A network quality identification method, device, apparatus and storage medium

Abstract

The application relates to the computer technical field, in particular to a network quality identification method and device, equipment and storage medium, the method comprises the following steps: acquiring network index data of a first time period; inputting the network index data of the first time period into a first model to obtain a network quality identification result output by the first model, wherein the network quality identification result is used for representing network quality conditions of the first time period; if the network quality of the first time period is abnormal, the network quality identification result is also used for determining an application type in which network quality abnormality is likely to occur.

Description

Technical Field

[0001] This application relates to the field of computer technology, specifically to a network quality identification method, apparatus, device, and storage medium. Background Technology

[0002] With the rapid development of wireless communication technology, the network has become an indispensable part of life. Network quality directly affects users' daily activities, including browsing the web, playing online games, and watching videos. However, network performance is affected by a variety of factors, including signal strength and latency, which can lead to abnormal network quality and thus affect user experience. Summary of the Invention

[0003] This application provides a network quality identification method, apparatus, device, and storage medium that can efficiently identify network quality.

[0004] In a first aspect, embodiments of this application provide a network quality identification method, the method comprising:

[0005] Obtain network metric data for the first time period;

[0006] The network indicator data of the first time period is input into the first model to obtain the network quality identification result output by the first model. The network quality identification result is used to characterize the network quality situation of the first time period. If the network quality of the first time period is abnormal, the network quality identification result is also used to determine the application type that may have network quality abnormalities.

[0007] In this embodiment, network indicator data for a first time period is input into a first model. The network quality identification result output by the first model can not only accurately identify the network quality situation during that time period, but also further identify the specific application type of the network quality anomaly during that time period if the network quality is determined to be abnormal. This method makes network quality identification more accurate and efficient, and can help technicians solve network problems in a targeted manner and improve user experience.

[0008] In one possible embodiment, the application type that may experience network anomalies includes one of video playback, games, web pages, or instant messaging.

[0009] In one possible embodiment, the network quality identification result includes a network quality score; if the network quality score falls within a first value range, the network quality identification result indicates that the network quality is normal during the first time period; if the network quality score falls within a second value range, the network quality identification result indicates that the network quality is abnormal during the first time period.

[0010] In this embodiment, by dividing the first value range and the second value range, the network quality situation in the first time period can be accurately determined based on which range the network quality score falls within.

[0011] In one possible embodiment, the second value range includes K sub-value ranges; the K sub-value ranges correspond one-to-one with K application types, where K is a positive integer; if the network quality score belongs to the second value range and the network quality score belongs to the i-th sub-value range, then the network quality identification result indicates that the application type that may have network quality anomalies is the application type corresponding to the i-th sub-value range, where the i-th sub-value range is one of the K sub-value ranges, i is a positive integer, and i≤K.

[0012] In this embodiment, the second value range includes multiple sub-value ranges, each of which corresponds to a different application type. If the network quality identification result output by the first model indicates network quality anomalies in the first time period, the application type that may have network quality anomalies in the corresponding time period can be quickly determined based on which sub-value range the network quality score belongs to.

[0013] In one possible embodiment, the first value range, the second value range, and the K sub-value ranges are determined based on the first network indicator dataset and the first user experience dataset. The first network indicator dataset is associated with the first user experience dataset. The first network indicator dataset includes historical network indicator data for M time periods. The first user experience dataset includes user experience data for the M time periods. The user experience data for each time period includes the network quality status for that time period. The user experience data for time periods with abnormal network quality also includes the application type with abnormal network quality. M is a positive integer.

[0014] In this embodiment, the user experience data for each time period includes the network quality status for that time period. When the network quality status is abnormal, the user experience data also includes the application type that has experienced network quality abnormalities, i.e., the user's actual experience data. By combining network indicator data and user experience data, a more accurate evaluation standard can be provided for the first model, accurately identifying the application type that may experience network quality abnormalities in the corresponding time period. This is beneficial for technicians to analyze the causes of network quality abnormalities and solve network problems in a targeted manner.

[0015] In one possible embodiment, determining the first value range and the second value range based on the first network indicator dataset and the first user experience dataset includes: inputting the first network indicator dataset into the first model to obtain the network quality scores for the M time periods; and obtaining the first value range and the second value range based on the network quality scores for the M time periods and the user experience data for each time period, including the network quality situation for the corresponding time period.

[0016] In this embodiment, the first network indicator dataset is input into the first model to obtain network quality scores for each time period. Based on the obtained network quality scores and the network quality status of the corresponding time period in the user experience data, a score range (first value range) representing normal network quality and a score range (second value range) representing abnormal network quality can be obtained.

[0017] In one possible embodiment, determining the K sub-value ranges based on a first network metric dataset and a first user experience dataset includes: determining the K sub-value ranges based on user experience data during time periods when the network quality is abnormal, and the network quality score during those time periods, wherein the network quality score during those time periods belongs to the network quality scores of the M time periods.

[0018] In this embodiment, the network quality scores for time periods with abnormal network quality are verified using the application types recorded in the first user experience dataset. Each network quality score is associated with the application type of the corresponding time period, and network quality scores with the same application type are grouped into the same value range. This process is repeated to form K sub-value ranges, each corresponding to one application type. Under this method, if the model output indicates abnormal network quality in the first time period, the application types that may have abnormal network quality can be accurately identified using the network quality score and the K sub-value ranges.

[0019] In one possible embodiment, if the network quality identification result indicates network quality anomalies in the first time period, user feedback results for the first time period are obtained, and the user feedback results indicate a first application type; if the application type that may experience network quality anomalies is different from the first application type, the network indicator data for the first time period is added to the first network indicator dataset.

[0020] In this embodiment, when the network quality identification result indicates that the network quality is abnormal in the first time period, based on the collected user network usage feedback in the corresponding time period, i.e. the user's actual network usage experience data, the data that the first model identified incorrectly can be added back to the first network indicator dataset, and the first model can be updated using the updated first network indicator dataset, which can form a more accurate first value range, second value range and K sub-value ranges, so as to improve the identification accuracy.

[0021] In one possible embodiment, if the network quality identification result indicates that the network quality is abnormal in the first time period, the cause of the network quality abnormality in the first time period is determined based on the network quality identification result and the network indicator data of the first time period.

[0022] In this embodiment, when the network quality identification result indicates that the network quality is abnormal in the first time period, the specific reason for the network quality abnormality in that time period can be determined based on the application type indicated by the network quality identification result and the network indicator data of the first time period. For example, during the first time period, the video buffering of the user is due to insufficient bandwidth or signal interference.

[0023] In one possible embodiment, an isolated forest model is trained based on a second network metric dataset to obtain the first model; wherein the second network metric dataset includes historical network metric data for N time periods, where N is a positive integer.

[0024] In this embodiment, the isolated forest model used is not only more suitable for handling anomaly detection tasks than other supervised machine learning algorithms such as random forest, but also does not require a large amount of data labeling, which greatly reduces the manual input in building the dataset and improves the iteration speed and accuracy of the model.

[0025] In one possible embodiment, the network metric data for the first time period includes at least one of channel utilization, signal-to-noise ratio, signal strength, noise floor, number of uplink retransmission packets, number of downlink retransmission packets, uplink retransmission rate, downlink retransmission rate, number of uplink packet loss, number of downlink packet loss, uplink packet loss rate, downlink packet loss rate, and total packet loss rate.

[0026] Secondly, embodiments of this application also provide a network quality identification device, the device including a communication unit and a processing unit;

[0027] The communication unit is used to acquire network indicator data for the first time period;

[0028] The processing unit is used to input network indicator data of the first time period into the first model to obtain the network quality identification result output by the first model, wherein the network quality identification result is used to characterize the network quality situation of the first time period; if the network quality of the first time period is abnormal, the network quality identification result is also used to determine the application type that may have network quality abnormalities.

[0029] Thirdly, this application provides a network quality identification device, comprising:

[0030] Memory, used to store program instructions;

[0031] A processor is configured to invoke program instructions stored in the memory and execute the steps of the method described in any one of the first aspects according to the obtained program instructions.

[0032] Fourthly, this application provides a computer-readable storage medium storing a computer program, the computer program including program instructions, which, when executed by a computer, cause the computer to perform the method described in any one of the first aspects.

[0033] Fifthly, this application provides a computer program product comprising: computer program code, which, when run on a computer, causes the computer to perform the method described in any one of the first aspects. Attached Figure Description

[0034] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0035] Figure 1 A flowchart of the network quality identification method provided in the embodiments of this application;

[0036] Figure 2 A schematic diagram illustrating the network quality identification result obtained based on the first model, provided as an embodiment of this application;

[0037] Figure 3 This is a schematic diagram of the network quality identification method provided in the embodiments of this application;

[0038] Figure 4 Schematic diagram of the network quality identification device provided in the embodiments of this application Figure 1 ;

[0039] Figure 5Schematic diagram of the network quality identification device provided in the embodiments of this application Figure 2 . Detailed Implementation

[0040] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0041] Home Wi-Fi experience refers to the difference in user experience caused by variations in network metrics when using mobile devices such as phones and tablets to connect to the internet through a router. Existing network quality assessment methods generally use a "one-size-fits-all" approach: when a certain network metric falls below an experienced value, it's considered an abnormal network quality. However, in reality, a user's actual internet experience depends not only on network metrics but also on the applications they are using. For example, while a user might experience buffering while watching a video, browsing the web or text chatting might not actually be affected.

[0042] Therefore, existing network quality identification methods often rely solely on network metric data to assess network quality, neglecting to incorporate real-world user experience data. This results in the ability to identify only the network quality level, failing to provide more granular information and hindering the resolution of network quality issues. Furthermore, traditional network quality identification methods are typically supervised approaches, requiring significant human and material resources to construct datasets and hindering flexible updates.

[0043] Based on the above issues, such as Figure 1 As shown in the embodiment of this application, a method for network quality identification is provided, which specifically includes:

[0044] Step 100: The server obtains network metric data for the first time period.

[0045] First, network metric data is collected using routers or other network devices deployed in the home network, and the collected data is uploaded to a server. Data collection can be performed at regular time intervals; this application does not specify a particular time interval.

[0046] The network metrics data include at least one of the following: channel utilization, signal-to-noise ratio, signal strength, noise floor, number of uplink retransmission packets, number of downlink retransmission packets, uplink retransmission rate, downlink retransmission rate, number of uplink packet loss, number of downlink packet loss, uplink packet loss rate, downlink packet loss rate, and total packet loss rate.

[0047] In one embodiment, if data collection starts at 9:00 AM and can be performed every X minutes (e.g., X = 60), the router can collect various network indicator data at the top of the hour (9:00, 10:00, 11:00, etc.) and upload the network indicator data for each time period to the server.

[0048] Step 110: The server inputs the network indicator data of the first time period into the first model and obtains the network quality identification result output by the first model. The network quality identification result is used to characterize the network quality of the first time period. If the network quality of the first time period is abnormal, the network quality identification result is also used to determine the application type that may have abnormal network quality. The application type includes one of video playback, games, web pages or instant messaging.

[0049] Before inputting the network metric data of the first time period into the first model, the server first obtains the first model based on the first network metric dataset, the second network metric dataset, and the first user experience dataset, and the first model is constructed based on the isolated forest model.

[0050] In one embodiment, the server may obtain the first model in, but is not limited to, the following ways:

[0051] 1. The server obtains historical datasets, which include the first network metric dataset, the second network metric dataset, and the first user experience dataset.

[0052] The dataset comprises two sets of data: a first network metrics dataset (including historical network metrics data for M time periods) and a first user experience dataset (including user experience data for M time periods). The first network metrics dataset and the first user experience dataset are associated. The user experience data for each time period includes the network quality information for that period. For time periods with abnormal network quality, the user experience data also includes the application type exhibiting abnormal network quality. The second network metrics dataset includes historical network metrics data for N time periods. M and N are positive integers, and the M time periods are different from the N time periods.

[0053] The parameters included in the historical network indicator data for each time period are the same as those included in the network indicator data for the first time period collected above, and will not be elaborated further here.

[0054] In one embodiment, if the acquired data is a historical dataset containing 100 time periods, the dataset includes historical network indicator data for 40 time periods as a first network indicator dataset. The first user experience dataset includes user experience data for these 40 time periods. The user experience data for each time period records the network quality situation in detail. If the network quality is abnormal, the user experience data also records the application type that experienced the abnormal network quality during that time period, such as video playback, games, web pages, instant messaging, etc. Alternatively, the user experience data for each time period records the specific application (APP) identifier, such as video playback application A, web page application B, instant messaging C, etc. This user experience data can be recorded manually. The second network indicator dataset includes historical network indicator data for the remaining 60 time periods, which do not contain corresponding user experience data.

[0055] In one embodiment, taking 9:00 to 21:00 on October 20, 2024 as an example, for a router, the server collected network indicator data for multiple time periods at certain time intervals. User experience data for these multiple time periods can be recorded simultaneously. For example, assuming each time period is 20 minutes, the data on internet access experience for multiple time periods recorded instances of game lag during 10:00-10:20, video buffering during 10:21-10:40, and inability to send images in chat during 12:01-12:20. The user experience data for multiple time periods also recorded smooth video playback using video playback application A during 10:00-10:20, and smooth web browsing using web application B during 10:21-11:00.

[0056] Second, the server obtains the first model based on the second network metric dataset.

[0057] The first model is built based on the Isolation Forest model. The Isolation Forest model is an unsupervised machine learning algorithm, commonly used for anomaly detection tasks. It identifies outliers in a dataset by constructing a set of random decision trees to isolate data points.

[0058] The server uses the second network metric dataset as input to train the Isolation Forest model, obtaining the first model. The first model's parameters can be adjusted using the first network metric dataset and the first user experience dataset; for example, through grid search, the number of trees, depth, number of subsets, and anomaly ratio can be adjusted.

[0059] Third, the server determines the first value range, the second value range, and K sub-value ranges based on the first network metric dataset and the first user experience dataset.

[0060] After determining the first model, the first network indicator dataset is input into the first model to obtain network quality scores for M time periods. Based on the network quality scores for the M time periods and the network quality status of the corresponding time period included in the user experience data for each time period, a first value range and a second value range are obtained. The first value range is used to represent the score range of normal network quality, and the second value range is used to represent the score range of abnormal network quality.

[0061] Subsequently, K sub-value ranges can be determined based on user experience data during the time period when network quality is abnormal, and the network quality score during the time period when network quality is abnormal. Among them, the network quality score during the time period when network quality is abnormal belongs to the network quality scores of the above M time periods.

[0062] In one embodiment, the network quality score for a time period with abnormal network quality is associated with the user experience data (including application type or APP identifier) ​​for the corresponding time period. Network quality scores with the same application type or APP identifier are grouped into the same value range, and so on, forming K sub-value ranges. Each sub-value range corresponds to one application type or multiple APP identifiers. The multiple APP identifiers corresponding to each sub-value range can be identified as corresponding application types according to the constructed classification rules. For example, video playback application A and video playback application B can be identified as video playback, and game application A and game application B can be identified as games. This application does not limit this; the following embodiments are all described using application types.

[0063] For example, the first and second value ranges can be obtained as shown in Table 1.

[0064] First value range Second range of values Example 1 [0,1] [-1,0) Example 2 [50,100] [0,50) Example 3 … …

[0065] Table 1

[0066] As shown in Table 1, the first value range can be [0,1], and the second value range can be [-1,0); for example, the first value range can be [50,100], and the second value range can be [0,50].

[0067] Furthermore, the second value range includes K sub-value ranges, each of which corresponds to an application type. These sub-value ranges can be used to represent the score range corresponding to the application type that has network quality anomalies. For example, taking the second value range as [-1, 0), the K sub-value ranges in the second value range can be as shown in Table 2.

[0068] Sub-value range Score Application type The first sub-range of values [-0.2,0) Instant Messaging The second sub-range of values [-0.5,-0.2) Web page The third sub-value range [-0.8,-0.5) video The fourth sub-value range [-1,-0.8) game

[0069] Table 2

[0070] As shown in Table 2, the second value range can include four sub-value ranges. Among them, [-0.2, 0) corresponds to the application type of instant messaging, indicating that the application type that may experience network quality abnormalities is instant messaging; [-0.5, -0.2] corresponds to the application type of web page, indicating that the application type that may experience network quality abnormalities is web page; (-0.8, -0.5] corresponds to the application type of video playback, indicating that the application type that may experience network abnormalities is video playback; and (-1, -0.8] corresponds to the application type of game, indicating that the application type that may experience network abnormalities is game.

[0071] After obtaining the first model, the first value range, the second value range, and K sub-value ranges, as follows: Figure 2 As shown, the server can input network indicator data for the first time period into the first model to obtain the network quality identification result output by the first model. The network quality identification result is used to characterize the network quality situation in the first time period. If the network quality is abnormal in the first time period, the network quality identification result is also used to determine the application type that may have network quality abnormalities.

[0072] In one possible implementation, the server inputs network indicator data for a first time period into a first model. The first model analyzes the input network indicator data and outputs a network quality identification result. The network quality identification result can be represented by two identifiers to indicate whether the network quality is normal or abnormal during the first time period. Furthermore, other different identifiers can be used to represent the application types that may experience network quality abnormalities.

[0073] In one embodiment, the network quality identification result output by the first model can be represented by two identifiers. For example, an output result of identifier "1" indicates that the network quality is normal during the first time period; an output result of identifier "-1" indicates that the network quality is abnormal during the first time period. Further, the output result can include a second identifier to indicate the application type that may experience network quality abnormalities. For example, an output result of identifier "-11" indicates that the network quality is abnormal during the first time period, and according to the first model analysis, the application type that may experience network quality abnormalities is gaming; an output result of identifier "-12" indicates that the network quality is abnormal during the first time period, and according to the first model analysis, the application type that may experience network quality abnormalities is video playback. The method of using identifiers for the output result is merely an example and is not intended to limit the scope of this application.

[0074] In one possible implementation, the server inputs network indicator data for a first time period into a first model, the first model analyzes the input network indicator data, and outputs network quality identification results, which may include network quality scores.

[0075] If the network quality score falls within the first value range, the network quality identification result indicates that the network quality is normal in the first time period.

[0076] In one embodiment, taking case A in Table 1 above as an example, if the network quality identification result output by the first model is 0.5, then based on the first value range of [0,1], and 0.5 belonging to the first value range, it can be determined that the network quality of the first time period is normal.

[0077] Specifically, if the network quality score falls within the second value range, the network quality identification result indicates an anomaly in network quality during the first time period. Further, if the network quality score falls within both the second and i-th sub-value ranges, the network quality identification result indicates that the application type potentially experiencing network quality anomalies corresponds to the i-th sub-value range, where the i-th sub-value range is one of the K sub-value ranges.

[0078] In one embodiment, taking case A in Table 1 above as an example, if the network quality identification result output by the first model is -0.7, then based on the second value range of [-1, 0), and -0.7 belonging to the second value range, it can be determined that the network quality is abnormal in the first time period. Further, taking Table 2 above as an example, since the second value range contains four sub-value ranges, and -0.7 belongs to one of these sub-value ranges [-0.8, -0.5), based on the application type (video playback) corresponding to this sub-value range, it can be determined that the application type likely to experience network quality abnormalities in the first time period is video playback.

[0079] Furthermore, when the network quality identification result output by the first model indicates network quality anomalies in the first time period, the cause of the network quality anomalies in the first time period can be determined based on the network quality identification result and the network indicator data of the first time period.

[0080] In one embodiment, when the network quality identification result indicates network quality anomalies in a first time period, the result can be uploaded to the server, triggering anomaly analysis. The server can identify the causes of the network quality anomalies in that time period based on the application types and various network indicator data indicated by the network quality identification result. Specifically, algorithms such as random forests can be used to evaluate the impact of each network indicator on the application type and output the importance ratio of each indicator. By analyzing network indicators with higher importance ratios, the specific causes of the network quality anomalies in that time period can be identified.

[0081] For example, if the network quality identification result indicates that the application type potentially experiencing network quality anomalies is video playback, then using algorithms such as Random Forest, the key network indicator affecting the video viewing experience can be identified as channel utilization. Based on this indicator, technicians can analyze the specific values ​​and trends of channel utilization to determine whether network bandwidth is being over-utilized or insufficient. As another example, if the network quality identification result indicates that the application type potentially experiencing network quality anomalies is gaming, then using algorithms such as Random Forest, the key network indicator affecting the gaming experience can be identified as packet loss rate. Based on this indicator, technicians can monitor the specific values ​​and trends of packet loss rate to determine whether the packet loss rate is higher or lower than the normal range.

[0082] Based on the analysis results, the server can optimize the network configuration in a targeted manner, propose network optimization suggestions, and record the analysis results and optimization measures in the analysis report to support future network management and optimization.

[0083] In one possible implementation, when the network quality identification result indicates that the network quality is abnormal in the first time period, the server can obtain the user feedback result of the first time period. The user feedback result indicates the first application type. If the application type that may have network quality abnormality is different from the first application type, the network indicator data of the first time period is added to the first network indicator dataset.

[0084] In one embodiment, when the first model outputs a network quality identification result indicating network quality anomalies in a first time period, user follow-up can be conducted. Feedback can be collected through questionnaires, phone calls, in-person visits, etc. This feedback includes the user's application usage within the first time period, inquiring whether they encountered issues such as video buffering, game lag, long webpage loading times, or other network-related problems. The user's actual feedback is compared with the first model's output. If the first model's output (application types that may exhibit network quality anomalies) matches the user's feedback, the effectiveness of the first model is verified, achieving accurate network quality identification. If the first model's output (application types that may exhibit network quality anomalies) differs from the user's feedback, the network indicator data for the first time period is added to the first network indicator dataset. This process is repeated to collect regional user feedback, adding network indicator data inconsistent with user feedback to the first network indicator dataset, enabling flexible dataset updates. Furthermore, updating the first model using the updated first network indicator dataset can generate more accurate first value range, second value range, and K sub-value ranges, contributing to improved identification accuracy.

[0085] In one embodiment, the test results obtained by actually visiting users in person are shown in Table 3.

[0086]

[0087] Table 3

[0088] As can be seen from Table 3, the accuracy rate of network quality anomaly detection is 100%, indicating that the first model can accurately identify all network quality anomalies, and technicians can solve most network problems based on the specific causes analyzed.

[0089] In summary, this application provides a network quality identification method that uses a combination of network indicator data and user experience data to obtain a first model. When the network indicator data for a first time period is input into the first model, not only can the network quality situation for that time period be accurately obtained, but also the application types that may be experiencing network quality anomalies can be identified when network quality anomalies are confirmed. This method makes network quality identification more accurate and efficient, which is conducive to technicians to propose targeted network solutions, thereby improving the efficiency of network fault handling.

[0090] like Figure 3 As shown, a schematic diagram of a network quality identification method is provided. The method can be divided into three parts: a data acquisition part, a model building part, and a platform analysis part. In the data acquisition part, the first database includes... Figure 1The historical datasets in step 110, namely the first network metric dataset, the second network metric dataset, and the first network experience dataset, include the second database. Figure 1 The network metric data for the first time period in step 100; used in the model building section. Figure 1 Step 110 utilizes data from the first database for model training, validation, and fine-tuning to obtain the first model, and also uses data from the second database to input into the first model to obtain network quality identification results; in the platform analysis section, it is used for Figure 1 Step 110 analyzes the reasons for network quality anomalies in the network quality identification results, conducts user follow-ups to obtain the application types actually used by users, and also uses data that is inconsistent between the results output by the first model (application types that may have network quality anomalies) and the application types actually reported by users to be fed back to the first database to optimize the first model.

[0091] Figure 4 and Figure 5 This is a schematic diagram illustrating the structure of a possible network quality identification device provided in an embodiment of this application. This network quality identification device can be used to implement the server functions described in the above method embodiments.

[0092] like Figure 4 As shown, the network quality identification device 400 includes a communication unit 410 and a processing unit 420.

[0093] When the network quality identification device 400 is used to achieve Figure 1 The server function in the method embodiment shown is as follows:

[0094] The communication unit 410 is used to acquire network indicator data for the first time period.

[0095] The processing unit 420 is used to input the network indicator data of the first time period into the first model to obtain the network quality identification result output by the first model, wherein the network quality identification result is used to characterize the network quality situation of the first time period; if the network quality of the first time period is abnormal, the network quality identification result is also used to determine the application type that may have network quality abnormalities.

[0096] In one possible implementation, the application type that may experience network anomalies includes one of video playback, games, web pages, or instant messaging.

[0097] In one possible implementation, the network quality identification result includes a network quality score; if the network quality score falls within a first value range, the network quality identification result indicates that the network quality is normal during the first time period; if the network quality score falls within a second value range, the network quality identification result indicates that the network quality is abnormal during the first time period.

[0098] In one possible implementation, the second value range includes K sub-value ranges; the K sub-value ranges correspond one-to-one with K application types, where K is a positive integer; if the network quality score belongs to the second value range and the network quality score belongs to the i-th sub-value range, then the network quality identification result indicates that the application type that may have network quality anomalies is the application type corresponding to the i-th sub-value range, where the i-th sub-value range is one of the K sub-value ranges, i is a positive integer, and i≤K.

[0099] In one possible implementation, the processing unit 420 is configured to determine the first value range, the second value range, and the K sub-value ranges based on the first network indicator dataset and the first user experience dataset. The first network indicator dataset is associated with the first user experience dataset, which includes historical network indicator data for M time periods. The first user experience dataset includes user experience data for the M time periods. Each time period's user experience data includes the network quality status for that time period. For time periods with abnormal network quality, the user experience data also includes the application type with abnormal network quality. M is a positive integer.

[0100] In one possible implementation, when determining the first value range and the second value range based on the first network indicator dataset and the first user experience dataset, the processing unit 420 is configured to input the first network indicator dataset into the first model to obtain the network quality scores for the M time periods; and to obtain the first value range and the second value range based on the network quality scores for the M time periods and the user experience data for each time period, including the network quality situation of the corresponding time period.

[0101] In one possible implementation, when determining the K sub-value ranges based on the first network indicator dataset and the first user experience dataset, the processing unit 420 is further configured to determine the K sub-value ranges based on the user experience data of the time period in which the network quality situation is abnormal, and the network quality score of the time period in which the network quality situation is abnormal, wherein the network quality score of the time period in which the network quality situation is abnormal belongs to the network quality scores of the M time periods.

[0102] In one possible implementation, if the network quality identification result indicates network quality anomalies in the first time period, the communication unit 410 is further configured to obtain user feedback results in the first time period, wherein the user feedback results indicate a first application type; if the application type that may experience network quality anomalies is different from the first application type, the processing unit 420 is further configured to add the network indicator data of the first time period to the first network indicator dataset.

[0103] In one possible implementation, if the network quality identification result indicates that the network quality is abnormal in the first time period, the processing unit 420 is further configured to determine the cause of the network quality abnormality in the first time period based on the network quality identification result and the network indicator data of the first time period.

[0104] In one possible implementation, the processing unit 420 is further configured to train an isolated forest model based on a second network indicator dataset to obtain the first model; wherein the second network indicator dataset includes historical network indicator data for N time periods, where N is a positive integer.

[0105] In one possible implementation, the first model is constructed based on the isolated forest model.

[0106] In one possible implementation, the network metric data for the first time period includes at least one of channel utilization, signal-to-noise ratio, signal strength, noise floor, number of uplink retransmission packets, number of downlink retransmission packets, uplink retransmission rate, downlink retransmission rate, number of uplink packet loss, number of downlink packet loss, uplink packet loss rate, downlink packet loss rate, and total packet loss rate.

[0107] For a more detailed description of the communication unit 410 and the processing unit 420, please refer to [link / reference needed]. Figure 1 The relevant descriptions in the method embodiments shown are directly obtained and will not be repeated here.

[0108] like Figure 5 As shown, the network quality identification device 500 includes a processor 510 and a communication interface 520. The processor 510 and the communication interface 520 are coupled to each other. It is understood that the communication interface 520 can be a transceiver or an input / output interface. Optionally, the network quality identification device 500 may also include a memory 530 for storing instructions executed by the processor 510, or storing input data required by the processor 510 to execute instructions, or storing data generated after the processor 510 executes instructions.

[0109] When the network quality identification device 400 is used to achieve Figure 1In the method shown, the processor 510 is used to implement the functions of the processing unit 420, and the communication interface 520 is used to implement the functions of the communication unit 410.

[0110] The unit division in this embodiment is illustrative and represents only one logical functional division. In actual implementation, other division methods may be used. Furthermore, the functional units in the various embodiments of this application can be integrated into a single processor, exist as separate physical units, or be integrated into a single unit. The integrated units described above can be implemented in hardware or as software functional units.

[0111] Based on the same inventive concept, embodiments of this application provide a computer-readable storage medium. The computer program product includes computer program code, which, when executed on a computer, causes the computer to perform any of the network quality identification methods discussed above. Since the principle by which the above-described computer-readable storage medium solves the problem is similar to that of the network quality identification method, the implementation of the above-described computer-readable storage medium can be found in the implementation of the method; repeated details will not be elaborated further.

[0112] Based on the same inventive concept, this application also provides a computer program product, which includes computer program code. When the computer program code is run on a computer, it causes the computer to execute any of the network quality identification methods discussed above. Since the principle by which the above-described computer program product solves the problem is similar to that of the network quality identification method, the implementation of the above-described computer program product can be referred to the implementation of the method, and repeated details will not be described again.

[0113] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0114] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0115] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0116] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A network quality identification method characterized by, The method includes: Obtain network metric data for the first time period; The network indicator data of the first time period is input into the first model to obtain the network quality identification result output by the first model. The network quality identification result is used to characterize the network quality situation of the first time period. If the network quality of the first time period is abnormal, the network quality identification result is also used to determine the application type that may have network quality abnormalities.

2. The method as described in claim 1, characterized in that, The types of applications that may experience network quality anomalies include video playback, games, web pages, or instant messaging.

3. The method as described in claim 1, characterized in that, The network quality identification results include a network quality score; If the network quality score falls within the first value range, the network quality identification result indicates that the network quality is normal during the first time period. If the network quality score falls within the second value range, the network quality identification result indicates that the network quality is abnormal during the first time period.

4. The method as described in claim 3, characterized in that, The second value range includes K sub-value ranges; each of the K sub-value ranges corresponds one-to-one with a K application type, where K is a positive integer; If the network quality score belongs to the second value range and the network quality score belongs to the i-th sub-value range, then the network quality identification result indicates that the application type that may have network quality anomalies is the application type corresponding to the i-th sub-value range, where the i-th sub-value range is one of the K sub-value ranges, i is a positive integer, and i≤K.

5. The method as described in claim 4, characterized in that, Also includes: The first value range, the second value range, and the K sub-value ranges are determined based on the first network indicator dataset and the first user experience dataset. The first network indicator dataset is associated with the first user experience dataset. The first network indicator dataset includes historical network indicator data for M time periods. The first user experience dataset includes user experience data for the M time periods. The user experience data for each time period includes the network quality status for that time period. The user experience data for time periods with abnormal network quality also includes the application type with abnormal network quality. M is a positive integer.

6. The method as described in claim 5, characterized in that, The first value range and the second value range are determined based on the first network metric dataset and the first user experience dataset, including: Input the first network metric dataset into the first model to obtain the network quality scores for the M time periods; Based on the network quality scores for the M time periods and the user experience data for each time period, including the network quality situation for the corresponding time period, the first value range and the second value range are obtained.

7. The method as described in claim 6, characterized in that, The K sub-value ranges are determined based on the first network metric dataset and the first user experience dataset, including: The K sub-value ranges are determined based on the user experience data of the time period when the network quality is abnormal, and the network quality score of the time period when the network quality is abnormal. The network quality score of the time period when the network quality is abnormal belongs to the network quality score of the M time periods.

8. The method as described in claim 5, characterized in that, The method further includes: If the network quality identification result indicates that the network quality is abnormal in the first time period, obtain the user feedback result for the first time period, and the user feedback result indicates the first application type; If the application type that may experience network quality anomalies is different from the first application type, the network indicator data for the first time period is added to the first network indicator dataset.

9. The method as described in claim 3, characterized in that, The method further includes: If the network quality identification result indicates that the network quality is abnormal in the first time period, the cause of the network quality abnormality in the first time period is determined based on the network quality identification result and the network indicator data of the first time period.

10. The method according to any one of claims 1 to 9, characterized in that, The method further includes: The first model is obtained by training an isolated forest model based on the second network metric dataset. The second network indicator dataset includes historical network indicator data for N time periods, where N is a positive integer.

11. The method as described in claim 1, characterized in that, The network metrics data for the first time period include at least one of the following: channel utilization, signal-to-noise ratio, signal strength, noise floor, number of uplink retransmission packets, number of downlink retransmission packets, uplink retransmission rate, downlink retransmission rate, number of uplink packet loss, number of downlink packet loss, uplink packet loss rate, downlink packet loss rate, and total packet loss rate.

12. A network quality identification device, characterized in that, The device includes: a communication unit and a processing unit; The communication unit is used to acquire network indicator data for the first time period; The processing unit is used to input network indicator data of the first time period into the first model to obtain the network quality identification result output by the first model, wherein the network quality identification result is used to characterize the network quality situation of the first time period; if the network quality of the first time period is abnormal, the network quality identification result is also used to determine the application type that may have network quality abnormalities.

13. A network quality identification device, characterized in that, include: Memory, used to store program instructions; A processor is configured to invoke program instructions stored in the memory and execute the steps of the method according to any one of claims 1-11 in accordance with the obtained program instructions.

14. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, the computer program including program instructions that, when executed by a computer, cause the computer to perform the method as described in any one of claims 1-11.

15. A computer program product, characterized in that, The computer program product includes: computer program code, which, when run on a computer, causes the computer to perform the method described in any one of claims 1-11.

Images