009] Preferably, in said S40, after generating the prediction data, the prediction data is input into a home overall network quality scoring model to determine the home overall network quality score; In the S50, when it is determined that the overall home network quality score is lower than the set value, the overall home network quality prediction result is sent to the terminal device connected to the home gateway.

[0010] Preferably, in S50, information on whether the home gateway can be restarted is sent to a terminal device connected to the home gateway.

[0011] Preferably, the activity status of the terminal devices within the period of time is determined based on the network traffic data of the period of time. When it is determined that the activity status of the terminal devices is inactive, it is determined whether information on whether the home gateway can be restarted is sent to the terminal devices under the home gateway; when it is determined that the activity status of the terminal devices is active, it is determined not to send information on whether the home gateway can be restarted to the terminal devices under the home gateway.

[0012] Preferably, each time the network traffic data for a period of time is obtained, the busyness of the home gateway and / or the activity of the terminal device are determined based on the network traffic data for the period of time, and whether the home gateway can be controlled to restart is determined accordingly.

[0013] Preferably, when it is determined that the home gateway device is not busy and / or the terminal device is not active, it is determined that the home gateway can be controlled to restart.

[0014] Home network monitoring system, including: A data acquisition unit, used to acquire network traffic data of a home gateway running for a period of time; A data processing unit, used for processing the network traffic data for a period of time to obtain data to be evaluated; An application scenario identification unit, used to determine the application scenario corresponding to the data to be evaluated according to the data to be evaluated; A score prediction unit, configured to, when determining that the application scenario is an air interface competition scenario, input the data to be evaluated into a generative adversarial network to generate prediction data, and input the prediction data into a network quality scoring model corresponding to the air interface competition scenario to determine a network quality score in the air interface competition scenario; When it is determined that the application scenario is a video scenario, a game scenario, or an office scenario, inputting the data to be evaluated into a generative adversarial network to generate prediction data, and inputting the prediction data into a network quality scoring model corresponding to the video scenario, the game scenario, or the office scenario to determine the network quality score in the video scenario, the game scenario, or the office scenario; A feedback unit is used to feedback the network quality prediction result in the air interface competition scenario, video scenario, game scenario or office scenario to the terminal device under the home gateway when it is determined that the network quality score in the air interface competition scenario, video scenario, game scenario or office scenario is lower than a set value.

[0015] A home network monitoring device includes a memory and a processor, wherein the memory stores a program executable by the processor, and when the program is executed, any of the above-mentioned home network monitoring methods is implemented.

[0016] The advantages of the technical solution of the present invention are mainly reflected in: The present invention simulates different network quality scenarios through TC instructions, and performs operations of different applications in different application scenarios through different types of terminal devices in different network quality scenarios, collects and processes corresponding data during the operation, and trains a generative adversarial network through the processed data to generate prediction data reflecting future network quality, and obtains a variety of different network quality scoring models through different data training, so that the network quality of the whole family, different application scenarios, and different applications can be predicted according to the predicted data, which is convenient for determining whether the overall family network is poor, or the network situation is poor in a specific application scenario or when using a certain application, and timely feedback to users when the network quality is poor, providing more accurate data support for improving user experience.

[0017] The present invention can determine whether to send information for restarting the home gateway to the user according to the activity status of the terminal device determined over a period of time, and can better provide processing suggestions for the user in combination with actual conditions and perform processing according to user feedback.

[0018] The present invention can continuously monitor the operation of the home gateway and the activity of the terminal devices according to the continuously collected network traffic data, and can restart the home gateway without affecting the user's use as much as possible without manual intervention, thereby ensuring the long-term and reliable operation of the home gateway. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a flow chart of the home network monitoring method of the present invention; Figure 2 The home network monitoring method of the present invention has a flow chart of whether to perform a home gateway restart reminder process; Figure 3 It is a schematic diagram of the process of determining whether the home gateway is restarted in the present invention. DETAILED DESCRIPTION

[0020] The purpose, advantages and features of the present invention will be illustrated and explained by the non-limiting description of the following preferred embodiments. These embodiments are only typical examples of the application of the technical solution of the present invention, and any technical solution formed by equivalent replacement or equivalent transformation falls within the scope of protection claimed by the present invention.

[0021] In the description of the scheme, it should be noted that the terms "center", "upper", "lower", "left", "right", "front", "back", "vertical", "horizontal", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience and simplification of description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and "third" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance.

[0022] Example 1 The home network monitoring method disclosed in the present invention is explained below in conjunction with the accompanying drawings. The home network monitoring method is based on application scenario identification and a trained generative adversarial network and a network quality scoring model corresponding to a variety of application scenarios to predict and score the network quality of a home user using a wireless network in each application scenario.

[0023] The following first explains how to obtain the generative adversarial network and the network quality scoring model corresponding to each application scenario.

[0024] The present invention first simulates different network quality scenarios by setting up a home gateway, then uses a terminal device to perform different network usage operations in the simulated various network quality scenarios, and uses a packet capture tool to collect network traffic data generated when performing different network usage operations in different network quality scenarios. After processing the collected network traffic data, the generative adversarial network and the network quality scoring model corresponding to each application scenario are trained.

[0025] Specifically, the TC command is used to set the delay and packet loss rate of the home gateway to simulate a variety of network quality scenarios with different delays and packet loss rates; the TC command can manage and control network traffic by defining different queue rules. Among them, the delay of different network quality scenarios is between 0-1000ms, and the packet loss rate of different network quality scenarios is between 0%-50%. For example, in the video scene, the delay of different network quality scenarios is set to 0ms, 300ms, 500ms, and 1000ms, respectively, and the packet loss rate is set to 0%, 10%, 20%, 30%, 40%, and 50%, respectively, so that 24 network quality scenarios can be obtained.

[0026] In each simulated network quality scenario, the wireless network is used in different test scenarios and network traffic data during use is obtained, and the obtained network traffic data is saved in the form of data streams according to a certain granularity.

[0027] The use of the wireless network in different test scenarios refers to using the wireless network in different application scenarios and through different applications through different types of terminal devices.

[0028] The different types of terminal devices include but are not limited to mobile phones, laptops, desktop computers, tablet computers, Internet TVs, smart speakers, etc.

[0029] The application scenarios include but are not limited to video scenarios, game scenarios, office scenarios and air interface competition scenarios. The video scenario is the application scenario when watching videos through a terminal device, for example, watching long videos, short videos normally, refreshing short video pages, watching live broadcasts, video skipping (i.e. pulling the video progress bar), etc. The game scenario is the application scenario when playing online games normally through a terminal device. The office scenario is the application scenario when using a terminal device for online office operations, for example, using conference software to conduct video conferences, share screens, and voice conferences; using remote software for remote office, etc. The air interface competition scenario is that multiple terminal devices are online at the same time. During testing, one terminal device is selected as the test device, and the remaining terminal devices are identified as interference devices. The interference devices are responsible for using applications with higher network speed requirements to interfere with the test devices.

[0030] The applications include video applications, game applications, and office applications.

[0031] The video applications include, for example, Tik Tok, Kuaishou, Xiaohongshu, iQiyi, Youku Video, Tencent Video, Bilibili, etc.; the game applications include, for example, Honor of Kings, Peace Elite, Genshin Impact, Onmyoji, Dream Star, etc.; the office applications include, for example, Tencent Meeting, DingTalk, Changlian, Zoom, Sunflower, ToDesk, WeChat, QQ, etc.

[0032] In addition, the duration of using the wireless network in each test scenario is not less than 30 minutes, so that the usage can be as consistent as possible with normal user behavior. Of course, the upper limit of the duration of using the wireless network can be set as needed. Considering the factors of data processing volume and traffic size, the duration of using the wireless network in each test scenario is preferably no more than 1 hour. When using the wireless network, the data of the home gateway is captured by a packet capture tool and stored at a certain granularity. The certain granularity is in minutes. The certain granularity is, for example, 1 minute, and of course, it can also be 2 minutes, 5 minutes, etc., which is specifically designed according to needs.

[0033] For example, a mobile phone is used as a terminal device, and a video is watched in a network quality scenario with a delay of 0 and a packet loss rate of 0%. The video application used is Tik Tok, and the video watching time is, for example, 40 minutes. The network traffic data of the 40-minute video watching on the mobile phone is collected through the packet capture tool, and the 40-minute network traffic data is stored as multiple data streams with a granularity of 1 minute.

[0034] At the same time, network traffic data of using wireless networks in different real home scenarios is obtained and saved in the form of data streams at a certain granularity. The real home scenario is a real user environment. The network traffic data of using wireless networks in real home scenarios can be processed and used as real data required for subsequent discriminator discrimination of the generated adversarial network.

[0035] All acquired data streams are processed separately to obtain the data sets required for model training.

[0036] The original data of each data flow includes but is not limited to the following fields: time, src_ip, dst_ip, src_mac, dst_port, protoname, splt_direction, splt_piat_ms, where time indicates the establishment time of a flow generated by a user accessing an application, src_ip and src_mac indicate the source IP address and source MAC address, dst_ip and dst_port indicate the destination IP address and destination port, protoname indicates the application layer protocol of each flow, splt_direction indicates the flow direction of the first 16 packets accessing a flow, and splt_piat_ms indicates the delay of accessing the first 16 packets of a flow.

[0037] All the data streams are filtered and aggregated. The filtering is to filter out outliers in the data streams. The dimensions of the data aggregation extension include the average delay of all streams of each data stream, the first packet delay, the total delay, the application layer delay, and the network layer delay.

[0038] In addition, a network quality experience label is added to each data stream, and the network quality experience label is used as a reference for model classification and a basis for verifying the accuracy of model prediction. The network quality experience label is determined based on the network jamming experience rating corresponding to each data stream, and the network quality experience label includes but is not limited to smooth, slight jamming, general jamming, very jamming, and jamming.

[0039] The network quality experience tag can determine the network jam rating based on the actual experience of the tester using the wireless network in a time period, and then add the same network quality experience tag to all data streams in the time period based on the network jam rating. For example, in the above-mentioned test of watching videos on Douyin, the tester takes 5 minutes as a time period. Assuming that the network jam rating determined by the tester based on the actual experience in a certain 5 minutes is smooth, the network quality experience tag added to each data stream in the 5-minute network traffic data is smooth.

[0040] When conducting actual experience, testers can use the following example criteria to determine the network lag rating when watching videos:

[0041] Of course, the network jamming experience rating can also be determined in other ways, for example, according to the determined average delay and / or first packet delay and / or total delay and / or application layer delay and / or network layer delay. Assuming that the network jamming experience rating is determined according to the average delay, the specific criteria can be expressed as follows:

[0042] Each processed data flow is taken as a data sample, which at least includes the number of successfully established flows, the number of packets flowing in, the number of packets flowing out, the total number of flows, the total number of downstream devices, the delay, the average delay, the first packet delay, the total delay, the application layer delay, the network layer delay, and the network quality experience label.

[0043] The processed multiple data samples are used as a data set, and the data samples in the data set are differentiated according to applications and application scenarios to facilitate the training of different network quality scoring models.

[0044] The constructed generative adversarial network is trained by the data set until the generative adversarial network meets the requirements. The generative adversarial network is a known technology, which includes a generator and a discriminator. The generator includes a multi-layer neural network, and the multi-layer neural network includes but is not limited to a fully connected layer or a convolutional neural network. The generator is used to generate data similar to the real network traffic data, and then the generated data is distinguished from the real data based on the discriminator. Through the alternating training of the generator and the discriminator, the network traffic data of the future period is finally generated for the subsequent network quality scoring.

[0045] During the training process, according to the accuracy, recall and other indicators returned by the training, the hyperparameters are adjusted to perform optimized training of the generative adversarial network until the accuracy, recall and other indicators meet the requirements.

[0046] The network quality scoring model corresponding to the air interface competition scenario is trained by using data samples obtained from network traffic data collected in the air interface competition scenario until the model converges.

[0047] The network quality scoring model corresponding to the video scene is trained by using data samples obtained from the network traffic data collected in the video scene until the model converges.

[0048] The network quality scoring model corresponding to the office scene is trained by using data samples obtained from the network traffic data collected in the office scene until the model converges.

[0049] The network quality scoring model corresponding to the game scenario is trained by using data samples obtained from the network traffic data collected in the game scenario until the model converges.

[0050] The network quality scoring model corresponding to the air interface competition scenario, the network quality scoring model corresponding to the video scenario, the network quality scoring model corresponding to the office scenario, and the network quality scoring model corresponding to the game scenario are respectively constructed by classification models in deep learning, for example, using known CNN networks (convolutional neural networks), DNN networks (deep neural networks), RNN networks (recurrent neural networks), GAN networks (generative adversarial networks) and other models.

[0051] For example, they use a generative adversarial network model for classification, which includes a generator and a discriminator. The reason for this design is that there is less network traffic data available in scenarios with poor network quality. Therefore, false data is constructed through the generator to alleviate the problem of uneven training category samples caused by data sparsity. The generator is preferably constructed using an RNN network. The discriminator is used for final classification. The discriminator is constructed using a CNN network, and the output layer is replaced by a softmax classifier to output the classification results and confidence (prob).

[0052] Furthermore, the network quality scoring models corresponding to the air interface competition scenario, video scenario, office scenario and game scenario all determine their corresponding network quality scores according to the classification results and confidence levels.

[0053] When the network quality experience label corresponding to the classification result is smooth, the network quality score is calculated according to the formula 100-10*(1-prob); When the network quality experience label corresponding to the classification result is slight lag, the network quality score is calculated according to the formula 90-10*(1-prob); When the network quality experience label corresponding to the classification result is general lag, the network quality score is calculated according to the formula 70+10*prob or 70-10*prob; for example, general lag can be further divided into above general lag and below general lag. When the confidence level corresponding to above general lag is higher, the formula 70+10*prob is selected to calculate the network quality score; otherwise, the formula 70-10*prob is selected.

[0054] When the network quality experience label corresponding to the classification result is very laggy, the network quality score is calculated according to the formula 50+10*prob or 50-10*prob; for example, very laggy can be further divided into very laggy or above and very laggy or below. When the confidence level corresponding to very laggy or above is higher, the formula 50+10*prob is selected to calculate the network quality score; otherwise, the formula 50-10*prob is selected.

[0055] When the network quality experience label corresponding to the classification result is stuck, the network quality score is calculated according to the formula 40-10*prob.

[0056] Furthermore, we also trained the network quality scoring model corresponding to each application, for example: In the video scenario, there are network quality scoring models corresponding to Douyin, Kuaishou, and iQiyi. Correspondingly, when training the network quality scoring model corresponding to Douyin, only the data samples obtained from the network traffic data collected when using Douyin are used for the training of the network quality scoring model corresponding to Douyin. When training the network quality scoring model corresponding to Kuaishou, only the data samples obtained from the network traffic data collected when using Kuaishou are used for the training of the network quality scoring model corresponding to Kuaishou. When training the network quality scoring model corresponding to iQiyi, only the data samples obtained from the network traffic data collected when using iQiyi are used for the training of the network quality scoring model corresponding to iQiyi.

[0057] In the game scenario, there are network quality scoring models corresponding to Honor of Kings, Peace for Gamers, and Genshin Impact, etc.; correspondingly, when training the network quality scoring model corresponding to Honor of Kings, only data samples obtained from network traffic data collected when using Honor of Kings are used for the training of the network quality scoring model corresponding to Honor of Kings; when training the network quality scoring model corresponding to Peace for Gamers, only data samples obtained from network traffic data collected when using Peace for Gamers are used for the training of the network quality scoring model corresponding to Peace for Gamers; when training the network quality scoring model corresponding to Genshin Impact, only data samples obtained from network traffic data collected when using Genshin Impact are used for the training of the network quality scoring model corresponding to Genshin Impact.

[0058] In the office scenario, the network quality scoring model corresponding to Tencent Conference, the network quality scoring model corresponding to DingTalk, the network quality scoring model corresponding to Changlian, etc.; correspondingly, when training the network quality scoring model corresponding to Tencent Conference, only the data samples obtained from the network traffic data collected when using Tencent Conference are used for the training of the network quality scoring model corresponding to Tencent Conference; when training the network quality scoring model corresponding to DingTalk, only the data samples obtained from the network traffic data collected when using DingTalk are used for the training of the network quality scoring model corresponding to DingTalk; when training the network quality scoring model corresponding to Changlian, only the data samples obtained from the network traffic data collected when using Changlian are used for the training of the network quality scoring model corresponding to Changlian, etc.

[0059] The network quality scoring model corresponding to each application reflects the network quality when using each application. The network quality scoring model corresponding to each application can also be constructed using the classification model of the above example. It also determines the network quality score corresponding to each application based on the classification result and confidence level, which will not be elaborated here.

[0060] Furthermore, a family overall network quality scoring model is trained, and the output result of the family overall network quality scoring model reflects the overall network quality of the family. It can also be constructed using the classification model of the above example, and it also determines the network quality score corresponding to each application based on the classification result and confidence level, which will not be described here. When training the family overall network quality scoring model, it is trained through all data samples in the data set.

[0061] When determining which application scenario to use, the network quality scoring model corresponding to the application scenario is first determined based on the total number of devices hanging down in the data to be processed. If the total number of devices hanging down is greater than 2, it can be determined that it is an air interface competition scenario. On the contrary, if the total number of devices hanging down is 1, it is determined that it is a non-air interface competition scenario. At this time, it is necessary to further determine the application corresponding to the data stream by combining the destination IP address of each stream in the data stream with the existing application identification library, and then the corresponding application scenario can be determined according to the identified application. The application identification library will store the destination IP address corresponding to each application name. When multiple applications are identified in a data stream, the application with the most occurrences is used as the application corresponding to the data stream, and the application scenario corresponding to the identified application is used as the application scenario corresponding to the data stream. For example, if the application corresponding to the destination IP address with the most occurrences in a data stream is vibrato, then the application scenario is determined to be a video scene, and the predicted data can be subsequently input into the network quality scoring models corresponding to the video scene and vibrato to determine the network quality score.

[0062] Correspondingly, as attached Figure 1 As shown, the home network monitoring method includes the following steps: S10, obtaining network traffic data of a home gateway running for a period of time; S20, processing the network traffic data for the period of time to obtain data to be evaluated; S30, determining an application scenario corresponding to the data to be evaluated according to the data to be evaluated; S40, when it is determined that the application scenario is an air interface competition scenario, inputting the data to be evaluated into a generative adversarial network to generate prediction data, and inputting the prediction data into a network quality scoring model corresponding to the air interface competition scenario to determine a network quality score in the air interface competition scenario; When it is determined that the application scenario is a video scenario, a game scenario, or an office scenario, the data to be evaluated is input into a generative adversarial network to generate prediction data, and the prediction data is input into a network quality scoring model corresponding to the video scenario, the game scenario, or the office scenario to determine the network quality score in the video scenario, the game scenario, or the office scenario; S50, when it is determined that the network quality score in the air interface competition scenario, video scenario, game scenario, or office scenario is lower than the set value, the network quality prediction result in the air interface competition scenario, video scenario, game scenario, or office scenario is fed back to the terminal device under the home gateway.

[0063] In S10, the network traffic data is also captured by using a packet capture tool set in the home gateway. The period of time is the same as the granularity used to store each data stream, such as 1 minute, 2 minutes, or 5 minutes.

[0064] The specific process of obtaining the data to be evaluated in S20 can refer to the process of obtaining data samples in model training, which will not be described in detail here.

[0065] In the S30, it is also first determined whether it is an air interface contention scenario according to the total number of downstream devices. When determining a non-air interface contention scenario, the specific application scenario is determined according to the application identified by the destination IP address.

[0066] In S50, the set value can be set as needed, for example, 65 points. When the network quality score determined by the network quality scoring model corresponding to each application scenario is lower than 65 points, which will significantly affect the user experience, the corresponding network quality prediction result will be sent to the user via an APP message. The network quality prediction result may include the network quality score and network quality experience label in a certain application scenario.

[0067] Furthermore, when a specific application is identified in S30, in S40, the prediction data is also input into a network quality scoring model corresponding to the identified application to determine a network quality score when the application is used.

[0068] In S50, when it is determined that the network quality score when using the application is lower than the set value, the network quality prediction result when using the application is sent to the terminal device under the home gateway. The network quality prediction result may also include the network quality score and network quality experience label when using the application.

[0069] Furthermore, in S40, after the prediction data is generated, the prediction data is input into a home overall network quality scoring model to determine the home overall network quality score.

[0070] Correspondingly, in S50, when it is determined that the overall home network quality score is lower than the set value, the overall home network quality prediction result is sent to the terminal device connected to the home gateway.

[0071] Further, in S50, when the APP message is fed back to the terminal device under the home gateway, information on whether the home gateway can be restarted can also be sent to the terminal device under the home gateway. When the user determines that the home gateway can be restarted, the home gateway is remotely controlled to restart. Conversely, when any user determines that the home gateway cannot be restarted, the home gateway is not restarted. The technology of remotely controlling the restart of the home gateway is a known technology and is not an innovation of the present invention, and will not be described in detail here.

[0072] As attached Figure 2 As shown, in S50, the activity status of the terminal device within the period of time can be determined based on the network traffic data of the period of time. When it is determined that the activity status of the terminal device is inactive, it is determined whether the information on whether the home gateway can be restarted is sent to the terminal device under the home gateway; when it is determined that the activity status of the terminal device is active, it is determined not to send the information on whether the home gateway can be restarted to the terminal device under the home gateway.

[0073] In specific implementation, the activity status of the terminal devices under the home gateway can be determined based on the network traffic data and activity classification model for the period of time.

[0074] In detail, the activity situation classification model may adopt a known binary classification model. For example, the activity situation classification model is constructed by adopting a K-nearest neighbor classification algorithm, a support vector machine, a decision tree, a random forest, a Bayesian classifier, and the like.

[0075] When training the activity classification model, the network traffic data of the home gateway for 7 days can be obtained and stored separately at the granularity determined above. The network traffic data of each granularity includes a property table of the device connected to the gateway, including but not limited to: data acquisition time, gateway number, gateway address, connected device number, device type, device uplink rate, device downlink rate, device uplink traffic, and device downlink traffic.

[0076] After the network traffic data stored at each granularity is normalized by the Z-score standardization method, data aggregation is performed. The data aggregation can be summing, averaging, maximum, minimum, median, quantile, variance, etc. Specifically, data aggregation is performed on the device uplink rate, device downlink rate, device uplink traffic, and device downlink traffic. The data obtained by data aggregation includes at least the average uplink rate, average downlink rate, total uplink traffic, total downlink traffic, and the total number of connected devices. The data obtained after data aggregation of the network traffic data at each granularity is used as a training data set. All training data sets for 7 days are split, 70% of which are used as training sets and 30% as test sets.

[0077] The constructed activity classification model is trained with the training set and tested with the test set. When selecting model parameters, it is evaluated based on precision, recall, and F1-score. Specifically, the model parameters are determined based on the size of the multiplication result of precision, recall, and F1-score, and the model parameter with the largest result is selected as the optimal parameter; if the multiplication results are the same, the model parameter with the highest precision is taken as the optimal parameter.

[0078] After the activity classification model corresponding to the home gateway is trained, the relevant data in the attribute table of the gateway-attached device in the network traffic data for a period of time in S10 can be normalized and aggregated, and then input into the activity classification model corresponding to the home gateway. The output result of the activity classification model is that the terminal device is active or the terminal device is inactive. The active terminal device means that the user is currently using the wireless network a lot, and it is not suitable to restart the home network; the inactive terminal device means that the user is not using the home network or only needs a small amount of network traffic, and the home gateway can be restarted.

[0079] Furthermore, in order to ensure the long-term reliable use of the home network, when obtaining network traffic data for a period of time, as shown in the attached Figure 3 As shown, the busyness of the home gateway and / or the activity of the terminal device are determined according to the network traffic data for the period of time, and whether the home gateway can be controlled to restart is determined accordingly. When it is determined that the home gateway device is not busy and / or the terminal device is not active, it is determined that the home gateway can be restarted.

[0080] The specific method for determining the activity status of the terminal device is described above and will not be repeated here.

[0081] At the same time, the network traffic data acquired for a period of time includes a gateway attribute table, which includes: data acquisition time, gateway number, gateway address, CPU usage, memory usage, gateway uplink rate, gateway downlink rate, gateway uplink traffic, gateway downlink traffic, and gateway startup time. Among them, the CPU usage, memory usage, gateway uplink rate, gateway downlink rate, gateway uplink traffic, gateway downlink traffic, and gateway startup time can be used to determine the busyness of the home gateway.

[0082] Specifically, the busy status of the home gateway is determined according to the following process: First, the gateway uplink rate, gateway downlink rate, gateway uplink traffic, and gateway downlink traffic in the gateway's own attribute table for a period of time are normalized respectively.

[0083] Next, the CPU usage rate, memory usage rate, gateway startup time and the normalized gateway uplink rate, gateway downlink rate, gateway uplink traffic, and gateway downlink traffic are multiplied by their corresponding weights and the average is summed to obtain the gateway status score, and the final score is between 0-100. Among them, the weights corresponding to the CPU usage rate, memory usage rate, gateway uplink rate, gateway downlink rate, gateway uplink traffic, gateway downlink traffic, and gateway startup time are calculated by the TOPSIS algorithm. Specifically, the network traffic data of the home gateway for 7 days is obtained, and then the data of the gateway's own attribute table in the obtained network traffic data is calculated by the TOPSIS algorithm to obtain the weights corresponding to the CPU usage rate, memory usage rate, gateway uplink rate, gateway downlink rate, gateway uplink traffic, gateway downlink traffic, and gateway startup time, and stored in the cloud storage. When evaluating later, it can be directly called.

[0084] Finally, the busy state of the home gateway is determined to be busy or not busy according to the gateway status score and the score threshold. Specifically, when the gateway status score is greater than the score threshold, the busy state of the home gateway is determined to be busy, otherwise the busy state of the home gateway is determined to be not busy. The score threshold can be designed as needed. According to the weight definition, the score threshold is preferably 50, that is, when the gateway status score is less than or equal to 50, the busy state of the home gateway is determined to be not busy, and when the gateway status score is greater than 50, the busy state of the home gateway is determined to be busy.

[0085] In order to avoid an erroneous restart caused by a misjudgment, whether to restart the home gateway is determined based on the busy status of the home gateway and the activity status of the terminal device. That is, when it is determined that the activity status of the terminal device is not active and the busy status of the home gateway is not busy, it is determined that the home gateway can be restarted.

[0086] In order to further ensure the accuracy of the judgment, at least two judgments can be made. If the result of each judgment is that the activity status of the terminal device is inactive and the busy status of the home gateway is not busy, it is determined that the home gateway can be restarted and the home gateway restart is controlled.

[0087] Example 2 This embodiment discloses a home network monitoring system, including: A data acquisition unit, used to acquire network traffic data of a home gateway running for a period of time; A data processing unit, used for processing the network traffic data for a period of time to obtain data to be evaluated; An application scenario identification unit, used to determine the application scenario corresponding to the data to be evaluated according to the data to be evaluated; A score prediction unit, configured to, when determining that the application scenario is an air interface competition scenario, input the data to be evaluated into a generative adversarial network to generate prediction data, and input the prediction data into a network quality scoring model corresponding to the air interface competition scenario to determine a network quality score in the air interface competition scenario; When it is determined that the application scenario is a video scenario, a game scenario, or an office scenario, inputting the data to be evaluated into a generative adversarial network to generate prediction data, and inputting the prediction data into a network quality scoring model corresponding to the video scenario, the game scenario, or the office scenario to determine the network quality score in the video scenario, the game scenario, or the office scenario; A feedback unit is used to feedback the network quality prediction result in the air interface competition scenario, video scenario, game scenario or office scenario to the terminal device under the home gateway when it is determined that the network quality score in the air interface competition scenario, video scenario, game scenario or office scenario is lower than a set value.

[0088] Example 3 This embodiment discloses a home network monitoring device, including a memory and a processor, wherein the memory stores a program executable by the processor, and when the program is executed, the home network monitoring method described above is implemented.

[0089] There are many implementation methods of the present invention, and all technical solutions formed by equivalent transformation or equivalent transformation fall within the protection scope of the present invention.

Claims

1. A home network monitoring method, characterized in that: The steps include: S10, obtaining network traffic data of a home gateway running for a period of time; S20, processing the network traffic data for the period of time to obtain data to be evaluated; S30, determining an application scenario corresponding to the data to be evaluated according to the data to be evaluated; S40, when it is determined that the application scenario is an air interface competition scenario, inputting the data to be evaluated into a generative adversarial network to generate prediction data, and inputting the prediction data into a network quality scoring model corresponding to the air interface competition scenario to determine a network quality score in the air interface competition scenario; When it is determined that the application scenario is a video scenario, a game scenario, or an office scenario, the data to be evaluated is input into a generative adversarial network to generate prediction data, and the prediction data is input into a network quality scoring model corresponding to the video scenario, the game scenario, or the office scenario to determine the network quality score in the video scenario, the game scenario, or the office scenario; S50, when it is determined that the network quality score in the air interface competition scenario, video scenario, game scenario, or office scenario is lower than the set value, the network quality prediction result in the air interface competition scenario, video scenario, game scenario, or office scenario is fed back to the terminal device under the home gateway.

2. The home network monitoring method according to claim 1, characterized in that: In S30, whether it is an air interface contention scenario is determined according to the total number of downstream devices.

3. The home network monitoring method according to claim 1, characterized in that: In the S30, determining whether the application scenario is a video scenario, a game scenario, or an office scenario according to the application identified by the data to be evaluated; In the S40, the predicted data is also input into a network quality scoring model corresponding to the identified application to determine a network quality score when the application is used; In the S50, when it is determined that the network quality score when using the application is lower than the set value, the network quality prediction result when using the application is sent to the terminal device connected to the home gateway.

4. The home network monitoring method according to claim 1, characterized in that: In said S40, after generating the prediction data, the prediction data is input into the home overall network quality scoring model to determine the home overall network quality score; In the S50, when it is determined that the overall home network quality score is lower than the set value, the overall home network quality prediction result is sent to the terminal device connected to the home gateway.

5. The home network monitoring method according to claim 1, characterized in that: In S50, information on whether the home gateway can be restarted is sent to the terminal device connected to the home gateway.

6. The home network monitoring method according to claim 5, characterized in that: Determine the activity of the terminal device within the period of time according to the network traffic data of the period of time, and when it is determined that the activity of the terminal device is inactive, determine whether to send information on whether to restart the home gateway to the terminal device under the home gateway; When it is determined that the activity status of the terminal device is active, it is determined not to send information on whether the home gateway can be restarted to the terminal device under the home gateway.

7. The home network monitoring method according to any one of claims 1 to 6, characterized in that: When the network traffic data for a period of time is obtained, the busyness of the home gateway and / or the activity of the terminal device are determined according to the network traffic data for the period of time, and whether the home gateway can be controlled to restart is determined accordingly.

8. The home network monitoring method according to claim 7, characterized in that: When it is determined that the home gateway device is not busy and / or the terminal device is not active, it is determined that the home gateway can be controlled to restart.

9. Home network monitoring system, characterized in that, include: A data acquisition unit, used to acquire network traffic data of a home gateway running for a period of time; A data processing unit, used for processing the network traffic data for a period of time to obtain data to be evaluated; An application scenario identification unit, used to determine the application scenario corresponding to the data to be evaluated according to the data to be evaluated; A score prediction unit, for, when it is determined that the application scenario is an air interface competition scenario, inputting the data to be evaluated into a generative adversarial network to generate prediction data, and inputting the prediction data into a network quality scoring model corresponding to the air interface competition scenario to determine the network quality score in the air interface competition scenario; when it is determined that the application scenario is a video scene, a game scene, or an office scene, inputting the data to be evaluated into a generative adversarial network to generate prediction data, and inputting the prediction data into a network quality scoring model corresponding to the video scene, the game scene, or the office scene to determine the network quality score in the video scene, the game scene, or the office scene; A feedback unit is used to feedback the network quality prediction result in the air interface competition scenario, video scenario, game scenario or office scenario to the terminal device under the home gateway when it is determined that the network quality score in the air interface competition scenario, video scenario, game scenario or office scenario is lower than a set value.

10. A home network monitoring device, comprising a memory and a processor, wherein the memory stores a program executable by the processor, wherein: When the program is executed, the home network monitoring method as described in any one of claims 1-8 is implemented.

Images