A broadband service quality identification method, device, equipment and medium

Abstract

The application provides a broadband service quality difference identification method, device, equipment and medium, and relates to the technical field of communication. The method comprises the following steps: obtaining user usage data generated by a user using a broadband network; inputting the user usage data into a trained quality difference identification model to obtain a quality difference identification result output by the quality difference identification model; the quality difference identification result is used to indicate a user with a quality difference problem; and the quality difference identification model is trained based on historical quality difference data of a body sense generated by the broadband network after a user uses the broadband network with added preset network loss. The application realizes active evaluation of the broadband service quality level, so as to accurately label each potential quality difference user, and then find problems in advance and timely repair the broadband, thereby improving user experience and stickiness.

Description

Technical Field

[0001] This invention relates to the field of communication technology, and in particular to a method, apparatus, device, and medium for identifying poor broadband service quality. Background Technology

[0002] As the number of broadband users increases, users' expectations for network quality are also constantly rising. Researching broadband quality differences is crucial for improving both network quality and user experience. Improving user satisfaction with broadband has become an urgent need for operators. However, broadband service processes are complex and involve users' subjective feelings. Issues such as slow internet speeds and video buffering can affect users' subjective experiences, and the user's broadband experience is key to their satisfaction.

[0003] There is often a discrepancy between network quality and actual user experience. Data generated during actual internet use cannot clearly identify which aspects exhibit poor network quality. For example, even with good broadband network metrics, users may experience poor performance and file complaints for services requiring low latency and low packet loss. Conversely, poor broadband network metrics may not elicit complaints due to a lack of activity. Therefore, accurately identifying users with poor network quality is crucial for improving broadband satisfaction. Identifying these users cannot rely solely on network metric thresholds; a user experience-centric approach is essential for measurable quality. However, existing methods for identifying poor network quality fail to accurately assess user perception, hindering effective evaluation of user experience and precise identification of users with poor perceived quality. Consequently, they cannot quickly pinpoint and address user complaints, resulting in operators lacking a rapid, closed-loop capability for improving user experience.

[0004] It is evident that objectively assessing user perception of broadband services is crucial for refined network quality management and proactive problem detection. Therefore, it is necessary to provide a broadband service quality poorness identification method that can accurately assess user perception and identify users with truly poor perception. Summary of the Invention

[0005] In view of this, the present invention proposes a method, apparatus, device and medium for identifying poor broadband service quality, in order to solve the problem of being unable to accurately assess users' active perception and identify users with truly poor perception.

[0006] The technical solution of this invention is implemented as follows:

[0007] According to a first aspect, embodiments of the present invention provide a method for identifying poor broadband service quality, the method comprising:

[0008] Acquire user usage data generated by users using broadband networks;

[0009] The user data is input into the trained poor quality identification model to obtain the poor quality identification result output by the poor quality identification model; the poor quality identification result is used to indicate users with poor quality problems; the poor quality identification model is trained based on historical poor quality data generated after users use broadband networks with preset network loss added to network parameters.

[0010] The poor quality identification model is trained through the following steps:

[0011] Add preset network loss to network parameters in broadband networks that exceed the level of abnormal impact;

[0012] Acquire historical data on poor network quality generated after users use broadband networks with preset network loss parameters;

[0013] Data cleaning was performed on historical poor-quality data to obtain historical network anomaly data;

[0014] The historical network anomaly data is used as input data for training. By adding preset noise to the historical network anomaly data and using unsupervised learning, the goal is to control the similarity between the input data and the output data within a preset similarity range, thus obtaining a noise reduction coding layer for output decoding data.

[0015] The historical decoding data output by the noise reduction coding layer is used as the input data for training, and the training is carried out in an unsupervised manner to obtain an isolated forest layer for outputting poor quality recognition results.

[0016] In conjunction with the first aspect, in the first embodiment of the first aspect, adding a preset network loss to network parameters in the broadband network that exceed the level of abnormal impact specifically includes:

[0017] Acquire key performance indicator data of broadband network users and extract network anomaly data from the key performance indicator data of broadband network users.

[0018] The system obtains the degree of impact of abnormal network data, arranges and combines the network parameters corresponding to abnormal network data that exceed the degree of impact, and adds preset network loss to the combined network parameters.

[0019] In conjunction with the first aspect, in the second embodiment of the first aspect, the step of cleaning historical poor user experience data to obtain historical network anomaly data specifically includes:

[0020] Fill in the missing values ​​in the historical data of poor sensory quality;

[0021] Historical network anomaly data is extracted from historical perceptual quality differences after null filling.

[0022] In conjunction with the first aspect, in the third embodiment of the first aspect, the noise reduction coding layer specifically includes:

[0023] The encoder input layer, noise addition layer, encoder hidden layer, and encoder output layer are connected in sequence.

[0024] The encoder input layer is used to receive and transmit historical network anomaly data downwards; the noise addition layer is used to add preset noise to the historical network anomaly data to obtain damaged data; the encoder hidden layer is used to encode the damaged data using the encoding function g(·) to obtain historical encoded data, and to decode the historical encoded data using the decoding function f(·) to obtain historical decoded data; the encoder output layer is used to receive and transmit the historical decoded data to the isolated forest layer.

[0025] In conjunction with the third embodiment of the first aspect, in the fourth embodiment of the first aspect, the loss function of the noise reduction coding layer is:

[0026]

[0027] Where L(·) represents the loss function; This represents the damaged data of the i-th feature after adding preset noise to historical network anomaly data; x i This represents the historical network anomaly data for the i-th feature; n represents the total number of features i; λ(ε) i ≠0) represents a penalty term. When the damaged data of the i-th feature has no added noise, a penalty term of size λ is added.

[0028] In conjunction with the third embodiment of the first aspect, in the fifth embodiment of the first aspect, the step of inputting the user usage data into the trained quality defect identification model to obtain the quality defect identification result output by the quality defect identification model specifically includes:

[0029] The user data is input into the noise reduction coding layer, which adds preset noise to the user data and then performs encoding and decoding processing to obtain decoded data.

[0030] The decoded data is input into the isolated forest layer to obtain the quality defect identification result output by the isolated forest layer.

[0031] In conjunction with the first aspect, in the sixth embodiment of the first aspect, the step of using the historical decoding data as input data for training and training it in an unsupervised manner to obtain an isolated forest layer for outputting poor quality recognition results specifically includes:

[0032] Filter a preset number of the historical decoded data and place them into the root node of the tree;

[0033] A cut point is randomly generated in the root node; the cut point is located between the maximum and minimum values ​​of a specified dimension in the root node data;

[0034] The root node is divided according to the cut point. Sample data with dimensions lower than the specified dimension are placed in one subspace of the root node, and sample data with dimensions not lower than the specified dimension are placed in the other subspace of the root node.

[0035] The training of the isolated forest layer is completed when each root node is continuously divided into subspaces until there is only one piece of the historical decoded data in the subspace of the root node or the subspace has reached the preset height.

[0036] According to a second aspect, embodiments of the present invention also provide a broadband service quality poor identification device, the device comprising:

[0037] The data acquisition module is used to acquire user usage data generated by users using the broadband network;

[0038] The poor quality identification module is used to input the user's usage data into the trained poor quality identification model to obtain the poor quality identification result output by the poor quality identification model; the poor quality identification result is used to indicate users with poor quality problems; the poor quality identification model is trained based on historical user experience poor quality data generated after the user used a broadband network with preset network loss added to the network parameters.

[0039] The poor quality identification model is trained through the following steps:

[0040] Add preset network loss to network parameters in broadband networks that exceed the level of abnormal impact;

[0041] Acquire historical data on poor network quality generated after users use broadband networks with preset network loss parameters;

[0042] Data cleaning was performed on historical poor-quality data to obtain historical network anomaly data;

[0043] The historical network anomaly data is used as input data for training. By adding preset noise to the historical network anomaly data and using unsupervised learning, the goal is to control the similarity between the input data and the output data within a preset similarity range, thus obtaining a noise reduction coding layer for output decoding data.

[0044] The historical decoding data output by the noise reduction coding layer is used as the input data for training, and the training is carried out in an unsupervised manner to obtain an isolated forest layer for outputting poor quality recognition results.

[0045] According to a third aspect, embodiments of the present invention also provide 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 program to implement the steps of any of the broadband service quality poor identification methods described above.

[0046] According to a fourth aspect, embodiments of the present invention also provide a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the broadband service quality poor identification method as described above.

[0047] According to a fifth aspect, embodiments of the present invention also provide a computer program product, including a computer program that, when executed by a processor, implements the steps of the broadband service quality poor identification method as described in any of the preceding claims.

[0048] The broadband service quality identification method, apparatus, device, and medium of the present invention have the following advantages over the prior art:

[0049] The poor quality identification model is trained to identify quality issues and obtain the identification results. This model adds a denoising coding layer in the form of an autoencoder before the isolated forest layer, solving the problem that the traditional isolated forest algorithm can only perform linear segmentation, while retaining the advantage that the isolated forest algorithm does not need to calculate distance and density. This significantly improves speed and reduces system overhead. The denoising coding layer reduces the impact of reported erroneous data on the subsequent identification accuracy, improving noise resistance. Furthermore, the training data is obtained by adding network losses to the PON network, thus constructing historical poor quality data that reflects the user's actual experience. This historical poor quality data contains the user's real experience information, thus filling the gap of unlabeled data in broadband user poor quality identification scenarios. Therefore, it improves the identification accuracy and robustness of the poor quality identification model, enabling proactive assessment of broadband service quality levels. This allows for accurate labeling of each potential poor quality user, enabling early detection of problems and timely broadband maintenance, improving user experience and user stickiness. Attached Figure Description

[0050] To more clearly illustrate the technical solutions in the embodiments of the present invention 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 only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0051] Figure 1 This is a flowchart illustrating the broadband service quality poor identification method of the present invention;

[0052] Figure 2This is a schematic diagram of the noise reduction coding layer in the broadband service quality poor identification method of the present invention;

[0053] Figure 3 This is a schematic diagram of the process of training the poor quality identification model in the broadband service poor quality identification method of the present invention.

[0054] Figure 4 This is a schematic diagram of the poor quality identification model performing service processing in the broadband service poor quality identification method of the present invention.

[0055] Figure 5 This is a schematic diagram of the broadband service quality identification device of the present invention;

[0056] Figure 6 This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0057] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0058] With the increasing number of broadband users, users' expectations for network quality are also constantly rising. Researching broadband quality differences is crucial for improving both network quality and user experience. Improving user satisfaction with broadband has become an urgent need for operators. However, broadband service processes are complex and involve users' subjective feelings. Issues such as slow internet speeds and video buffering can affect users' subjective experiences, and the user's broadband experience is key to satisfaction. For example, the network link for home broadband services typically consists of terminal equipment, routers, switches, optical network units (ONUs), optical distribution networks (ODNs), passive optical networks (PONs), optical line terminals (OLTs), bearer networks, and content source networks. User satisfaction is influenced by the usage of these user-owned terminals, as well as users' subjective behaviors and habits. There is often a difference between network quality and actual user service perception; data generated by users during actual internet access cannot clearly identify which data points indicate poor perceived quality. For example, even if broadband network indicators are good, users may experience poor performance and file complaints for services requiring low latency and low packet loss. Conversely, if broadband network indicators are poor, users may not file complaints due to a lack of service. Therefore, accurately identifying users with poor network quality issues is a prerequisite for improving broadband experience satisfaction. Identifying users with poor quality issues cannot simply rely on network indicator thresholds; it should be centered on user experience, thereby achieving measurable user experience. However, existing methods for identifying poor quality cannot accurately assess users' proactive perceptions, and suffer from problems such as the inability to effectively evaluate the quality of user experience and the inability to accurately identify users with perceived poor quality. Consequently, they cannot quickly define and locate the problems reported by users, resulting in operators lacking a rapid closed-loop capability for improving user experience.

[0059] Existing methods for identifying poor network quality cannot accurately assess users' proactive perceptions and identify users with truly poor perceptions. In conclusion, objectively assessing users' perceptions of broadband services is crucial for refined network quality management and proactive problem detection.

[0060] The broadband service quality identification method provided in this specification can be applied to electronic devices with broadband service communication data processing capabilities. It aims to more accurately identify broadband service quality issues and improve user experience. These electronic devices may include laptops, desktop computers, smartphones, smart wearable devices (virtual reality glasses, smartwatches, etc.), tablets, etc. Of course, the broadband service quality identification method provided in this specification can also be applied to applications running on the aforementioned electronic devices. For example, this broadband service quality identification method can be applied to browsers with broadband service communication data processing capabilities, or to software with broadband service communication data processing capabilities.

[0061] Please see Figure 1 , Figure 1 A flowchart of a broadband service quality poor identification method according to an embodiment of the present invention is shown. The method may include the following steps:

[0062] S101. Obtain user usage data generated by the user's use of the broadband network.

[0063] In this embodiment, user data can be pre-stored in the electronic device or obtained by the electronic device from an external source. For example, the electronic device can obtain the data from an external broadband service acquisition device. There are no restrictions on the specific form in which the user data is obtained, as long as the electronic device can acquire the user data.

[0064] S102. Input the user's usage data into the trained poor quality identification model to obtain the poor quality identification result output by the poor quality identification model. The poor quality identification result is used to indicate users with poor quality problems. The poor quality identification model is trained based on historical poor quality data generated after the user used a broadband network with preset network loss added to the network parameters.

[0065] By adding network losses to the PON network, historical data reflecting the user's actual experience of poor quality can be constructed. This historical data contains the user's actual experience information, thus filling the gap of unlabeled data in the broadband user quality identification scenario.

[0066] In this embodiment, the poor quality identification model is trained in the following way:

[0067] S201. Add preset network loss to network parameters in the broadband network that exceed the abnormal impact level.

[0068] Specifically, step S201 includes:

[0069] S2021. Obtain key performance indicator (KPI) data of broadband network users and extract network anomaly data from user KPI data.

[0070] Similarly, in this embodiment, the user KPI data can be pre-stored in the electronic device or obtained by the electronic device from an external source. For example, the electronic device can obtain the data from an external broadband service acquisition device. There are no restrictions on the specific form of user KPI data acquisition, as long as the electronic device can obtain the user KPI data.

[0071] Since the user KPI data contains a large amount of data, it is also necessary to extract network anomaly data that is highly correlated with network anomalies. In this embodiment, the network anomaly data includes eight types of data: user-side average round-trip time (networkSideAverageRtt), network-side average round-trip time jitter (networkSideRttAverageJitter), user-side average round-trip time (userSideAverageRtt), user-side average round-trip time jitter (userSideRttAverageJitter), downlink network-side packet loss rate (flowDownLinkNetworkSideDiscardRatio), uplink network-side packet loss rate (flowUplinkNetworkSideDiscardRatio), downlink user-side packet loss rate (flowDownLinkUserSideDiscardRatio), and uplink user-side packet loss rate (flowUplinkUserSideDiscardRatio).

[0072] Preferably, the user KPI data comes from the reports of the OLT board. The user_kpi data reported by the OLT board has a lot of fields, so it is also necessary to extract the fields that are highly correlated with network anomalies.

[0073] Fields highly correlated with network anomalies include user-side average round-trip time (networkSideAverageRtt), network-side average round-trip time jitter (networkSideRttAverageJitter), user-side average round-trip time (userSideAverageRtt), user-side average round-trip time jitter (userSideRttAverageJitter), total downlink packets (flowDownLinkTotalPackets), downlink network-side packet loss (flowDownLinkNetworkSideDiscardPackets), downlink user-side packet loss (flowDownLinkUserSideDiscardPackets), total uplink packets (flowUpLinkTotalPackets), uplink network-side packet loss (uplink network-side packet loss), and uplink user-side packet loss (uplink user-side packet loss).

[0074] Then, based on the packet loss count and total packet count data, obtain the required packet loss rate data, where:

[0075]

[0076] In formula (1), flowDownLinkNetworkSideDiscardRatio represents the downlink network-side packet loss rate.

[0077]

[0078] In formula (2), flowDownLinkUserSideDiscardRatio represents the downlink user-side packet loss rate.

[0079]

[0080] In formula (3), flowUpLinkNetworkSideDiscardRatio represents the uplink network side packet loss rate.

[0081]

[0082] In formula (4), flowUpLinkUserSideDiscardRatio represents the uplink user-side packet loss rate.

[0083] S2022. Obtain the degree of abnormal impact of network abnormal data, arrange and combine the network parameters corresponding to the network abnormal data that exceeds the degree of abnormal impact, and add preset network loss to the combined network parameters.

[0084] Among the network anomaly data, average jitter and uplink packet loss rate have a relatively small impact on perceived quality, while the other four data points have a significant impact. Further permutations and combinations of the network parameters (including round-trip time and packet loss rate) that generate these four data points yield 15 possible combinations. For each combination, multiple gradient values ​​of network loss parameters are set. For example, the round-trip time is increased from 100ms to 300ms using a network impairment meter, with increments of 50ms or 100ms; the packet loss rate is increased from 10% to 25% using a network impairment meter, with increments of 5% or 10%.

[0085] By arranging and combining network anomaly data exceeding a preset anomaly impact level and adding preset network loss parameters, a specific network loss parameter can be added to the PON network. When a router is connected to the ONU, the user terminal can experience the network status under the preset network loss parameter after connecting to broadband.

[0086] S202. Obtain historical user experience quality degradation data generated after users use broadband networks with preset network loss parameters. This historical user experience quality degradation data is also part of the obtained user KPI data.

[0087] Users generate various types of user data when using various applications on a broadband network with added network loss. By reading the user data of the applications from systems such as Kafka and recording the time periods when poor user experience occurred, the data is marked as 1, and the data during the time periods when poor user experience did not occur is marked as 0. This way, historical poor user experience data is obtained.

[0088] S203. Perform data cleaning on historical poor-quality data to obtain historical network anomaly data.

[0089] In this embodiment, data cleaning includes null value imputation and feature filtering. Specifically, step S203 includes:

[0090] S2031. Fill in the missing values ​​in the historical poor sensory quality data. For any missing values ​​that may exist in the historical poor sensory quality data, fill in the missing values ​​with 0.

[0091] S2032. Extract historical network anomaly data from the historical perceptual quality difference after null filling.

[0092] In this embodiment, when obtaining the packet loss rate of abnormal network data as in step S202, the upper limit of the packet loss rate is set to 1 to prevent dirty data from causing the packet loss rate to exceed 1. For other extraction processes, please refer to step S2021.

[0093] S204. Using historical network anomaly data as input data for training, and by adding preset noise to the historical network anomaly data and using unsupervised learning to train, with the goal of controlling the similarity between input data and output data within a preset similarity level, a denoising coding layer for output decoding data is obtained.

[0094] The existing isolated forest technology has two problems: firstly, it can only segment outlier data on a single feature plane each time, which leads to bias in areas not covered by the data; secondly, when processing high-dimensional data, the distance between data points becomes blurred, making it difficult to accurately detect outliers.

[0095] An autoencoder is a deep learning model used for unsupervised learning. Its goal is to compress input data into a low-dimensional representation and then reconstruct the original data from this low-dimensional representation. In this embodiment, the denoising autoencoder is based on an autoencoder, but noise is added to the input data of the input layer to prevent overfitting, thus making the trained denoising encoder more robust.

[0096] Because the collected historical sensory data may contain a few errors, the source data is mixed with noisy data. A noise factor is added to the loss function of the denoising autoencoder. In addition to ensuring that the reconstructed data is close to the original data, the denoising autoencoder will also give an extra reward when reconstructing noisy original data, thereby improving the model's denoising ability.

[0097] The poor quality identification model adds an improved denoising encoder to the Isolation Forest algorithm to learn non-linear data relationships and compress high-dimensional data. The addition of the denoising encoder effectively solves the problems of existing Isolation Forest algorithms, such as only segmenting outliers on a single feature plane and the blurring of distances between data points.

[0098] Please see Figure 2 The network structure of the noise reduction autoencoder includes an encoder input layer, a noise addition layer, an encoder hiding layer, and an encoder output layer.

[0099] The encoder input layer is used to input historical network anomaly data, and the noise addition layer is used to add preset noise to the historical network anomaly data. Specifically:

[0100]

[0101] In formula (5), This represents the damaged data of the i-th feature with added preset noise; x irepresents the original historical network anomaly data of the i-th feature; p represents the preset probability, that is, the noise addition layer has a probability of p to set the historical network anomaly data to 0, and a probability of 1-p to keep the original value unchanged.

[0102] The encoder hidden layer is used to encode the damaged data using the encoding function g(·) to obtain historical encoded data, and then to decode the historical encoded data using the decoding function f(·) to obtain historical decoded data. Specifically, the encoding process is as follows:

[0103]

[0104] In formula (6), h represents damaged data; h represents data that has been damaged. Historical encoded data (features) for encoding processing; u represents the activation function; w1 represents the first weight, which is the weight from the encoder input layer to the encoder hidden layer; b1 represents the first bias term, which is the bias term of the encoder hidden layer; the activation function u is specifically:

[0105] u(z)=(1+e -z ) -1 (7)

[0106] The decoding process is as follows:

[0107] y=f(h)=u(w2h+b2) (8)

[0108] In formula (8), y represents the reconstructed historical decoding data (features); w2 represents the second weight, which is the weight from the encoder hidden layer to the encoder output layer; b2 represents the second bias term, which is the bias term of the encoder output layer.

[0109] In this embodiment, to determine whether the similarity or proximity between the data received and transmitted downwards by the encoder input layer and the data input to the encoder output layer is within a preset similarity range, a loss function L(·) is set in the noise reduction encoder in this embodiment, specifically:

[0110]

[0111] In formula (9), n represents the total number of features i; λ(ε i ≠0) indicates that when the damaged data of the i-th feature has no noise added, a penalty term of size λ is added. By adding an extra penalty term, a balance point in the model training process can be found, so that the model can effectively remove noise and maintain the stability of model training.

[0112] Assuming the parameters θ = (w1, b1, w2, b2), θ can be obtained by minimizing the loss function L(·). The objective function J(·) of the improved denoising autoencoder is set as follows:

[0113] J(θ)=∑ x∈s L(x,f(g(x))) (10)

[0114] In formula (10), S represents the total number of damaged data.

[0115] S205. Using historical decoding data as input data for training, an unsupervised training method is employed to obtain an isolated forest layer used to output poor quality recognition results.

[0116] Please see Figure 3 After training the denoising encoder, this improved denoising encoder is used as a feature extractor, and the historical decoding data output by the final denoising encoder output layer is used as the input data for subsequent network structures. In this embodiment, the subsequent network structure is an isolated forest layer, and the processing method of this isolated forest layer is as follows:

[0117] S2051. Randomly select a preset number of N sample data from the historical decoded data as sub-samples and place them in the root node of the tree.

[0118] S2052. Randomly set a specified dimension and randomly generate a cut point in the root node. The cut point is generated between the maximum and minimum values ​​of the specified dimension in the current root node data.

[0119] S2053. A hyperplane is generated using this cutting point. Then, based on the cutting point, the data space of the current root node is divided into two child nodes, i.e., subspaces: sample data with dimensions lower than the specified dimension are placed to the left of the current root node, and sample data with dimensions not lower than the specified dimension are placed to the right of the current root node.

[0120] S2054. Recursively repeat steps S2052 and S2053 in the subspace of the current root node, continuously constructing new subspaces until only one sample data in the subspace cannot be further segmented or the subspace has reached the preset height. After obtaining t subtrees, the training ends.

[0121] In this embodiment, the isolated forest layer does not output true or false, but outputs a value of 0-1, which represents the degree of data abnormality. The larger the value, the higher the degree of abnormality.

[0122] Accordingly, step S102 specifically includes:

[0123] S1021. Input the user data into the noise reduction coding layer. The noise reduction coding layer adds preset noise to the user data and then performs encoding and decoding processing to obtain decoded data.

[0124] S1022. Input the decoded data into the isolated forest layer to obtain the quality poor identification result output by the isolated forest layer.

[0125] Please see Figure 4 In this process, user data can be stored in Kafka, and the poor quality identification model can be integrated into a Spark task. The poor quality identification model consumes data from Kafka to obtain the poor quality identification results and stores them in Elasticsearch.

[0126] The broadband service quality poor identification method of this invention identifies quality problems through a trained quality poor identification model, obtaining the quality poor identification result. The quality poor identification model adds a noise reduction coding layer in the form of an autoencoder before the isolated forest layer, which solves the problem that the traditional isolated forest algorithm can only perform linear segmentation, while retaining the advantage that the isolated forest algorithm does not need to calculate distance and density. This can significantly improve speed and reduce system overhead. The noise reduction coding layer can reduce the impact of reported erroneous data on the subsequent identification accuracy and improve noise resistance. Furthermore, the data used for training is obtained by adding network loss to the PON network, thus constructing historical user experience quality poor data that reflects the user's real experience. This historical user experience quality poor data contains the user's real experience information, thus filling the gap of unlabeled data in the broadband user quality poor identification scenario. Therefore, it improves the identification accuracy and robustness of the quality poor identification model, realizes the proactive assessment of broadband service quality level, so as to accurately label each potential quality poor user, thereby discovering problems in advance and carrying out broadband maintenance in a timely manner, improving user experience and stickiness.

[0127] The apparatus provided in the embodiments of the present invention will be described below. The apparatus described below can be referred to in correspondence with the method described above.

[0128] Please see Figure 5 , Figure 5 A schematic diagram of a broadband service quality poor identification device according to an embodiment of the present invention is shown. The device may include:

[0129] The data acquisition module 10 is used to acquire user usage data generated by the user using the broadband network.

[0130] In this embodiment, user data can be pre-stored in the electronic device or obtained by the electronic device from an external source. For example, the electronic device can obtain the data from an external broadband service acquisition device. There are no restrictions on the specific form in which the user data is obtained, as long as the electronic device can acquire the user data.

[0131] The poor quality identification module 20 is used to input user data into the trained poor quality identification model and obtain the poor quality identification result output by the poor quality identification model. The poor quality identification result is used to indicate users with poor quality problems. The poor quality identification model is trained based on historical poor quality data generated after users use a broadband network with preset network loss added to the network parameters.

[0132] By adding network losses to the PON network, historical data reflecting the user's actual experience of poor quality can be constructed. This historical data contains the user's actual experience information, thus filling the gap of unlabeled data in the broadband user quality identification scenario.

[0133] The broadband service quality poor identification device of this invention identifies quality problems through a trained quality poor identification model, and obtains the quality poor identification result. The quality poor identification model adds a noise reduction coding layer in the form of an autoencoder before the isolated forest layer, which solves the problem that the traditional isolated forest algorithm can only perform linear segmentation, while retaining the advantage that the isolated forest algorithm does not need to calculate distance and density. This can significantly improve speed and reduce system overhead. The noise reduction coding layer can reduce the impact of reported erroneous data on the subsequent identification accuracy and improve noise resistance. Furthermore, the data used for training is obtained by adding network loss to the PON network, which can construct historical user experience quality poor data that reflects the user's real experience. The historical user experience quality poor data contains the user's real experience information, thus filling the gap of unlabeled data in the broadband user quality poor identification scenario. Therefore, it improves the identification accuracy and robustness of the quality poor identification model, realizes the proactive assessment of broadband service quality level, so as to accurately label each potential quality poor user, thereby discovering problems in advance and carrying out broadband maintenance in a timely manner, improving user experience and stickiness.

[0134] Figure 6 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 6 As shown, the electronic device may include: a processor 610, a communication interface 620, a memory 630, and a communication bus 640, wherein the processor 610, the communication interface 620, and the memory 630 communicate with each other through the communication bus 640. The processor 610 can call logical commands in the memory 630 to execute a broadband service quality poor identification method, which includes:

[0135] Acquire user usage data generated by users using broadband networks;

[0136] The user data is input into the trained poor quality identification model to obtain the poor quality identification result output by the poor quality identification model; the poor quality identification result is used to indicate users with poor quality problems; the poor quality identification model is trained based on historical poor quality data generated after users use broadband networks with preset network loss added to network parameters.

[0137] The poor quality identification model is trained through the following steps:

[0138] Add preset network loss to network parameters in broadband networks that exceed the level of abnormal impact;

[0139] Acquire historical data on poor network quality generated after users use broadband networks with preset network loss parameters;

[0140] Data cleaning was performed on historical poor-quality data to obtain historical network anomaly data;

[0141] The historical network anomaly data is used as input data for training. By adding preset noise to the historical network anomaly data and using unsupervised learning, the goal is to control the similarity between the input data and the output data within a preset similarity range, thus obtaining a noise reduction coding layer for output decoding data.

[0142] The historical decoding data output by the noise reduction coding layer is used as the input data for training, and the training is carried out in an unsupervised manner to obtain an isolated forest layer for outputting poor quality recognition results.

[0143] Furthermore, the logical instructions in the aforementioned memory 630 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 the present invention, 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 the present invention. 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.

[0144] On the other hand, the present invention also provides a computer program product, the computer program product comprising a computer program stored on a non-transitory computer-readable storage medium, the computer program comprising program instructions, wherein when the program instructions are executed by a computer, the computer is able to execute the broadband service quality poor identification method provided by the above methods, the method comprising:

[0145] Acquire user usage data generated by users using broadband networks;

[0146] The user data is input into the trained poor quality identification model to obtain the poor quality identification result output by the poor quality identification model; the poor quality identification result is used to indicate users with poor quality problems; the poor quality identification model is trained based on historical poor quality data generated after users use broadband networks with preset network loss added to network parameters.

[0147] The poor quality identification model is trained through the following steps:

[0148] Add preset network loss to network parameters in broadband networks that exceed the level of abnormal impact;

[0149] Acquire historical data on poor network quality generated after users use broadband networks with preset network loss parameters;

[0150] Data cleaning was performed on historical poor-quality data to obtain historical network anomaly data;

[0151] The historical network anomaly data is used as input data for training. By adding preset noise to the historical network anomaly data and using unsupervised learning, the goal is to control the similarity between the input data and the output data within a preset similarity range, thus obtaining a noise reduction coding layer for output decoding data.

[0152] The historical decoding data output by the noise reduction coding layer is used as the input data for training, and the training is carried out in an unsupervised manner to obtain an isolated forest layer for outputting poor quality recognition results.

[0153] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the aforementioned methods for identifying poor broadband service quality, the method comprising:

[0154] Acquire user usage data generated by users using broadband networks;

[0155] The user data is input into the trained poor quality identification model to obtain the poor quality identification result output by the poor quality identification model; the poor quality identification result is used to indicate users with poor quality problems; the poor quality identification model is trained based on historical poor quality data generated after users use broadband networks with preset network loss added to network parameters.

[0156] The poor quality identification model is trained through the following steps:

[0157] Add preset network loss to network parameters in broadband networks that exceed the level of abnormal impact;

[0158] Acquire historical data on poor network quality generated after users use broadband networks with preset network loss parameters;

[0159] Data cleaning was performed on historical poor-quality data to obtain historical network anomaly data;

[0160] The historical network anomaly data is used as input data for training. By adding preset noise to the historical network anomaly data and using unsupervised learning, the goal is to control the similarity between the input data and the output data within a preset similarity range, thus obtaining a noise reduction coding layer for output decoding data.

[0161] The historical decoding data output by the noise reduction coding layer is used as the input data for training, and the training is carried out in an unsupervised manner to obtain an isolated forest layer for outputting poor quality recognition results.

[0162] 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.

[0163] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention 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; and these 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 the present invention.

Claims

1. A method for identifying poor broadband service quality, characterized in that: The method includes: Acquire user usage data generated by users using broadband networks; The user data is input into the trained poor quality identification model to obtain the poor quality identification result output by the poor quality identification model; the poor quality identification result is used to indicate users with poor quality problems; the poor quality identification model is trained based on historical poor quality data generated after users use broadband networks with preset network loss added to network parameters. The poor quality identification model is trained through the following steps: Add preset network loss to network parameters in broadband networks that exceed the level of abnormal impact; Acquire historical data on poor network quality generated after users use broadband networks with preset network loss parameters; Data cleaning was performed on historical poor-quality data to obtain historical network anomaly data; The historical network anomaly data is used as input data for training. By adding preset noise to the historical network anomaly data and using unsupervised learning, the goal is to control the similarity between the input data and the output data within a preset similarity range, thus obtaining a noise reduction coding layer for output decoding data. The historical decoding data output by the noise reduction coding layer is used as the input data for training, and the training is carried out in an unsupervised manner to obtain an isolated forest layer for outputting poor quality recognition results. The step of using the historical decoded data output by the noise reduction coding layer as input data for training, and training in an unsupervised manner to obtain an isolated forest layer for outputting poor quality recognition results, specifically includes: Filter a preset number of the historical decoded data and place them into the root node of the tree; A cut point is randomly generated in the root node; the cut point is located between the maximum and minimum values ​​of a specified dimension in the root node data; The root node is divided according to the cut point. Sample data with dimensions lower than the specified dimension are placed in one subspace of the root node, and sample data with dimensions not lower than the specified dimension are placed in the other subspace of the root node. The training of the isolated forest layer is completed when each root node is continuously divided until there is only one piece of the historical decoded data in the subspace of the root node or the subspace has reached the preset height. The loss function of the noise reduction coding layer is: ； in, Represents the loss function; This indicates the historical network anomaly data after adding preset noise. Damaged data for each feature; Indicates the first Historical network anomaly data with several characteristics; Representation of features The total number; Indicates the penalty item, when the first... When no noise is added to the damaged data of each feature, add a noise of size 1. The penalty items.

2. The broadband service quality poor identification method as described in claim 1, characterized in that: The addition of preset network loss to network parameters in the broadband network that exceed the level of abnormal impact specifically includes: Acquire key performance indicator data of broadband network users and extract network anomaly data from the key performance indicator data of broadband network users. The system obtains the degree of impact of abnormal network data, arranges and combines the network parameters corresponding to abnormal network data that exceed the degree of impact, and adds preset network loss to the combined network parameters.

3. The broadband service quality poor identification method as described in claim 1, characterized in that: The process of cleaning historical poor-quality data to obtain historical network anomaly data specifically includes: Fill in the missing values ​​in the historical data of poor sensory quality; Historical network anomaly data is extracted from historical perceptual quality differences after null filling.

4. The broadband service quality poor identification method as described in claim 1, characterized in that: The noise reduction coding layer specifically includes: The encoder input layer, noise addition layer, encoder hidden layer, and encoder output layer are connected in sequence. The encoder input layer is used to receive and transmit historical network anomaly data downwards; the noise addition layer is used to add preset noise to the historical network anomaly data to obtain damaged data; the encoder hidden layer is used to transmit data through an encoding function. The damaged data is encoded to obtain historical encoded data, and then decoded using a decoding function. The historical encoded data is decoded to obtain historical decoded data; the encoder output layer is used to receive and transmit the historical decoded data to the isolated forest layer.

5. The broadband service quality poor identification method as described in claim 4, characterized in that: The step of inputting the user data into the trained poor quality identification model to obtain the poor quality identification result output by the poor quality identification model specifically includes: The user data is input into the noise reduction coding layer, which adds preset noise to the user data and then performs encoding and decoding processing to obtain decoded data. The decoded data is input into the isolated forest layer to obtain the quality defect identification result output by the isolated forest layer.

6. A broadband service quality poor identification device, characterized in that: The device includes: The data acquisition module is used to acquire user usage data generated by users using the broadband network; The poor quality identification module is used to input the user's usage data into the trained poor quality identification model to obtain the poor quality identification result output by the poor quality identification model; the poor quality identification result is used to indicate users with poor quality problems; the poor quality identification model is trained based on historical user experience poor quality data generated after the user used a broadband network with preset network loss added to the network parameters. The poor quality identification model is trained through the following steps: Add preset network loss to network parameters in broadband networks that exceed the level of abnormal impact; Acquire historical data on poor network quality generated after users use broadband networks with preset network loss parameters; Data cleaning was performed on historical poor-quality data to obtain historical network anomaly data; The historical network anomaly data is used as input data for training. By adding preset noise to the historical network anomaly data and using unsupervised learning, the goal is to control the similarity between the input data and the output data within a preset similarity range, thus obtaining a noise reduction coding layer for output decoding data. The historical decoding data output by the noise reduction coding layer is used as the input data for training, and the training is carried out in an unsupervised manner to obtain an isolated forest layer for outputting poor quality recognition results. The poor quality identification module is also used to filter a preset number of the historical decoded data and put them into the root node of the tree; A cut point is randomly generated in the root node; the cut point is located between the maximum and minimum values ​​of a specified dimension in the root node data; The root node is divided according to the cut point. Sample data with dimensions lower than the specified dimension are placed in one subspace of the root node, and sample data with dimensions not lower than the specified dimension are placed in the other subspace of the root node. The training of the isolated forest layer is completed when each root node is continuously divided until there is only one piece of the historical decoded data in the subspace of the root node or the subspace has reached the preset height. The loss function of the noise reduction coding layer is: ； in, Represents the loss function; This indicates the historical network anomaly data after adding preset noise. Damaged data for each feature; Indicates the first Historical network anomaly data with several characteristics; Representation of features The total number; Indicates the penalty item, when the first... When no noise is added to the damaged data of each feature, add a noise of size 1. The penalty items.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the broadband service quality poor identification method as described in any one of claims 1 to 5.

8. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When a computer program is executed by a processor, it implements the steps of the broadband service quality poor identification method as described in any one of claims 1 to 5.

Images