A root cause analysis method for poor user video quality, electronic device and storage medium

By combining set-top box and gateway probe data, a knowledge base is built, and the problem of difficulty in positioning fault points and relying on experience diagnosis in user video quality root cause analysis is solved, and a multi-faceted analysis and diagnosis of user video quality root cause is realized. Improved troubleshooting efficiency and accuracy.

CN114997560BActive Publication Date: 2024-11-01BEIJING ZZNODE TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210372229.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-04-11
Publication Date
2024-11-01
Estimated Expiration
2042-04-11

AI Technical Summary

Technical Problem

The existing technology has problems in the analysis of poor user video quality, such as difficulty in positioning fault points, easy errors based on experience diagnosis, high pressure on data processing, and inability to effectively improve network quality.

Method used

By combining set-top box probe data and gateway probe data, data preprocessing, association rule analysis and clustering algorithms are used to build a knowledge base to realize multi-faceted analysis and diagnosis of the root causes of poor video quality of users, reduce data storage pressure, and improve sequencing Barrier efficiency.

Benefits of technology

It realizes accurate location of home LAN fault points, improves troubleshooting efficiency, effectively refines knowledge, supports a variety of diagnostic methods, and continuously improves as data increases.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114997560B_ABST
    Figure CN114997560B_ABST
Patent Text Reader

Abstract

A method for analyzing the root causes of poor video quality of users, an electronic device and a storage medium. In the process of troubleshooting the root causes of poor quality of Internet TV services, by combining set-top box probe data and gateway probe data, the fault point in a home local area network can be more accurately located, the troubleshooting efficiency can be improved, historical data can be effectively processed, the pressure of data storage can be reduced, knowledge can be effectively extracted, and root cause analysis in multiple aspects can be performed at the same time.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to network management technology, in particular to a root cause analysis method for poor quality of user videos, electronic equipment and storage medium. Background Art

[0002] A large number of gateways are pre-installed with probes to monitor service quality in real time, including but not limited to hardware information, software information, service information, and startup information. Probes are pre-installed on set-top boxes to monitor video service quality in real time, including but not limited to playback time, freeze information, program information, network delay, etc. The reason for poor quality of Internet TV services is usually on the user side, and it is often difficult for operation and maintenance personnel to directly find the location of the problem when they visit. On the other hand, it is difficult to identify the fault point using only set-top box probes, and the problem is difficult to solve. The current solution mainly relies on set-top box probes. The set-top box probes detect poor quality and use the data collected by themselves to make judgments. However, the use of other devices in the home affects the operation of the set-top box, and the analysis is inevitably not comprehensive enough. If all set-top box data is uploaded to a unified platform, it is inevitable that too much pressure will be placed on the platform, and on the other hand, the storage and computing pressure will be too high. Another solution is to realize communication between the set-top box probe and the gateway probe on the user side to obtain home network information in real time. This method involves a lot of human experience in diagnosing the root cause, and it is easy to make mistakes. At the same time, the data is actually gradually lost as it disappears, which does not play a key role in improving the overall network quality.

[0003] There are many reasons for poor user video quality, including content source quality, core network quality, access network quality, and home network quality. The quality issues of the core network and access network are relatively easy to monitor, and the occurrence of degradation can be quickly identified through the monitoring of key performance indicators. However, it is difficult to identify the specific causes of problems on the user side due to the large number of users, complex practical scenarios, and diverse home broadband devices. This brings difficulties to the root cause analysis of poor quality. Real-time methods lack the accumulation of knowledge and rely too much on empirical conclusions. The problem of lack of empirical conclusions cannot be effectively solved and cannot face new challenges. On-site detection is affected by multiple factors such as time, location, and usage scenarios, and is difficult to implement. Accordingly, this application attempts to propose a method that can effectively process historical data, reduce data storage pressure, effectively extract knowledge, and at the same time perform root cause analysis in multiple aspects. Summary of the invention

[0004] The present invention aims at the defects or shortcomings in the prior art and provides a root cause analysis method for poor quality of user videos, an electronic device and a storage medium.

[0005] The technical solution of the present invention is as follows:

[0006] A method for analyzing root causes of poor user video quality, characterized in that it comprises the following steps:

[0007] Step S201, preprocessing the reported data to form valid data, wherein the reported data includes report-related data formed by collecting the operation data of the user video terminal;

[0008] Step S202, constructing a data set using the complaint data and the valid data;

[0009] Step S203, performing association rule analysis on the data in the data set construction and forming a knowledge base, the knowledge base includes data tables with the following column names: support, confidence, lift, reason, and result.

[0010] The terminal operation data in step S201 includes set-top box probe data and gateway probe data, and is uploaded periodically. The period is limited by a first threshold. In the process of troubleshooting the root cause of poor quality of Internet TV services, the combination of set-top box probe data and gateway probe data can more accurately locate the fault point in the home local area network and improve troubleshooting efficiency.

[0011] The terminal operation data includes user complaint data and access network resource tree data, and the terminal operation data is reported to the network server via TCP.

[0012] The data preprocessing includes filtering data, and the filtering data includes filtering out data without viewing behavior; the data set construction includes associating complaint data with terminal operation data, screening out user data with a time difference that meets a second threshold as a complaint data set, and using other data as a non-complaint data set, filling in incomplete data through an AP clustering algorithm, and marking the successfully filled data as a silent poor quality data set. So far, a total of three data sets are constructed, namely: a complaint data set, a silent poor quality data set, and a non-complaint data set; the association rule analysis includes constructing a user perception index for records in the complaint data set and the silent poor quality data set, and discretizing them, performing association rule analysis on the discretized data, and outputting association rule records as knowledge base data. The knowledge base also includes optimization suggestions for different root causes.

[0013] The method includes using a knowledge base to perform real-time diagnosis on user videos of user terminals: the user terminal actively reports set-top box probe data in real time, the real-time data is discretized, compared with the knowledge base, and root cause analysis results and optimization suggestions are returned; or, by remotely dialing the user terminal, a collection command is issued, real-time data is uploaded, compared with the knowledge base, and root cause analysis results and optimization suggestions are returned; or, the operator performs periodic dialing of the user terminal to obtain the network quality status of the user terminal, compares with the knowledge base, and returns the root cause analysis results and optimization suggestions.

[0014] The data set construction in step S202 includes the following steps:

[0015] Step S2021, data association is performed on the resource tree, gateway probe data, set-top box probe data, and complaint data;

[0016] Step S2022, classifying the data to form a complaint data set and a non-complaint data set;

[0017] Step S2023, expanding the complaint data set and forming a silent poor quality data set;

[0018] Step S2024, constructing a user perception index using the complaint data set and the silent poor quality data set; Step S2025, discretizing the data and forming a correlation analysis data set.

[0019] The complaint data set expansion in step S2023 is implemented by using the AP clustering algorithm, including the following steps:

[0020] Step S20231, algorithm initialization, including initialization of matrices R, A, S;

[0021] Step S20232, update the attraction matrix R;

[0022] Step S20233, updating the attribution matrix A;

[0023] Step S20234, performing an attenuation operation according to the attenuation coefficient;

[0024] Step S20235, determine whether the matrix is ​​stable or the maximum number of iterations has been reached, if not, return to step S20232, if yes, proceed to step S20236;

[0025] Step S20236, selecting the point with the largest A+R as the cluster center;

[0026] Step S20237, exporting a silent poor quality data set;

[0027] Step S20238, update the non-complaint data set.

[0028] An application network using the above-mentioned root cause analysis method for poor user video quality is characterized by comprising a user terminal, a network server, a database, and a knowledge base, which jointly perform the following steps:

[0029] Step S401, data reporting;

[0030] Step S402, analyzing historical data;

[0031] Step S403, knowledge base assists in real-time diagnosis and optimization;

[0032] Step S401 includes the following steps:

[0033] Step S4011, the user terminal is turned on;

[0034] Step S4012, reporting user terminal startup data to the network server;

[0035] Step S4013, the network server stores the user terminal startup data in the database;

[0036] Step S4014, the user terminal periodically reports data to the network server;

[0037] Step S4015, the network server stores the periodic reporting data in a database;

[0038] Step S402 includes the following steps:

[0039] Step S4021, the network server extracts historical data from the database;

[0040] Step S4022, the network server performs association rule analysis to form a knowledge base;

[0041] Step S4023, the network server updates and maintains the knowledge base;

[0042] Step S403 includes the following steps:

[0043] Step S4031, the user terminal reports real-time measurement data to the network server to assist diagnosis;

[0044] Step S4032, the network server extracts relevant knowledge from the knowledge base;

[0045] Step S4033, the network server calculates the reason for the poor quality;

[0046] Step S4034, the network server sends the diagnosis result and optimization plan to the user terminal;

[0047] Step S4035, the user terminal feeds back the optimization effect to the network server;

[0048] Step S4036, the network server updates and maintains the knowledge base.

[0049] An electronic device, characterized in that it includes: a memory, a processor, a receiver, a display and a computer program, wherein the computer program is stored in the memory, and the processor runs the computer program to execute the above-mentioned user video quality poor root cause analysis method.

[0050] A storage medium, characterized in that it comprises: a readable storage medium and a computer program stored in the readable storage medium, wherein the computer program is used to implement the above-mentioned method for analyzing the root causes of poor user video quality.

[0051] The technical effects of the present invention are as follows: the present invention provides a method for analyzing the root causes of poor video quality of users, an electronic device and a storage medium. In the process of troubleshooting the root causes of poor quality of Internet TV services, the combination of set-top box probe data and gateway probe data can more accurately locate the fault point in the home local area network, improve troubleshooting efficiency, effectively process historical data, reduce data storage pressure, effectively extract knowledge, and simultaneously perform root cause analysis in multiple aspects.

[0052] The features of the present invention are as follows: 1. Offline analysis. This application summarizes multiple root causes by analyzing historical data and deriving a knowledge base to achieve offline analysis. 2. Online real-time query. By uploading set-top box probe data in real time, real-time diagnosis is performed, and the diagnosis is accurate, fast, and more efficient. 3. Multiple diagnostic methods can be performed simultaneously. It supports multiple diagnostic methods such as active diagnosis of user terminals, active diagnosis of installation and maintenance personnel, and periodic diagnosis of operators. 4. It supports updating and maintenance of the knowledge base. With multiple update and maintenance methods, the accuracy continues to improve as the data increases. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 The present invention is a schematic diagram of a data collection and reporting process involved in a method for analyzing root causes of poor video quality of a user. Figure 1 The process includes step S101, collecting terminal operation data; and step S102, reporting relevant data.

[0054] Figure 2 The present invention is a flowchart of a method for analyzing the root causes of poor video quality of a user. Figure 2 The method includes step S201, preprocessing the reported data and forming valid data; step S202, constructing a data set using the complaint data and the valid data; and step S203, performing association rule analysis and forming a knowledge base (or knowledge database).

[0055] Figure 3 yes Figure 2 Schematic diagram of the data preprocessing process in step S201. Figure 3 The method includes step S2011, filtering set-top box probe data; step S2012, associating the set-top box probe data and forming valid data; step S2013, filtering gateway probe data; and step S2014, associating the gateway probe data and forming valid data.

[0056] Figure 4 yes Figure 2 Schematic diagram of the data set construction process of step S202. Figure 4 The method includes step S2021, associating the resource tree, gateway probe data, set-top box probe data and complaint data; step S2022, classifying the data to form a complaint data set and a non-complaint data set respectively; step S2023, expanding the complaint data set and forming a silent poor quality data set; step S2024, constructing a user perception index using the complaint data set and the silent poor quality data set; and step S2025, discretizing the data and forming an association analysis data set.

[0057] Figure 5 yes Figure 4 Schematic diagram of the process of expanding the complaint data set in step S2023 using the AP clustering algorithm. Figure 5 The process includes step S20231, algorithm initialization, including initialization matrices R, A, and S; step S20232, updating the attraction matrix R; step S20233, updating the attribution matrix A; step S20234, performing attenuation operation according to the attenuation coefficient; step S20235, judging whether the matrix is ​​stable or the maximum number of iterations has been reached, if not, returning to step S20232, if yes, proceeding to step S20236; step S20236, selecting the point with the largest A+R as the cluster center; step S20237, exporting the silent poor quality data set; step S20238, updating the non-complaint data set.

[0058] Figure 6 This is a flowchart of real-time diagnosis of the root cause of poor video quality for users. Figure 6 The method includes step S301, the user terminal sends a request; step S302, calculating the user experience index; step S303, discretizing the reported data; step S304, querying the knowledge base; step S305, returning the root cause and returning relevant suggestions.

[0059] Figure 7 The present invention is a schematic diagram of the overall workflow of a method for analyzing root causes of poor user video quality based on network application. Figure 7 It includes user terminals, network servers, databases, and knowledge bases; Figure 7The process includes step S401, data reporting; step S402, analyzing historical data; step S403, knowledge base assisting real-time diagnosis and optimization. Step S401 includes step S4011, user terminal booting; step S4012, reporting user terminal booting data to the network server; step S4013, the network server stores the user terminal booting data in the database; step S4014, the user terminal periodically reports data to the network server; step S4015, the network server stores the periodically reported data in the database. Step S402 includes step S4021, the network server extracts historical data from the database; step S4022, the network server performs association rule analysis to form a knowledge base; step S4023, the network server updates and maintains the knowledge base. Step S403 includes step S4031, the user terminal reports real-time measurement data to the network server to assist in diagnosis; step S4032, the network server extracts relevant knowledge from the knowledge base; step S4033, the network server calculates the cause of poor quality; step S4034, the network server sends the diagnosis results and optimization plan to the user terminal; step S4035, the user terminal feeds back the optimization effect to the network server; step S4036, the network server updates and maintains the knowledge base.

[0060] Figure 8 This is a schematic diagram of a knowledge base data sample. Figure 8 The first column is the sequence number in the table, the second column is the support, the third column is the confidence, the fourth column is the lift, the fifth column is the reason or root cause, and the sixth column is the result. DETAILED DESCRIPTION

[0061] Below is the attached figure ( Figure 1-Figure 8 ) and Examples illustrate the present invention.

[0062] The present invention proposes a method and device for analyzing the root cause of poor video quality, a user terminal, and a network server. In the process of troubleshooting the root cause of poor quality of Internet TV services, the combination of set-top box probe data and gateway probe data can more accurately locate the fault point in the home local area network and improve troubleshooting efficiency. In order to solve the above problems, the embodiments of the present invention provide the following technical solutions:

[0063] A data feedback device is located in a user terminal, comprising: collecting terminal operation data, including set-top box probe data and gateway probe data, and periodically uploading the data, wherein the period is limited by a first threshold.

[0064] Report terminal operation data to the server via TCP.

[0065] In particular, it should also include user complaint data and access network resource tree data.

[0066] A method for analyzing the root cause of poor video quality of users, used in a network server, comprising: preprocessing the terminal operation data, filtering the data, and filtering the data without viewing behavior. Associating the complaint data with the terminal operation data, screening out the user data whose time difference meets the second threshold as the complaint data set, and taking the other data as the non-complaint data set. Filling the incomplete data through the AP clustering algorithm, marking the successfully filled data as the silent poor quality data set. So far, three data sets are constituted, namely: complaint data set, silent poor quality data set, and non-complaint data set.

[0067] The user perception index is constructed for the records in the complaint data set and the silent quality-poor data set, and discretized. The discretized data is analyzed for association rules, and the association rule records are output as a knowledge base.

[0068] In particular, the knowledge base should also include optimization suggestions for different root causes.

[0069] A real-time diagnosis method for a user terminal includes: the user terminal actively reports set-top box probe data in real time, discretizes the real-time data, compares it with a knowledge base, and returns root cause analysis results and optimization suggestions. Alternatively, by remotely dialing the user terminal, issuing a collection command, and uploading real-time data for root cause analysis. Alternatively, the operator periodically dials the user terminal to obtain the network quality status of the user terminal.

[0070] Figure 1 The process of Embodiment 1 is included, and the specific steps are as follows: S101: collecting terminal operation data. The terminal operation data includes set-top box probe data and gateway probe data.

[0071] Furthermore, the set-top box probe data includes set-top box startup data, set-top box hardware data, set-top box software data, set-top box network data, set-top box service data, and set-top box operation data.

[0072] The set-top box startup data includes: the set-top box startup time and the set-top box last operation time.

[0073] The set-top box hardware data includes: set-top box manufacturer and licensee.

[0074] The set-top box software data includes: probe provider, probe version, system version information, set-top box firmware version, and CPU model.

[0075] The set-top box network data includes: network connection mode, m3u8 file request success rate, media file request success rate, EPG request success rate, m3u8 file request response average delay, media file request response average delay, EPG request response average delay, TCP connection success rate, TCP connection average duration, TCP average retransmission rate, if the network connection mode is Wifi, it should also include: the bandwidth of wireless hotspots, wireless network signal strength. Wherein EPG (Electronic Program Guide) is an electronic program menu. M3u8 is a file format, and TCP is a transmission control protocol.

[0076] The set-top box service data includes: an installed application package name list, a watched program list, an accessed TCP resource list, an m3u8 request failure list, a media file request failure list, an EPG request failure list, and a TCP request failure list.

[0077] The set-top box operation data includes: user viewing time, number of freezes, total freeze duration, continuous operation time of the set-top box, whether the licensee's application is running, CPU occupancy, and memory usage.

[0078] Furthermore, the gateway probe data includes gateway hardware data, gateway software data, gateway network data, gateway service data, gateway operation data, and gateway downstream device data.

[0079] The gateway hardware data includes: CPU model, hardware version number, gateway model, received optical power, and transmitted optical power.

[0080] The gateway software data includes: software version number, soft probe middleware version number, and interface protocol version number.

[0081] The gateway network data includes: TCP connection success rate, TCP retransmission rate, HTTP response delay, HTTP download rate, and HTTP request success rate.

[0082] The gateway service data includes: uplink traffic period average, downlink traffic period average, uplink traffic, downlink traffic, uplink peak rate, downlink peak rate, protocol type, service type, number of surrounding WiFi, and WiFi channel.

[0083] The gateway operation data includes: gateway operation time, CPU occupancy, and memory occupancy.

[0084] The gateway downstream device data includes: the number of downstream devices, the MAC address list of downstream devices, the periodic average list of upstream traffic of downstream devices, the periodic average list of downstream traffic of downstream devices, the upstream traffic list, and the downstream traffic list. The downstream devices are a group of data, and the number is determined by the number of downstream devices.

[0085] In particular, the set-top box probe data should also include a set-top box account, a set-top box unique identifier, and a set-top box MAC address. The gateway probe data should also include a gateway account and a gateway unique identifier.

[0086] S102: Upload relevant data. The gateway probe data and the set-top box probe data are reported in TCP mode.

[0087] Furthermore, the set-top box startup data, set-top box hardware information, and set-top box software data are uploaded once after startup; the set-top box network data, set-top box service data, and set-top box operation data are reported periodically.

[0088] Furthermore, the gateway hardware information and gateway software data are uploaded once after startup; the gateway network data, gateway service data, gateway operation data, and gateway downstream device data are reported periodically.

[0089] Furthermore, the required data also needs to include network alarm data and user complaint data.

[0090] Furthermore, the collection period of the gateway data and the set-top box data should be consistent.

[0091] Furthermore, the collection period of the periodically reported data is a first threshold value. In this embodiment, the first threshold value is set to 15 minutes.

[0092] Furthermore, it should also include access network resource tree data.

[0093] Figure 2 The process of Example 2 is included, and the specific steps are as follows: S201: Data preprocessing. The content of the reported data is as in Example 1. The processing process of the reported data is as in Example 1. Figure 3 shown.

[0094] S2011: Filter the set-top box probe data. The filtering rules are: data with user viewing time less than or equal to 0, and unknown network connection mode. Filter out the data that meets any of the filtering rules.

[0095] S2012: Associating set-top box data. Associating set-top box startup data, set-top box hardware data, set-top box software data, set-top box network data, set-top box service data, and set-top box operation data, using the first threshold as the granularity, and constructing a wide table.

[0096] After association, the set-top box probe data is in the following format:

[0097] Y STB ={y 1 ,y 2 ,···,y 34}

[0098] Y STB It contains 34 fields in total, in order: set-top box manufacturer, licensee, probe provider, probe version, system version information, set-top box firmware version, CPU model, whether the licensee's application is running, set-top box startup time, set-top box last running time, m3u8 file request success rate, media file request success rate, EPG request success rate, m3u8 file request response average delay, media file request response average delay, EPG request response average delay, TCP connection success rate, TCP connection establishment average time, TCP average retransmission rate, user viewing time, number of freezes, total freeze time, set-top box continuous operation time, CPU occupancy, memory usage, installed application package name list, viewing program list, access to TCP resource list, m3u8 request failure list, media file request failure list, EPG request failure list, TCP request failure list, network connection method, wireless network signal strength.

[0099] S2013: Filter the gateway probe data. The filtering rules are: the downstream device is empty, the data with empty value,

[0100] S2014: Associating gateway probe data. Associating gateway hardware data, gateway software data, gateway network data, gateway service data, gateway operation data, and gateway attached device data.

[0101] After the gateway probe data is associated, it forms the following format:

[0102] X HGU ={x 1 , x 2 ,···,x 32}

[0103] X HGU It contains 32 fields in total, in order: CPU model, hardware version number, gateway model, software version number, soft probe middleware version number, interface protocol version number, protocol type, service type, WiFi channel, received optical power, transmitted optical power, TCP connection success rate, TCP retransmission rate, HTTP response delay, HTTP download rate, HTTP request success rate, upstream traffic cycle average, downstream traffic cycle average, upstream traffic, downstream traffic, upstream peak rate, downstream peak rate, gateway running time, CPU occupancy, memory occupancy, number of surrounding WiFi, number of downstream devices, MAC address list of downstream devices, upstream traffic cycle average list of downstream devices, downstream traffic cycle average list, upstream traffic list, downstream traffic list.

[0104] S202: Dataset construction. The process of data set construction is as follows Figure 4 shown.

[0105] S2021: Data association. Associate the gateway probe data, set-top box probe data, and complaint data. In particular, since the unique user identifiers used by the gateway probe data, set-top box probe data, and complaint data are different, resource tree data is required to participate in the association. The resource tree data includes at least: the user's gateway unique identifier, set-top box unique identifier, and complaint account unique identifier. After data association, any record L has the following data format:

[0106] L = {X HGU , Y STB , Z TS}

[0107] In particular, X IHGU As described in S2014, YSTB as described in S2012, where Z TS There are three cases, as shown below.

[0108]

[0109] S2022: Data classification. Extract data with complaint behavior. The complaint behavior is defined as the time difference between the complaint event and the occurrence of the set-top box viewing behavior is less than the second threshold. In Example 2, the second threshold is 120 hours. The second threshold is used to control the delay of user complaint behavior. The data with complaint behavior is used as the complaint data set. The other data is used as the non-complaint data set.

[0110] S2023: Complaint dataset expansion. The dataset expansion is implemented using the AP clustering algorithm. The process is as follows: Figure 5 shown.

[0111] S20231: Algorithm initialization. Use the existing complaint data set as the initial centroid, which can also be called the cluster center point. Initialize the attraction matrix R and the attribution matrix A to 0 matrices. Initialize the similarity matrix S. The similarity matrix is ​​initialized as follows:

[0112] The data described in the first embodiment has multiple types. In order to calculate the similarity of different types of data, the data types need to be classified. The similarity calculation uses normalized data:

[0113] (1) Set-top box probe data is divided into four categories:

[0114] The first data type includes: set-top box manufacturer, licensee, probe provider, probe version, system version information, set-top box firmware version, CPU model, and whether the licensee's application is running.

[0115] The second data type includes: the length of time the set-top box is turned on, the length of time the set-top box was last running, the success rate of m3u8 file requests, the success rate of media file requests, the success rate of EPG requests, the average delay in responding to m3u8 file requests, the average delay in responding to media file requests, the average delay in responding to EPG requests, the success rate of TCP connection, the average time it takes to establish a TCP connection, the average TCP retransmission rate, the length of time the user watches the video, the number of freezes, the total length of time the set-top box is running continuously, the CPU occupancy rate, and the memory usage rate.

[0116] The third data type includes: installed application package name list, watched program list, accessed TCP resource list, m3u8 request failure list, media file request failure list, EPG request failure list, and TCP request failure list.

[0117] The fourth data type includes: network connection mode and wireless network signal strength.

[0118] (2) Gateway probe data is divided into four categories:

[0119] The fifth data type includes: CPU model, hardware version number, gateway model, software version number, soft probe middleware version number, interface protocol version number, protocol type, service type, and WiFi channel.

[0120] The sixth data type includes: received optical power, transmitted optical power, TCP connection success rate, TCP retransmission rate, HTTP response delay, HTTP download rate, HTTP request success rate, upstream traffic cycle average, downstream traffic cycle average, upstream traffic, downstream traffic, upstream peak rate, downstream peak rate, gateway operating time, CPU occupancy, and memory occupancy.

[0121] The seventh data type includes: the number of surrounding WiFi networks.

[0122] The eighth data type includes: the number of downstream devices, the MAC address list of downstream devices, the periodic average list of upstream traffic of downstream devices, the periodic average list of downstream traffic of downstream devices, the upstream traffic list, and the downstream traffic list.

[0123] (3) For any first data sample A and any second data sample B in the data set, the similarity calculation method of the first data type to the eighth data type is:

[0124] The method for calculating the similarity of the first data type is as follows:

[0125]

[0126] Wherein my is the first weight. In this embodiment, my={0.2, 0.1, 0.2, 0.1, 0.1, 0.1, 0.1, 0.1}, and f(a, b) is calculated as follows:

[0127]

[0128] The method for calculating the similarity of the second data type is as follows:

[0129]

[0130] The method for calculating the similarity of the third data type is as follows:

[0131]

[0132] Each quantity in the third data type is in the form of a list, num() means calculating the number of items in the list, ∪ means finding the union of two lists, ∩ means finding the intersection of two lists, w is the second weight, and in this embodiment w = (0.08, 0.08, 0.08, 0.19, 0.19, 0.19, 0.19).

[0133] The fourth data type calculates the similarity Dis 4 The method is: If the network connection methods are the same and both are not Wifi, Dis 4 =0; if the network connection methods are all non-Wifi and different, then Dis 4 =1; if the network connection mode is Wifi, then Dis 4 The two are defined as the absolute value of the difference in wireless network signal strength.

[0134] The method for calculating the similarity of the fifth data type is as follows:

[0135]

[0136] Where m x is the third weight. In this embodiment, m x ={0.05, 0.05, 0.1, 0.05, 0.05, 0.2, 0.2, 0.1, 0.2}, the calculation method of f(a, b) is the same as the first data type.

[0137] The method for calculating the similarity of the sixth data type is as follows:

[0138]

[0139] The seventh data type calculates the similarity Dis 7 The method is: if the network connection methods in the set-top box data are all non-WIFI methods, Dis 7=0; if the network connection mode in the set-top box data is WIFI, then the calculation method Dis 7 =|x Ai -x Bi |, i = 26; if one is WIFI mode and the other is non-WIFI mode, how to Dis 7 =1.

[0140] The eighth data type calculates the similarity as follows:

[0141] Use the MAC address of the downstream device to match the set-top box, which is the kth in the MAC address list of the gateway downstream device. The similarity is expressed as follows:

[0142]

[0143] (4) The value of its similarity matrix S is:

[0144] S AB =-(1+Dis 1 +Dis 5 )×(1+Dis 2 +Dis 4 +Dis 7 +Dis 8 )×(Dis 3 +Dis 6 )

[0145] S20232: Update the attraction matrix. The update formula is as follows:

[0146]

[0147] S20233: Update the attribution matrix. The update formula is as follows:

[0148]

[0149] S20234: Perform an attenuation operation according to the attenuation coefficient. The attenuation coefficient λ is a first adjustable parameter. In this embodiment, the first adjustable parameter λ=0.5. The formula for the attenuation operation is as follows:

[0150] R t+1 (i, k) = λR t (i,k)+(1-λ)R t+1 (i, k)

[0151] A t+1 (i, k) = λA t (i,k)+(1-λ)A t+1 (i, k)

[0152] S20235: Repeat steps S20232 to S20234 until the matrix is ​​stable or the maximum number of iterations is reached, and the algorithm ends.

[0153] S20236: Select the point with the largest A+R as the cluster center.

[0154] S20237: Export the silent poor quality data set. Export the data that meets the following conditions as the silent poor quality data set. Export condition: The cluster contains at least one complaint user data, and the other non-complaint data is exported as the silent poor quality data set.

[0155] S20238: Update the non-complaint data set. Delete the data marked as silent poor quality in the non-complaint data set.

[0156] So far, three data sets are formed: complaint data set, non-complaint data set, and silent poor quality data set.

[0157] S2024: Construct a user experience index. In order to better describe the actual user experience, a user experience index is constructed. After adding the user experience index, each first record in the data set is in the following format:

[0158] L new ={X HGU , Y STB , Z TS , T UE}

[0159] The user experience indicators include: freeze ratio, freeze frequency, network score, resource request score, and resource request success rate.

[0160] The calculation method of the freeze ratio is: the total freeze duration divided by the user's viewing time,

[0161] The freeze frequency is calculated as follows: the number of freezes is divided by the user's viewing time, and the user's viewing time needs to be converted into hours.

[0162] The network score is calculated as follows: the average value of the average duration of TCP connection establishment, the average TCP retransmission rate, and the TCP connection failure rate (1-TCP connection success rate) in the set-top box probe data.

[0163] The resource request score is calculated as follows: the average of the average delay of m3u8 file request response, the average delay of media file request response, and the average delay of EPG request response in the set-top box probe data.

[0164] The resource request success rate is calculated as follows: the average value of the m3u8 file request success rate, the media file request success rate, and the EPG request success rate in the set-top box probe data.

[0165] S2025: Data discretization. Association rule analysis is performed using the complaint data set and the silent quality-poor data set. Discretization is required before association rule analysis. Discretization uses equidistance discretization, and the continuous data type is discretized into 5 parts.

[0166] S203: Association rule analysis. Association analysis can be performed by a variety of algorithms. This application uses the Apriori algorithm. The steps of association rules are as follows:

[0167] S2031: When scanning the transaction data set for the kth time, a frequent k-item set is generated.

[0168] S2032: Clear candidate sets that do not meet the condition. The condition is that the minimum support count, that is, the number of times the item set appears, is controlled by a third threshold. In this embodiment, the third threshold is 10,000.

[0169] S2033: Repeat S2031 until no more frequent itemsets can be generated. At this point, the generation of frequent itemsets is complete.

[0170] S2034: Traverse the item set combinations in the frequent item set to generate a rule set.

[0171] S2035: Calculate the relevant indicators of the rule set, including: support Sup, confidence Cof and lift Lift.

[0172] S2036: Filter the lift to filter out rules whose support count SupNum is greater than the third threshold, confidence Cof is greater than the fourth threshold, and lift Lift is greater than the fifth threshold, and the rules contain any field of the user perception index. Update the rules that meet the conditions into the knowledge database. In this embodiment, the fourth threshold is 0.4 and the fifth threshold is 2.

[0173] Each rule in the knowledge database consists of five parts: root cause of poor quality, user perception index, support, confidence, and improvement.

[0174] Furthermore, the root causes in the knowledge database are divided into four types: set-top box hardware problems, set-top box software problems, content source problems, and network problems.

[0175] The set-top box hardware problems are mainly reflected in long-term high CPU and memory occupancy, hardware aging, etc.

[0176] The set-top box software problems are mainly reflected in the general poor quality of software, firmware, system and probe versions.

[0177] The content source problem is mainly reflected in the fact that the user's network quality is excellent, but the quality of a specific content source is poor.

[0178] The network problems mentioned above mainly refer to the problems of too many devices connected to the home gateway and insufficient bandwidth.

[0179] Furthermore, the knowledge database should also include the operations and maintenance personnel's solutions to various root causes.

[0180] Example 3 shows how to perform real-time diagnosis. The process of Example 3 is as follows Figure 6 shown.

[0181] S301: The user terminal sends a request to report the set-top box probe data of the most recent minute.

[0182] S302: Calculate the user experience index.

[0183] S303: Discretize the reported data. Discretize the reported data according to the steps described in S2025.

[0184] S304: Query the knowledge base. By comparing the discretized reported fields with the knowledge base, select several rules with high similarity to the root causes of poor quality and user experience index of the knowledge base rules. Sort the rules that meet the requirements by confidence.

[0185] The definition of high similarity is: the number of fields in the knowledge base rules that are the same as the reported data fields, divided by the number of knowledge base rule fields.

[0186] S305: Return the root cause and related suggestions.

[0187] In particular, real-time diagnosis can be actively issued and performed by dialing the user terminal and issuing a command to report real-time data for diagnosis. This application does not restrict the initiator of the real-time diagnosis request.

[0188] In particular, operators can periodically perform real-time diagnosis on the network.

[0189] Embodiment 4, Figure 7 This is a flow chart of Example 4 of the present invention, which is a complete workflow of the present application. It is a comprehensive application of Example 1, Example 2, and Example 3 to demonstrate an overall workflow of the present invention. Example 4 is only a preferred embodiment of the present invention.

[0190] Embodiment 4 involves four topics: user terminal, network server, database, and knowledge base. The user terminal is as described in Embodiment 1. The network server is as described in Embodiment 2. The knowledge base is as described in Embodiment 2. The database is not a key part of the present invention and is not limited thereto.

[0191] The specific steps of Example 4 are as follows:

[0192] S401: Data reporting. Embodiment 4 is an implementation of the overall process, and its calculation method is the same as that described in Embodiment 1.

[0193] S4011: Turn on the terminal. Actively turn on the user terminal for the user.

[0194] S4012: Reporting user terminal startup data. The startup data is collected by the method described in the first embodiment, and the content of the startup data is as described in the first embodiment.

[0195] S4013: Data storage. The startup data is stored.

[0196] S4014: Periodically reporting data. The periodically reported data includes set-top box probe data and gateway probe data.

[0197] S4015: Data storage: The periodically reported data is stored.

[0198] S402: Analyze historical data. The historical data includes set-top box probe data and gateway probe data. In particular, the historical data should also include broadband user complaint data, and the complaint data should include information such as device number, complaint reason classification, etc.

[0199] S4021: Extract historical data. Extract required data from the database. The specific required data range is as described in the second embodiment.

[0200] S4022: Calculate the reasons for poor quality and form a knowledge base. The process of generating the knowledge base is as described in the second embodiment.

[0201] S4023: Update and maintain the knowledge base. The update and maintenance process includes: adding, deleting, and modifying. The specific process is as described in the second embodiment. Figure 8 This is a sample of knowledge base data.

[0202] S403: Knowledge base assists in real-time diagnosis and optimization. The "assisting real-time diagnosis and optimization" refers specifically to the following scenario: installation and maintenance personnel visit user terminals to detect problems, conduct real-time tests, determine network conditions, and provide corresponding solutions. Assisting real-time diagnosis and optimization is an application of the knowledge base described in Example 3. The "diagnosis" refers to determining the cause of poor user quality. The "optimization" refers to the process of improving the user experience by modifying specific parameters.

[0203] S4031: Report real-time measurement data to assist diagnosis. This process requires installation and maintenance personnel to operate. After the real-time measurement is performed, the real-time measurement data will be uploaded to the network server. The real-time measurement data includes: set-top box probe data. The range of the set-top box probe data is as described in Example 1.

[0204] S4032: Extract relevant knowledge from the knowledge base. Obtain relevant information from the knowledge base for diagnosis.

[0205] S4033: Calculate the quality difference reason. The method for calculating the quality difference is as shown in the third embodiment.

[0206] S4034: Send the diagnosis result and optimization plan. Send the calculation result of S4033 to achieve network optimization.

[0207] S4035: Feedback on the optimization effect. The optimization effect is returned by the installation and maintenance personnel. This step is not the core of this application and is not limited by this application.

[0208] S4036: Update and maintain the knowledge base. By updating and maintaining the knowledge base with external knowledge, the accuracy of the knowledge base can be effectively improved. This step is not the core of this application and is not limited by this application.

[0209] Example 4 is an implementation of the application of this application for the purpose of facilitating understanding. The content described in Example 4 is not within the scope of protection of this application. The contents described in Example 1, Example 2, and Example 3 are preferred embodiments of this application.

[0210] A root cause analysis method for poor quality of user videos, comprising:

[0211] Build a knowledge base through analysis of historical data;

[0212] Report real-time data through various means, analyze the root causes of poor quality by comparing with the knowledge base, and issue root causes and optimization plans.

[0213] Use historical data to form a knowledge base through association rule analysis.

[0214] The historical data includes set-top box probe data and gateway probe data.

[0215] The set-top box probe data is associated with the gateway probe data and the complaint data, and the associated data is divided into a complaint data set and a non-complaint data set.

[0216] Use clustering methods to construct silent poor quality datasets.

[0217] Use complaint data as the initial cluster center of the clustering algorithm.

[0218] Different similarity calculation methods are used for the first to eighth data types in the construction of the silent poor quality data set.

[0219] The rule for determining the silent poor quality data set is: the cluster contains at least one complaint user data.

[0220] The user perception index is calculated for the data in the complaint dataset and the silent poor quality dataset.

[0221] The user experience index includes: freeze ratio, freeze frequency, network score, resource request score, and resource request success rate;

[0222] The calculation method of the jam ratio is: the total jam duration divided by the user's viewing time;

[0223] The freeze frequency is calculated as follows: the freeze frequency is divided by the user's viewing time, and the user's viewing time needs to be converted into hours;

[0224] The network score is calculated as follows: the average value of the average TCP connection establishment time, the average TCP retransmission rate, and the TCP connection failure rate (1-TCP connection success rate) in the set-top box probe data;

[0225] The resource request score is calculated as follows: the average of the average delay of the m3u8 file request response, the average delay of the media file request response, and the average delay of the EPG request response in the set-top box probe data;

[0226] The resource request success rate is calculated as follows: the average value of the m3u8 file request success rate, the media file request success rate, and the EPG request success rate in the set-top box probe data.

[0227] The real-time diagnosis method is implemented by comparing the knowledge base.

[0228] The knowledge base rules include five parts: root cause of poor quality, user perception index, support, confidence and improvement.

[0229] The comparative knowledge base relies on the root cause of poor quality of knowledge base rules and the similarity of user perception index;

[0230] The similarity is defined as: the number of fields in the knowledge base rules that are identical to the reported data fields, divided by the number of knowledge base rule fields.

[0231] The root cause of poor quality of the user video is determined by the rules with high similarity.

[0232] When comparing, first match the user perception index in the rule, then calculate the similarity, and sort by the confidence of the rule as the suspected root cause of the poor quality.

[0233] The root causes of poor quality are divided into four categories, including: set-top box hardware problems, set-top box software problems, content source problems, and network problems.

[0234] The knowledge base should also include solutions to various root causes provided by operations and maintenance personnel.

[0235] An electronic device comprises: a memory, a processor, a receiver, a display and a computer program, wherein the computer program is stored in the memory, and the processor runs the computer program to execute any one of the above-mentioned methods for analyzing root causes of poor user video quality.

[0236] A storage medium comprises: a readable storage medium and a computer program stored in the readable storage medium, wherein the computer program is used to implement any of the above-mentioned root cause analysis methods for poor user video quality.

[0237] The contents not described in detail in the specification of the present invention belong to the prior art known to the professional and technical personnel in the field. It is pointed out here that the above description helps those skilled in the art to understand the invention, but does not limit the protection scope of the invention. Any equivalent replacement, modification and / or simplification of the above description without departing from the essence of the invention falls within the protection scope of the invention.

Claims

1. A method for analyzing the root causes of poor user video quality, characterized in that: The following steps are involved: Step S201, preprocessing the reported data to form valid data, wherein the reported data includes report-related data formed by collecting the operation data of the user video terminal; Step S202, constructing a data set using the complaint data and the valid data; Step S203, performing association rule analysis on the data in the data set construction and forming a knowledge base, the knowledge base including data tables with the following column names: support, confidence, lift, reason, and result; The terminal operation data in step S201 includes set-top box probe data and gateway probe data, and is uploaded periodically, and the period is limited by the first threshold. In the process of troubleshooting the root cause of poor quality of Internet TV services, the combination of set-top box probe data and gateway probe data can more accurately locate the fault point in the home local area network, thereby improving troubleshooting efficiency; The data preprocessing includes filtering data, and the filtering data includes filtering out data without viewing behavior; the data set construction includes associating complaint data with terminal operation data, screening out user data with a time difference that meets a second threshold as a complaint data set, and using other data as a non-complaint data set, filling in incomplete data through an AP clustering algorithm, and marking the successfully filled data as a silent poor quality data set. So far, a total of three data sets are constructed, namely: a complaint data set, a silent poor quality data set, and a non-complaint data set; the association rule analysis includes constructing a user perception index for records in the complaint data set and the silent poor quality data set, and discretizing them, performing association rule analysis on the discretized data, and outputting association rule records as knowledge base data. The knowledge base also includes optimization suggestions for different root causes.

2. The method for analyzing root causes of poor user video quality according to claim 1, characterized in that: The terminal operation data includes user complaint data and access network resource tree data, and the terminal operation data is reported to the network server via TCP.

3. The method for analyzing root causes of poor user video quality according to claim 1, characterized in that: The method includes using the knowledge base to perform real-time diagnosis on user videos of user terminals: the user terminal actively reports the set-top box probe data in real time, the real-time data is discretized, compared with the knowledge base, and the root cause analysis results and optimization suggestions are returned; Alternatively, you can remotely dial the user terminal, issue a collection command, upload real-time data, compare it with the knowledge base, and return the root cause analysis results and optimization suggestions. Alternatively, the operator performs periodic dial testing on the user terminal to obtain the network quality status of the user terminal, compares it with the knowledge base, and returns the root cause analysis results and optimization suggestions.

4. The method for analyzing root causes of poor user video quality according to claim 1, characterized in that: The data set construction in step S202 includes the following steps: Step S2021, data association is performed on the resource tree, gateway probe data, set-top box probe data, and complaint data; Step S2022, classifying the data to form a complaint data set and a non-complaint data set; Step S2023, expanding the complaint data set and forming a silent poor quality data set; Step S2024, constructing a user perception index using the complaint data set and the silent poor quality data set; Step S2025, discretizing the data and forming a correlation analysis data set.

5. The method for analyzing the root causes of poor user video quality according to claim 4, characterized in that: The complaint data set expansion in step S2023 is implemented by using the AP clustering algorithm, including the following steps: Step S20231, algorithm initialization, including initialization of matrices R, A, S; Step S20232, update the attraction matrix R; Step S20233, updating the attribution matrix A; Step S20234, performing an attenuation operation according to the attenuation coefficient; Step S20235, determine whether the matrix is ​​stable or the maximum number of iterations has been reached, if not, return to step S20232, if yes, proceed to step S20236; Step S20236, selecting the point with the largest A+R as the cluster center; Step S20237, exporting a silent poor quality data set; Step S20238, update the non-complaint data set.

6. An application network using the method for analyzing the root cause of poor video quality according to any one of claims 1 to 5, characterized in that: The system includes a user terminal, a network server, a database, and a knowledge base, and jointly performs the following steps: Step S401, data reporting; Step S402, analyzing historical data; Step S403, knowledge base assists in real-time diagnosis and optimization; Step S401 includes the following steps: Step S4011, the user terminal is turned on; Step S4012, reporting user terminal startup data to the network server; Step S4013, the network server stores the user terminal startup data in the database; Step S4014, the user terminal periodically reports data to the network server; Step S4015, the network server stores the periodic reporting data in a database; Step S402 includes the following steps: Step S4021, the network server extracts historical data from the database; Step S4022, the network server performs association rule analysis to form a knowledge base; Step S4023, the network server updates and maintains the knowledge base; Step S403 includes the following steps: Step S4031, the user terminal reports real-time measurement data to the network server to assist diagnosis; Step S4032, the network server extracts relevant knowledge from the knowledge base; Step S4033, the network server calculates the reason for the poor quality; Step S4034, the network server sends the diagnosis result and optimization plan to the user terminal; Step S4035, the user terminal feeds back the optimization effect to the network server; Step S4036, the network server updates and maintains the knowledge base.

7. An electronic device, characterized in that: include: A memory, a processor, a receiver, a display and a computer program, wherein the computer program is stored in the memory, and the processor runs the computer program to execute the method for analyzing the root cause of poor quality of user videos as described in any one of claims 1-5.

8. A storage medium, characterized in that: include: A readable storage medium and a computer program stored in the readable storage medium, wherein the computer program is used to implement the method for analyzing root causes of poor quality of user videos as described in any one of claims 1 to 5.