Fixed network user network quality control method and device and electronic equipment

Abstract

The invention discloses a fixed network user network quality control method and device and electronic equipment. The method comprises the following steps of: firstly, obtaining problem data in network operation of a fixed network user; the problem data comprises perception problem data fed back by a user and abnormal performance data monitored by a network side; inputting the perception problem data into a user perception poor quality model to obtain a first quality reason of a user perception level; and inputting the abnormal performance data into a network poor quality early warning model to obtain a second quality reason of a network side performance level, fusing the first quality reason and the second quality reason to obtain a poor-quality root reason of the fixed network user network; and finally, according to the poor-quality root cause, poor-quality order dispatching and problem repair are implemented, so that the network quality of the fixed network user is managed and controlled. According to the method, the user perception and the network performance in the network quality of the fixed network user can be considered, so that the user experience is improved, and the network reliability and service level are enhanced.

Description

Technical Field

[0001] The present invention belongs to the field of communication technology, and in particular relates to a method, device and electronic equipment for controlling the network quality of fixed-line users. Background Art

[0002] With the rapid growth of fixed-line user services and increasingly fierce market competition, operators urgently need to improve the overall quality of fixed-line services and user experience to attract new users and retain existing ones. Especially with the prevalence of diverse applications such as HD video, online gaming, smart homes, and the Internet of Things, users have higher expectations for network stability and user experience. However, current monitoring methods for fixed-line network performance and user experience remain limited. Many network issues are only discovered through feedback channels such as complaints after user performance issues have been resolved. This leads to a reactive "after-the-fact" approach, making it difficult to achieve early warning and proactive optimization.

[0003] Gaining network quality insight and analysis for fixed-line users faces numerous challenges. First, the business process is long and complex, involving multiple devices and layers, including set-top boxes, optical modems, routers, optical splitters, PONs, OLTs, switches, BRASs, and core routers. The complex link topology and numerous nodes significantly increase the difficulty of data collection and management. Second, due to the large scale and complex structure of network links, the numerous professional indicators involved, and the involvement of multiple devices and data sources, unified data collection, integration, and analysis become extremely difficult, severely limiting monitoring efficiency and diagnostic capabilities. Finally, there are numerous implicit relationships between user perception and various network performance indicators. Individual users have different network requirements in different usage scenarios, making accurately exploring the implicit connection between user perception and network quality a significant challenge.

[0004] Existing technologies rely primarily on two approaches to assessing network quality for fixed-line users: one relies solely on user feedback for problem diagnosis and quality management, while the other relies solely on early warning based on network performance parameters. Both approaches have limitations: the former fails to fully reflect the actual network status due to the subjective nature of user perception; while the latter, while capable of detecting performance anomalies, often overlooks the user's subjective experience, making it difficult to achieve comprehensive and accurate quality management from the user's perspective. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to address the above-mentioned shortcomings of the existing technology and propose a method, device and electronic equipment for controlling the network quality of fixed-line users. The method can take into account both user perception and network performance in the network quality of fixed-line users, thereby improving user experience and enhancing network reliability and service level.

[0006] In a first aspect, the present invention provides a method for controlling network quality of fixed-line users, the method comprising the following steps:

[0007] Obtain problem data on fixed-line users' network operations;

[0008] Among them, problem data includes perceived problem data reported by users and abnormal performance data monitored by the network side;

[0009] Input the perceived problem data into the user-perceived poor quality model to obtain the primary quality cause at the user-perceived level; and input the abnormal performance data into the network-side poor quality early warning model to obtain the secondary quality cause at the network-side performance level.

[0010] Combining the primary and secondary quality factors, we can determine the root cause of poor quality in fixed-line user networks.

[0011] Based on the root causes of poor quality, quality dispatch and problem repair are implemented to achieve network quality control for fixed-line users.

[0012] Furthermore, before inputting the perceived problem data into the user perceived quality difference model, the method further comprises the steps of: constructing the user perceived quality difference model;

[0013] Constructing a user-perceived quality difference model includes the following steps:

[0014] Obtaining user perceived quality difference data of different platforms, different types, and different dimensions to obtain a first data set;

[0015] performing data preprocessing on the first data set to obtain a second data set;

[0016] Divide the second data set into multiple category indicators based on network levels and business scenarios;

[0017] Filter out the key factors influencing user perception from category indicators;

[0018] According to the preset user perceived quality threshold, set the perceived quality judgment rules corresponding to the key influencing factors;

[0019] All perceived quality difference judgment rules are associated with all user perceived quality differences to form a user perceived quality difference model.

[0020] Furthermore, the key factors influencing user perception are screened out from the category indicators, which specifically includes the following steps:

[0021] Standardize the category indicators to obtain standardized data;

[0022] According to the covariance matrix of the standardized data, eigenvalue decomposition is performed to obtain eigenvectors and eigenvalues;

[0023] Arrange the eigenvalues ​​in descending order and select the top K eigenvalues ​​as the principal components based on the cumulative contribution rate;

[0024] According to the eigenvectors corresponding to the principal components, the main influencing indicators of user perception are obtained;

[0025] The main influencing indicators of user perception are taken as the key influencing factors of user perception.

[0026] Furthermore, before inputting the abnormal performance data into the network poor quality early warning model, the method further comprises the steps of: constructing the network poor quality early warning model;

[0027] Constructing a network quality warning model includes the following steps:

[0028] Obtain network performance data;

[0029] The network performance data is divided into BRAS index performance data, SW index performance data, MER index performance data, MAR index performance data, OLT index performance data, PON index performance data, and ONU index performance data;

[0030] According to a preset BRAS indicator threshold, a first network poor quality determination rule corresponding to the BRAS indicator performance data is set; and, according to a preset SW indicator threshold, a second network poor quality determination rule corresponding to the SW indicator performance data is set; and, according to a preset MER indicator threshold, a third network poor quality determination rule corresponding to the MER indicator performance data is set; and, according to a preset MAR indicator threshold, a fourth network poor quality determination rule corresponding to the MAR indicator performance data is set; and, according to a preset OLT indicator threshold, a fifth network poor quality determination rule corresponding to the OLT indicator performance data is set; and, according to a preset PON indicator threshold, a sixth network poor quality determination rule corresponding to the PON indicator performance data is set; and, according to a preset ONU indicator threshold, a seventh network poor quality determination rule corresponding to the ONU indicator performance data is set;

[0031] setting a first network adjustment rule for intelligent dynamic adjustment based on the first network poor quality determination rule; and setting a second network adjustment rule for intelligent dynamic adjustment based on the second network poor quality determination rule; and setting a third network adjustment rule for intelligent dynamic adjustment based on the third network poor quality determination rule; and setting a fourth network adjustment rule for intelligent dynamic adjustment based on the fourth network poor quality determination rule; and setting a fifth network adjustment rule for intelligent dynamic adjustment based on the fifth network poor quality determination rule; and setting a sixth network adjustment rule for intelligent dynamic adjustment based on the sixth network poor quality determination rule; and setting a seventh network adjustment rule for intelligent dynamic adjustment based on the seventh network poor quality determination rule;

[0032] Associating the BRAS indicator performance data, the first network poor quality determination rule, and the first network adjustment rule to form a BRAS indicator network poor quality early warning model; and associating the SW indicator performance data, the second network poor quality determination rule, and the second network adjustment rule to form a SW indicator network poor quality early warning model; associating the MER indicator performance data, the third network poor quality determination rule, and the third network adjustment rule to form a MER indicator network poor quality early warning model; associating the MAR indicator performance data, the fourth network poor quality determination rule, and the fourth network adjustment rule to form a MAR indicator network poor quality early warning model; associating the OLT indicator performance data, the fifth network poor quality determination rule, and the fifth network adjustment rule to form an OLT indicator network poor quality early warning model; associating the PON indicator performance data, the sixth network poor quality determination rule, and the sixth network adjustment rule to form a PON indicator network poor quality early warning model; associating the ONU indicator performance data, the seventh network poor quality determination rule, and the seventh network adjustment rule to form an ONU indicator network poor quality early warning model;

[0033] The network poor quality warning model is obtained by summarizing the BRAS indicator network poor quality warning model, SW indicator network poor quality warning model, MER indicator network poor quality warning model, MAR indicator network poor quality warning model, OLT indicator network poor quality warning model, PON indicator network poor quality warning model, and ONU indicator network poor quality warning model.

[0034] Furthermore, after implementing the poor quality dispatch and problem repair, the method further includes the steps of:

[0035] Verify the data of fixed-line user network operation based on the repair results after the problem is repaired.

[0036] Furthermore, the first quality reason and the second quality reason are integrated to obtain the root cause of the poor quality of the fixed-line user network, which specifically includes the following steps:

[0037] According to the network topology, all historical quality reasons are clustered and grouped;

[0038] Based on the same clustering, the potential correlation between the first quality cause and the second quality cause is identified, and the first quality cause and the second quality cause with the potential correlation are mapped and matched to form a fused quality cause;

[0039] Based on the two-way closed loop formed by network-to-user perception verification and user-to-network feedback verification, the reasons for the converged quality are verified, and the root causes of poor quality in the fixed-line user network are identified.

[0040] Furthermore, before clustering all the quality reasons according to the network topology, the method further comprises the steps of: constructing the network topology;

[0041] Building a network topology structure includes the following steps:

[0042] Determine the network element information of each network device;

[0043] According to the home network element information, the network device is set as a topological node of the network topology structure;

[0044] Use neighbor information exchanged in the LLDP protocol to identify link relationships between topological nodes;

[0045] Integrate the topological nodes and link relationships to obtain the end-to-end network topology structure.

[0046] In a second aspect, the present invention provides a device for controlling network quality of a fixed-line user, the device comprising:

[0047] An acquisition unit, used to acquire problem data of fixed-line users in network operation;

[0048] Among them, problem data includes perceived problem data reported by users and abnormal performance data monitored by the network side;

[0049] An input unit, connected to the acquisition unit, is used to input the perceived problem data into the user-perceived poor quality model to obtain the first quality cause at the user-perceived level; and is also used to input the abnormal performance data into the network poor quality warning model to obtain the second quality cause at the network-side performance level;

[0050] a fusion unit connected to the input unit, configured to fuse the first quality reason and the second quality reason to obtain a root cause of poor quality of the fixed-line user network;

[0051] The implementation unit is connected to the integration unit and is used to implement quality problems dispatch and problem repair according to the root cause of the quality problems, thereby achieving the management and control of the network quality of fixed-line users.

[0052] Furthermore, the fusion unit includes:

[0053] The clustering module is used to cluster all historical quality reasons according to the network topology;

[0054] an identification module, connected to the clustering module, for identifying a potential association between the first quality cause and the second quality cause based on the same cluster grouping;

[0055] A matching module, connected to the identification module, configured to map and match the first quality cause and the second quality cause that have a potential correlation to form a fused quality cause;

[0056] The verification module is connected to the matching module and is used to verify the quality reasons of the fusion based on the two-way closed loop formed by the perception verification from the network side to the user and the feedback verification from the user to the network side, so as to obtain the root cause of the poor quality of the fixed network user network.

[0057] In a third aspect, the present invention provides an electronic device comprising a memory and a processor, wherein a computer program is stored in the memory. When the processor runs the computer program stored in the memory, the processor executes the method for controlling the network quality of fixed-line users according to the first aspect.

[0058] This method integrates user feedback, perceived problem data, and abnormal network performance indicators to establish a multi-source information analysis model. This model can comprehensively and accurately identify the root causes of fixed-line user networks, improving the scientific nature and timeliness of fault diagnosis. While achieving a "user-perception-oriented" approach, it effectively integrates "network performance monitoring," thereby balancing user perception and network performance in fixed-line user network quality, providing a solid foundation for continuous network optimization. Specific beneficial effects are as follows:

[0059] 1. Improve the accuracy of fault diagnosis: By integrating user perception data and network performance indicators, the present invention can comprehensively and accurately identify the root causes in the network, reduce the misdiagnosis rate, and improve the efficiency of troubleshooting.

[0060] 2. Improve the timeliness of fault response: The multi-source information analysis model established by this invention can provide early warning of potential network problems, thereby shortening the time for fault discovery and repair and reducing the negative user experience.

[0061] 3. Enhance the scientific nature of network fault repair: This invention combines the perception quality difference model and the early warning model to ensure that the repair measures are targeted, improve the repair success rate, and reduce the cost of repeated maintenance.

[0062] 4. Improve user experience and satisfaction: This invention reduces the impact of network interruptions and performance degradation on users through early warning and precise repair, thereby enhancing users' overall satisfaction with network services.

[0063] 5. Support continuous optimization and intelligent operation and maintenance: The data-driven analysis foundation provided by this invention helps to achieve continuous optimization and remote intelligent scheduling of the network, and promotes the development of operation and management towards intelligence.

[0064] 6. Promote scientific decision-making in network management: This invention uses multi-source information fusion analysis to establish a scientific and quantifiable network quality assessment system to improve the decision-making level and management efficiency of network operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1Schematic diagram of a method for controlling network quality of fixed-line users in an embodiment of the present invention;

[0066] Figure 2 Schematic diagram of the process of controlling the network quality of fixed-line users in an embodiment of the present invention;

[0067] Figure 3 Schematic diagram of the fixed-line user network quality insight and analysis process in an embodiment of the present invention;

[0068] Figure 4 A schematic diagram of the indicator system in an embodiment of the present invention;

[0069] Figure 5 A schematic diagram of constructing a user perception model in an embodiment of the present invention;

[0070] Figure 6 A schematic diagram of a network quality warning model according to an embodiment of the present invention;

[0071] Figure 7 Schematic diagram of NtoC and CtoN bidirectional closed-loop driving in an embodiment of the present invention;

[0072] Figure 8 This is a flowchart of a multi-dimensional poor quality user network aggregation process according to an embodiment of the present invention;

[0073] Figure 9 This is a user-perceived full-link insight recognition display diagram in an embodiment of the present invention;

[0074] Figure 10 Schematic diagram of a device for controlling network quality of fixed-line users in an embodiment of the present invention;

[0075] Figure 11 2 is a diagram illustrating the architecture of an electronic device in an embodiment of the present invention.

[0076] Reference numerals: 10, acquisition unit, 20, input unit, 30, fusion unit, 40, implementation unit, 100, processor, 200, memory. DETAILED DESCRIPTION

[0077] In order to enable those skilled in the art to better understand the technical solutions of the present invention, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0078] It should be understood that the specific embodiments and drawings described herein are only used to explain the present invention rather than to limit the present invention.

[0079] It is understood that, in the absence of conflict, the various embodiments of the present invention and the various features in the embodiments may be combined with each other.

[0080] It can be understood that, for the convenience of description, the drawings of the present invention only show parts related to the present invention, while parts unrelated to the present invention are not shown in the drawings.

[0081] It can be understood that each unit and module involved in the embodiments of the present invention may correspond to only one physical structure, or may be composed of multiple physical structures, or multiple units and modules may be integrated into one physical structure.

[0082] It will be understood that, without conflict, the functions and steps marked in the flowcharts and block diagrams of the present invention may occur in an order different from that marked in the drawings.

[0083] It is understood that the flowcharts and block diagrams of the present invention illustrate the possible architectures, functions, and operations of the systems, devices, equipment, and methods according to various embodiments of the present invention. Each box in the flowchart or block diagram may represent a unit, module, program segment, or code, which contains executable instructions for implementing the specified functions. Moreover, each box or combination of boxes in the block diagram and flowchart may be implemented using a hardware-based system that implements the specified functions, or may be implemented using a combination of hardware and computer instructions.

[0084] It is understandable that the units and modules involved in the embodiments of the present invention may be implemented by software or hardware. For example, the units and modules may be located in a processor.

[0085] Example 1:

[0086] This embodiment provides a method for managing and controlling the network quality of fixed-line users. This method is applicable to daily network maintenance and optimization scenarios for home broadband service operators, particularly when faced with the diverse network needs of a large number of users and complex network environments. It enables real-time monitoring, rapid diagnosis, and precise repair. By combining user perception with network performance data, this method not only provides early warning of potential faults and shortens fault response time, but also supports continuous optimization of network quality and enhances user experience. This ensures that the network can stably and efficiently meet the diverse needs of users in multiple application scenarios, such as high-definition video, online gaming, smart homes, and the Internet of Things, thereby enhancing operators' competitive advantage and customer satisfaction.

[0087] like Figure 1 As shown, the method for controlling the network quality of fixed-line users in this embodiment includes the following steps:

[0088] Step S1: Obtain problem data during fixed-line user network operation; the problem data includes perceived problem data fed back by users and abnormal performance data monitored by the network side.

[0089] Step S2: Input the perceived problem data into the user perceived quality model to obtain the first quality reason at the user perception level; and input the abnormal performance data into the network quality warning model to obtain the second quality reason at the network performance level.

[0090] As a specific implementation method, before inputting the perception problem data into the user perception quality difference model, the method further includes the steps of: constructing the user perception quality difference model;

[0091] Constructing a user-perceived quality difference model includes the following steps:

[0092] Obtaining user perceived quality difference data of different platforms, different types, and different dimensions to obtain a first data set;

[0093] performing data preprocessing on the first data set to obtain a second data set;

[0094] Divide the second data set into multiple category indicators based on network levels and business scenarios;

[0095] Filter out the key factors influencing user perception from category indicators;

[0096] According to the preset user perceived quality threshold, set the perceived quality judgment rules corresponding to the key influencing factors;

[0097] All perceived quality difference judgment rules are associated with all user perceived quality differences to form a user perceived quality difference model.

[0098] As a specific implementation method, the key factors affecting user perception are screened out from the category indicators, which specifically includes the following steps:

[0099] Standardize the category indicators to obtain standardized data;

[0100] According to the covariance matrix of the standardized data, eigenvalue decomposition is performed to obtain eigenvectors and eigenvalues;

[0101] Arrange the eigenvalues ​​in descending order and select the top K eigenvalues ​​as the principal components based on the cumulative contribution rate;

[0102] According to the eigenvectors corresponding to the principal components, the main influencing indicators of user perception are obtained;

[0103] The main influencing indicators of user perception are taken as the key influencing factors of user perception.

[0104] As a specific implementation method, before inputting the abnormal performance data into the network poor quality early warning model, the method further includes the steps of: constructing the network poor quality early warning model;

[0105] Constructing a network quality warning model includes the following steps:

[0106] Obtain network performance data;

[0107] The network performance data is divided into BRAS index performance data, SW index performance data, MER index performance data, MAR index performance data, OLT index performance data, PON index performance data, and ONU index performance data;

[0108] According to a preset BRAS indicator threshold, a first network poor quality determination rule corresponding to the BRAS indicator performance data is set; and, according to a preset SW indicator threshold, a second network poor quality determination rule corresponding to the SW indicator performance data is set; and, according to a preset MER indicator threshold, a third network poor quality determination rule corresponding to the MER indicator performance data is set; and, according to a preset MAR indicator threshold, a fourth network poor quality determination rule corresponding to the MAR indicator performance data is set; and, according to a preset OLT indicator threshold, a fifth network poor quality determination rule corresponding to the OLT indicator performance data is set; and, according to a preset PON indicator threshold, a sixth network poor quality determination rule corresponding to the PON indicator performance data is set; and, according to a preset ONU indicator threshold, a seventh network poor quality determination rule corresponding to the ONU indicator performance data is set;

[0109] setting a first network adjustment rule for intelligent dynamic adjustment based on the first network poor quality determination rule; and setting a second network adjustment rule for intelligent dynamic adjustment based on the second network poor quality determination rule; and setting a third network adjustment rule for intelligent dynamic adjustment based on the third network poor quality determination rule; and setting a fourth network adjustment rule for intelligent dynamic adjustment based on the fourth network poor quality determination rule; and setting a fifth network adjustment rule for intelligent dynamic adjustment based on the fifth network poor quality determination rule; and setting a sixth network adjustment rule for intelligent dynamic adjustment based on the sixth network poor quality determination rule; and setting a seventh network adjustment rule for intelligent dynamic adjustment based on the seventh network poor quality determination rule;

[0110] Associating the BRAS indicator performance data, the first network poor quality determination rule, and the first network adjustment rule to form a BRAS indicator network poor quality early warning model; and associating the SW indicator performance data, the second network poor quality determination rule, and the second network adjustment rule to form a SW indicator network poor quality early warning model; associating the MER indicator performance data, the third network poor quality determination rule, and the third network adjustment rule to form a MER indicator network poor quality early warning model; associating the MAR indicator performance data, the fourth network poor quality determination rule, and the fourth network adjustment rule to form a MAR indicator network poor quality early warning model; associating the OLT indicator performance data, the fifth network poor quality determination rule, and the fifth network adjustment rule to form an OLT indicator network poor quality early warning model; associating the PON indicator performance data, the sixth network poor quality determination rule, and the sixth network adjustment rule to form a PON indicator network poor quality early warning model; associating the ONU indicator performance data, the seventh network poor quality determination rule, and the seventh network adjustment rule to form an ONU indicator network poor quality early warning model;

[0111] The network poor quality warning model is obtained by summarizing the BRAS indicator network poor quality warning model, SW indicator network poor quality warning model, MER indicator network poor quality warning model, MAR indicator network poor quality warning model, OLT indicator network poor quality warning model, PON indicator network poor quality warning model, and ONU indicator network poor quality warning model.

[0112] Step S3: The first quality reason and the second quality reason are integrated to obtain the root cause of the poor quality of the fixed-line user network.

[0113] As a specific implementation method, the first quality reason and the second quality reason are integrated to obtain the root cause of poor quality of the fixed-line user network, which specifically includes the following steps:

[0114] According to the network topology, all historical quality reasons are clustered and grouped;

[0115] Based on the same clustering, the potential correlation between the first quality cause and the second quality cause is identified, and the first quality cause and the second quality cause with the potential correlation are mapped and matched to form a fused quality cause;

[0116] Based on the two-way closed loop formed by network-to-user perception verification and user-to-network feedback verification, the reasons for the converged quality are verified, and the root causes of poor quality in the fixed-line user network are identified.

[0117] As a more specific implementation, before clustering all quality reasons according to the network topology, the method further includes the steps of: constructing the network topology;

[0118] Building a network topology structure includes the following steps:

[0119] Determine the network element information of each network device;

[0120] According to the home network element information, the network device is set as a topological node of the network topology structure;

[0121] Use neighbor information exchanged in the LLDP protocol to identify link relationships between topological nodes;

[0122] Integrate the topological nodes and link relationships to obtain the end-to-end network topology structure.

[0123] Step S4: Implement quality control and problem repair based on the root cause of the quality problem, thereby achieving network quality control for fixed-line users.

[0124] As a specific implementation method, after implementing poor quality dispatch and problem repair, the method further includes the steps of:

[0125] Verify the data of fixed-line user network operation based on the repair results after the problem is repaired.

[0126] This embodiment proposes a benchmarking method for network quality insight and analysis for home broadband fixed-line users. Based on basic information such as collected SA data, RMS data reported by other systems, access network performance data, metropolitan area network performance data, and line data, a two-wheeled system is constructed, encompassing both network and user problem discovery and closed-loop driving. This system aggregates and analyzes user quality differences across the entire link, from BRAS, switches (SWs), OLTs, PON interfaces, and primary and secondary optical splitters, enhancing network analysis capabilities while also creating a comprehensive profile of individual broadband users and the ability to locate quality differences. Ultimately, this method enables the automatic identification and independent definition of multi-dimensional user quality difference models, the automatic detection and early warning of network quality differences, and comprehensive network insights from both the C2N and N2C perspectives. This achieves end-to-end user network perception and penetration, thereby comprehensively improving the efficiency of network problem discovery and resolution.

[0127] Figure 2The demonstration showcases a comprehensive fixed-line user quality insight and analysis implementation plan. This solution analyzes and processes network data through a series of steps and modules, aiming to improve network performance and user satisfaction. The solution begins with quality root cause analysis and quality degradation aggregating. Performance data from broadband access servers (BRAS), switches (SW), optical line terminals (OLT), optical network units (ONU), and passive optical networks (PON) is analyzed to identify and identify the causes of network quality issues. Next, the solution moves into quality modeling, constructing a perceived quality model and a quality degradation model to assess network quality and identify quality degradation issues, respectively. Then, through multi-dimensional user-perceived quality correlation assessment, data from BRAS, SW, OLT, PON, and ONU devices is aggregated and analyzed to assess user perceived quality. The solution then configures quality assessment rules (ERAs), service quality (SVs), quality assessment thresholds (OLs), and ONU quality assessment parameters. During the demarcation and location phase of perceived poor quality issues, the solution identifies network issues, such as Wi-Fi signal coverage and optical attenuation, and determines the causes of these issues through network-to-consumer (NtoC) and user-to-network (CtoN) analysis. The problem remediation execution phase involves generating remediation tasks based on identified poor quality issues, triggering alarms for network elements, and ensuring problem resolution through a work order management system. This phase also includes statistics for various problem tickets and SLA closed-loop statistics, as well as the transfer and resolution of regulatory and initial issues. The effectiveness evaluation phase focuses on changes in network quality key performance indicators (KPIs), the network quality improvement rate for users, changes in the proportion of poor quality users, and the improvement rate for poor quality users. Finally, in the phase of further root cause improvement for poor quality users, the solution provides one-click diagnosis of network link problem nodes, prominently displays network link data, and displays PON and port quality and attribute information by clicking on a link pipeline. Through these steps, the entire flowchart systematically analyzes and resolves poor quality issues in fixed-line user networks, thereby improving network quality and user satisfaction.

[0128] The solution for fixed-line user network quality analysis covers several key steps, including: building a perceived quality degradation model, customizing the model, clustering and evaluating users with perceived quality degradation based on multi-dimensional indicators, identifying and locating quality degradation issues, dispatching and managing work orders for these issues, and ultimately evaluating outcomes and providing end-to-end user perception insights. The entire solution encompasses the entire process from discovery, clustering, identification, to remediation of quality degradation issues, providing a systematic solution for efficient and accurate network quality management. The following sections will detail the specific content and operational methods for each step.

[0129] Step 1: Building a perception indicator system:

[0130] The premise for building a perceptual quality difference model is to establish a perceptual indicator system. Currently, the sources of network quality difference data are diverse. The B domain includes: user information, broadband terminal information, IPTV account information, contract speed, product brand, etc.; the O domain includes: SA data, RMS data, access network data, metropolitan area network data, IP integrated network management data, resource data, ESB data, etc. It is necessary to comprehensively manage the data from these different data sources to lay the foundation for the construction of the perceptual quality difference model.

[0131] Before establishing an indicator system, data from multiple data sources must be preprocessed, including key steps such as data cleaning, conversion, parsing, and aggregation. The specific process is as follows: First, acquire home broadband data from different platforms, types, and time dimensions to ensure standardized data access and unified processing. Next, clean the raw data, including removing duplicates, filling missing values, converting data formats, converting units, and verifying data integrity and accuracy. After cleaning, analyze the data to extract key indicators, integrate information from different dimensions, and classify and process it to uncover valuable insights and transform it into reliable model foundation data. Then, through data correlation and aggregation, the processed multi-source data is integrated to enrich the dataset and provide a solid foundation for subsequent data analysis and model building. Finally, based on the above steps, a broadband perception indicator model set is formed to meet the needs of subsequent network quality assessment and optimization. After these preprocessing steps, the resulting model indicator set can effectively support further analysis and decision-making. The specific indicator set for fixed-line user network quality insight and analysis is shown in Table 1.

[0132] Figure 3This paper describes a hierarchical process for gaining insights and analyzing fixed-line user network quality. This process processes and analyzes data across four key layers: the Data Detail Layer (ODS), the Data Convergence Layer (DWD), the Data Intermediary Layer (DWI), and the Data Service Layer (DMK) to improve network performance and user satisfaction. The ODS layer collects raw data, including service provisioning information, dedicated line information, resource topology data, user information, line number information, LLDP data, performance data from broadband access servers (BRAS), switches (SW), optical line terminals (OLTs), and passive optical networks (PONs), as well as data from integrated network management, Huawei Service Awareness (SA), Resource Management System (RMS), and Enterprise Service Bus (ESB). The DWD layer cleans, parses, and correlates this raw data to ensure accuracy and consistency. It also establishes topological relationships between network devices and collects user-related data, such as Wi-Fi signal coverage and optical attenuation. The data intermediate layer (DWI) further aggregates and processes data to form user-level 15-minute indicator data. These data include performance indicators of BRAS, SW, OLT, and PON, as well as user perception indicators such as the number of abnormal ONU disconnections and optical attenuation, which are aggregated by user and time granularity (15 minutes). Finally, at the data service layer (DMK), the data from the intermediate layer is used to build a network quality insight and analysis model, including data association, fixed-line user network quality insight and analysis model, service availability (SA) perception quality difference model, IPTV quality difference model, and user-defined indicator quality difference model. The entire process, from the collection, cleaning, association, aggregation of raw data to the final analysis and service, achieves comprehensive insight and analysis of the network quality of fixed-line users, providing solid data support for network optimization and improvement of user satisfaction. Table 1: Fixed-line user network quality insight and analysis indicator set:

[0133]

[0134]

[0135]

[0136] Step 2: Build a broadband user quality model:

[0137] After obtaining the model indicator set, the next step is to build a broadband user perceived quality difference model. Specific models include: SA perceived quality difference model, IPTV quality difference model, BIP quality difference model, and indicator quality difference custom model.

[0138] Figure 4This article describes in detail the process for building a broadband user network poor quality (QoQ) model, including the configuration of various QoQ models and the customization of key performance indicator thresholds. The entire process aims to identify and resolve network issues by monitoring and analyzing various network services and quality indicators, thereby improving user satisfaction. The figure illustrates four main QoQ models: the SA (Service Availability) QoQ model, the BIP (Business Impact Indicator) QoQ model, the IPTV (Interactive Internet Television) QoQ model, and the custom QoQ indicator model. Each model monitors specific network services and performance indicators. In the SA QoQ model, QoQ judgment rules are based on the proportion of the top three poor quality applications and the proportion of poor quality occurrences across all applications. The BIP QoQ model focuses on both upstream and downstream BIP values. The IPTV QoQ model focuses on the number of freezes, screen artifacts, packet loss rate, and the number of consecutive poor quality events. The custom QoQ indicator model allows users to customize QoQ judgment rules based on indicators such as ONU (Optical Network Unit) receive power, ONU bit error rate, and the number of abnormal AAA (Authentication, Authorization, and Accounting) disconnections. The figure also mentions the model indicator set, which includes key indicators such as application name, number of poor application quality, percentage of poor application quality, uplink BIP, downlink BIP, number of freezes, number of screen distortion, MOS value (mean opinion score), and number of consecutive poor quality. Feature extraction methods such as PCA (principal component analysis) can be used to extract the most useful features for model analysis from the raw data. Users can customize the thresholds for each indicator according to their actual needs to more accurately identify and locate poor quality issues in the network. Through these steps, network operators can build a comprehensive poor quality model to monitor and analyze network quality, promptly identify and resolve network issues, and thus improve user satisfaction.

[0139] The broadband user quality model is constructed in three steps:

[0140] Model indicator set construction: In actual models, there are many indicators that affect the quality difference model, but not every type of indicator has the same weight on broadband user perception. Since there are many features obtained, using all features to calculate the indicator score will affect computational efficiency. Therefore, principal component analysis (PCA) is used here for feature selection.

[0141] The main process of principal component analysis is as follows:

[0142] 1) Data standardization:

[0143] Since the principal component analysis method is sensitive to dimension, the data needs to be standardized first.

[0144]

[0145] i represents the i-th sample data, j represents the j-th feature;

[0146] x ij is the original value of the jth feature of the i-th sample;

[0147] max(x j )The maximum value of the j-th feature;

[0148] min(x j )The minimum value of the j-th feature;

[0149] z ij is the standardized result;

[0150] 2) Calculate the covariance matrix. The covariance matrix can be calculated by the following formula

[0151]

[0152] m represents the number of samples;

[0153] z ij is the standardized result;

[0154] 3) Solve the covariance matrix to calculate the directional variance of each eigenvector, that is, the importance of the feature

[0155]

[0156] z j represents the normalized eigenvector;

[0157] Represents the covariance matrix result of the indicator.

[0158] 4) Sort the directional variance of the feature vectors and select the features corresponding to the first k values ​​as the final selected features.

[0159] Model configuration: After feature selection for each type of perception model, the next step is model configuration. Model configuration mainly involves two steps: one is to customize the indicator threshold, and the other is to assign association rules based on the threshold combination of each indicator to ultimately form a quality difference model.

[0160] Customizable indicator threshold settings: The thresholds for each indicator are not set blindly. Instead, they are determined by initially acquiring a large amount of complaint data as sample training. AI learning and training are then used to predict network-wide user performance. Finally, after verification, the thresholds for each indicator are determined as the degradation thresholds for model judgment.

[0161] Setting of quality difference judgment rules: In order to obtain a more accurate model, we initially restricted the model quality difference rules more strictly. By combining indicators and assigning quality difference association rules, we output a list of quality difference users and jointly verify it with cities and counties, which has been polished for a long time.

[0162] Ultimately, the SA perception quality difference model, IPTV quality difference model, BIP quality difference model, and indicator quality difference custom model are formed, and each model realizes dynamic adjustment of network self-intelligence L3.

[0163] Step 3: Build a network quality warning model:

[0164] User perception is impacted by multiple dimensions and links. Therefore, establishing the ability to proactively identify and localize poor quality issues requires a comprehensive approach across the entire broadband user network topology. This approach, combined with segmented identification of network pipeline indicators, automatically and accurately identifies poor quality at the BRAS, SW, MER, MAR, OLT, PON, optical splitter, and ONU levels, providing detailed cause analysis and ultimately localizing poor quality issues across the entire network. Due to previous data collection challenges, this network poor quality model primarily focuses on five dimensions: BRAS, SW, OLT, PON, and ONU.

[0165] Figure 5This paper presents a comprehensive process for building a broadband user network poor quality (QoQ) model. This process, comprising several key steps and components, aims to identify and resolve network issues and improve user satisfaction by meticulously monitoring and analyzing network services and quality indicators. The figure details four key QoQ models: the SA (Service Availability) QoQ model, the BIP (Business Impact Indicator) QoQ model, the IPTV (Interactive Internet Television) QoQ model, and a custom QoQ indicator model. Each model focuses on different network services and performance indicators. In the SA QoQ model, QoQ judgment rules are based on the proportion of the top three poor quality applications and the proportion of poor quality events across all applications. The BIP QoQ model focuses on upstream and downstream BIP values, while the IPTV QoQ model focuses on the number of freezes, screen artifacts, packet loss rate, and the number of consecutive poor quality events. The custom QoQ indicator model offers flexibility, allowing users to customize QoQ judgment rules based on metrics such as ONU (Optical Network Unit) receive power, ONU bit error rate, and the number of abnormal AAA (Authentication, Authorization, and Accounting) disconnections. The figure also shows the model metric set, which includes key indicators such as application name, number of poor application quality, percentage of poor application quality, uplink BIP, downlink BIP, number of freezes, number of screen distortion, mean opinion score (MOSS), and number of consecutive poor quality events. These indicators use feature extraction methods such as PCA (principal component analysis) to extract the most useful features for model analysis from the raw data. Furthermore, users can customize the thresholds for each metric based on their actual needs to more accurately identify and locate poor quality issues in the network. Through these steps, network operators can build a comprehensive poor quality model to monitor and analyze network quality, promptly identify and resolve network issues, and thus improve user satisfaction. The entire process embodies a systematic approach from data collection, model configuration, metric threshold setting, to problem solving, providing strong support for network management and optimization.

[0166] The network quality warning model is constructed in four steps:

[0167] Model indicator set construction: After acquiring performance data of all network elements and ONUs across multiple platforms and systems and normalizing them, the final indicator set for BRAS, SW, MER, MAR, OLT, PON, and ONU is formed as the output parameters for model construction;

[0168] Quality difference model configuration: There are two main steps in model configuration: the first is to customize the indicator threshold, and the second is to assign association rules based on the threshold combination of each indicator to finally form the quality difference model.

[0169] Customized indicator threshold settings: Customized indicator threshold settings follow a set of rules. The model obtains indicators from all network elements and ONUs at a 15-minute granularity per day. The worst 5% of each indicator are then filtered out and the average is calculated as the quality threshold. The threshold is not fixed but is dynamically adjusted based on the indicator value over different periods. Currently, dynamic threshold adjustment is implemented on a weekly basis.

[0170] Poor quality judgment rule: When the actual value of one or N 15-minute granularity indicators is worse than the poor quality threshold value, it is judged as a poor quality network element.

[0171] Poor quality warning trigger: Judging whether a network is of poor quality based solely on the quality of its indicators is unscientific. Extensive analysis and research have revealed that some networks have excellent indicators, but user perception doesn't reflect this. Therefore, to ensure authentic and accurate judgment of network perception, we incorporate user perception into our analysis when identifying poor quality networks. Consequently, we have developed a set of judgment rules for network quality warnings, as shown in the figure above. These rules have been extensively verified to be reliable.

[0172] Poor quality warning model: After the above steps, the BRAS poor quality warning model, SW poor quality warning model, MER poor quality warning model, MAR poor quality warning model, OLT poor quality warning model, PON poor quality warning model, and ONU poor quality warning model are finally formed. Each model realizes dynamic adjustment of network self-intelligence L3.

[0173] Step 4: Multi-dimensional network aggregation capabilities for poor-quality users: NtoC and CtoN bidirectional closed-loop driving:

[0174] The purpose of multi-dimensional network clustering of poor-quality users is to cluster users with common poor-quality characteristics in the existing network layer by layer, analyze whether poor-quality users are concentrated on the network, discover network-side clustering problems through user poor-quality clustering and sudden changes in poor-quality rates, analyze and verify models, and issue early warnings and closed-loop solutions for clustered poor-quality problems. At the same time, it also reversely analyzes network element and link performance indicators to determine whether poor quality will also affect user application perception. Through two-way mining, NtoC and CtoN bidirectional closed-loop drives are formed. This provides effective and fast support for problems such as complex end-to-end processes for fixed-line users, difficulty in locating poor-quality network problems, and the inability of traditional methods to accurately identify the causes of poor quality perceived by users.

[0175] Figure 6This figure details a multi-dimensional process for monitoring and troubleshooting fixed-line user network quality. This process, driven by a bidirectional closed-loop mechanism (NtoC (Network to Customer) and CtoN (Customer to Network), comprehensively identifies and resolves network quality issues. The left side of the process illustrates the multi-dimensional collection of poor-quality user networks, including a list of users with perceived poor service availability (SA), a list of users with poor service impact indicators (BIP), and a list of users with poor interactive network television (IPTV) quality. These lists are generated by analyzing data collected from network devices and systems such as BRAS, SW, MER, MAR, OLT, PON, and ONU. The network architecture in the center of the figure illustrates how user devices (such as set-top boxes, TVs, computers, and mobile phones / tablets) access the network through devices such as ONUs, PONs, and OLTs, and then connect to the backbone and metropolitan area networks through BRAS, SW, and MER. This involves different transmission media, such as optical fiber and copper cables. The right side of the diagram focuses on identifying and locating poor-quality network elements. This process assesses and analyzes the causes of poor quality in key network equipment, including switches (SWs), broadband access servers (BRASs), multi-service edge routers (MERs), multi-access routers (MARs), optical line terminals (OLTs), passive optical networks (PONs), and optical network units (ONUs). For example, an OLT or PON might be identified as poor quality due to excessively high bit errors at the PON interface, while an ONU might be judged as poor quality due to 10 abnormal disconnections. The far right side of the diagram represents the application layer, encompassing the content delivery network (CDN) and various application services, which are directly related to the user's network experience. Overall, this process enables network operators to comprehensively monitor network quality, promptly identify and accurately locate network issues, and then take effective measures to optimize and improve the network, ultimately improving overall user satisfaction.

[0176] Step 5: Perceive poor quality and dispatch orders and fix problems:

[0177] Based on step 4, poor-quality users are grouped into the multi-dimensional network of BRAS, SW, OLT, PON, primary splitter, and secondary splitter to trigger an alert. This creates a dispatch database, and through a series of work order dispatching, processing, verification, and closed-loop processes, the goal of improving network quality and perception is ultimately achieved.

[0178] Figure 7This article details a closed-loop process for handling fixed-line user network quality issues. This multi-step process systematically identifies, issues, handles, and evaluates network quality issues to improve user satisfaction. The process begins with the collection and dispatch of poor-quality user networks, including a list of various issues, such as IPTV lag, frequent abnormal ONU disconnections, weak PON light, excessive packet loss, OLT traffic congestion, SW port issues, and BRAS NAT board problems. During the departmental dispatch and processing phase, the problem list is first distributed to the appropriate department by opening interfaces and invoking processing capabilities. The Smart Home or Cloud Network / Integrated Network Management department then handles the issues based on the dispatched issues. After the issues are resolved, the effectiveness evaluation phase begins, evaluating the effectiveness of the handling to determine whether the issues have been resolved. If the issue is not resolved (NOK), the one-click assisted location phase begins; if the issue is resolved (OK), the process ends. In the one-click assisted location phase, for any unresolved issues, the assisted location function is used to further analyze the cause and develop a resolution strategy. This includes analyzing the resolution of poor quality issues in network elements such as BRAS, SW, OLT, and PON. If the problem is still not resolved (NOK), it is necessary to continue optimizing and exploring the root cause of the poor quality; if the problem is resolved (OK), the process ends. The figure also shows the interface of the broadband user's full-link insight capability, which can intuitively display network link data and abnormal link data, and allows users to click on the link topology to intuitively view the network link markings and network element and port attributes. Overall, this process provides network operators with a complete closed loop from problem identification to final resolution, ensuring that network quality issues can be systematically handled and effectively resolved, thereby improving user satisfaction.

[0179] Based on the poor quality users identified in step 4, they are grouped into the BRAS, SW, OLT, PON, primary splitter, and secondary splitter multi-dimensional network. After triggering an alert, a dispatch database is formed according to different dimensions. This is then intuitively presented through the BRAS quality analysis interface, SW quality analysis interface, OLT quality analysis interface, PON quality analysis interface, and splitter quality analysis interface. The root cause of the poor quality is located by combining network anomaly attribute analysis.

[0180] Open interfaces and call capabilities based on the pain points of each department, supporting the operation and maintenance side to optimize network quality and locate problems before line personnel and Zhijia on-site treatment.

[0181] After a period of capability output verification, the overall effect evaluation is ideal. In order to more accurately locate the perception problems of individual broadband users, we are considering adding an identification model for broadband users' perception of end-to-end network penetration to better assist in the one-click diagnosis of single-user perception problems.

[0182] Identifying broadband user perception issues solely based on network element quality is insufficient. Some network elements may have poor performance but excellent user perception, so there's no necessary correlation between network element quality and user perception. Our early analysis revealed that the poor perception experienced by some low-quality users is related to the perception of the relay links between each network element. To better identify potential issues with perception, we built an end-to-end network penetration identification model for broadband user perception. This model identifies user perception issues at a more refined level and presents them intuitively in a topological manner.

[0183] Figure 8 This diagram details a network topology and its performance data, which is used to analyze and monitor network device performance. The left side of the figure shows a typical network topology consisting of multiple layers of network devices, including a core router (CR), broadband network gateway (BNG), switch (SW), multi-service edge router (MER), multi-access router (MAR), optical line terminal (OLT), and optical network unit (ONU). Users access the OLT through the ONU, which then connects to the MAR and MER via a PON (passive optical network), and to the CR via the SW and BNG, forming a complete network path. The two tables on the right side of the figure present performance data for the BNG equipment in Block Z and the CR equipment in Block A, respectively. The performance data table for the BNG equipment in Block Z lists the BNG port, port name, port optical power (dBm), port received optical power (dBm), port packet loss, and port packet error. For example, port 1 has an optical power of -13.35 dBm and a received optical power of -3.17 dBm, with no packet loss or errors. The performance data table for Block A-CR lists the ports, port names, port emitting power (dBm), port receiving power (dBm), and port utilization (%) of the CR equipment. For example, the emitting power of port 1 is -13.35dBm, the receiving power is -3.17dBm, and the port utilization is 50%. The bottom of the figure shows statistics on user complaints and user activity, which are used to evaluate network service quality. User complaints may be related to network performance issues such as packet loss, high latency, or low bandwidth. In addition, the figure also mentions relay link awareness analysis, which may involve monitoring and analyzing the performance of network relay links to ensure link stability and performance. Through this view, network administrators can intuitively understand the performance status of network equipment, quickly identify potential network problems, and take appropriate measures for optimization and maintenance, thereby ensuring efficient network operation and user satisfaction.

[0184] The broadband user-perceived end-to-end network penetration identification model only requires the following steps:

[0185] Step 1: Network topology construction:

[0186] Fixed-line users have complex network topologies. Automatic and accurate identification of network topology is the cornerstone of end-to-end quality demarcation. The following describes the most common and accurate network topology identification methods. The specific approach is as follows:

[0187] Data collection: Daily granularity RADIUS call record data is obtained through interconnection with the data sharing platform for building home network association topology. Metropolitan area network LLDP and smart metropolitan area network LLDP data is obtained through interconnection with the data sharing platform for building home network and smart metropolitan area network association topology.

[0188] Identification of home network element: In the OLT-PON access network, in the home broadband FTTH / FTTR networking mode, the OLT determines the logical port information of the user access based on the PPPoE / DHCP / DHCPv6 message initiated by the home gateway and adds a PPPoE tag to the PPPoE / DHCP / DHCPv6 message.

[0189] / Option82 / Option18. After receiving the OLT authentication and billing message, the BRAS adds the BRAS physical port and user SVLAN and CVLAN information to the user access tag field to form the access identifier logicalportno string field in the home broadband authentication call log. At the same time, RADIUS appends the BRAS's network-wide unique public network loopback address to the ISNIP field in the authentication call log.

[0190] Based on the principle of generating access identifiers in call logs, the access topology attribution of single-user BRAS, OLT, PON port, and ONU_ID network elements can be completed.

[0191] Association and identification of home networks and smart metropolitan area networks: LLDP data of smart metropolitan area networks is obtained from the group and updated daily, stored in a local large database. The LLDP data of the metropolitan area network and the LLDP data of the smart metropolitan area network are then associated using the primary key OLT IP field to achieve topological association and identification of the home network and smart metropolitan area network.

[0192] The topology of the home network and the smart metropolitan area network is realized based on the LLDP data of the metropolitan area network and the LLDP data of the smart metropolitan area network.

[0193] Step 2: Broadband user perception of end-to-end network penetration identification

[0194] After establishing a single-user network topology based on step one, the relay performance data of each section the user passes through is obtained and combined with the topology for intuitive presentation. The topology allows for a clear understanding of network link perception and abnormal link data. By clicking on the link markings, network element and port attributes, such as device peak bandwidth, average bandwidth, latency, and temperature exceeding thresholds, are prominently displayed, enabling timely detection of abnormal points and assisting in addressing the root causes of poor quality.

[0195] Figure 9 This article describes in detail two network device data structures: metropolitan area network (MAN) Link Layer Discovery Protocol (LLDP) data and intelligent MAN LLDP data. These structures are used to record and describe the connection information and attributes between network devices. MAN LLDP data contains fields such as the A-side device name (a_device_name), A-side device ID (a_device_id), A-side device IP address (a_loopback_ip), A-side device model (a_device_model), A-side device hierarchy (a_resource_name), A-side port name (a_port_name), B-side device name (b_device_name), B-side device ID (b_device_id), B-side device IP address (b_loopback_ip), B-side device model (b_device_model), B-side device hierarchy (b_resource_name), and B-side port name (b_port_name). These fields cover basic network device attributes and connection information. Smart MAN LDP data builds on this foundation by adding the OLT IP (olt_ip) and OLT name (olt_name) fields, and includes the corresponding A-side and B-side device information fields in the MAN LDP data. These fields not only provide basic device information but also link it to the MAN LDP data using the OLT IP address as the primary key. The arrows in the figure indicate how the two data structures are linked. Smart MAN LDP data is matched with MAN LDP data using the OLT IP address, thus linking the same OLT device. This linkage mechanism allows network administrators to track and manage the same OLT device in both datasets. This flowchart provides a clear view for network administrators to understand and analyze the connectivity and attributes of network devices, enabling more efficient network management and maintenance. These detailed data structures and linkage methods ensure the consistency and accuracy of network device information, supporting network optimization and troubleshooting.

[0196] This embodiment proposes a comprehensive method for controlling the quality of fixed-line user networks, encompassing the entire process from data preprocessing, indicator system construction, and quality difference model development to network quality difference warning, problem aggregation, and closed-loop remediation. By cleaning, parsing, and aggregating multi-source, multi-dimensional data, a perception quality difference model and warning model are constructed, effectively enabling real-time monitoring and automatic warning of network quality. Combining user perception with network performance, a multi-model joint analysis approach is employed to identify potential faults and root causes, supporting fine-grained fault location and diagnosis across multiple links and devices. The solution also incorporates end-to-end network penetration detection to enhance the accuracy of user experience perception and ensure network stability in scenarios such as high-definition video, gaming, and smart homes. A closed-loop dispatching and work order management mechanism, combined with a visual interface, enables rapid problem resolution and effectiveness verification, effectively improving network operational efficiency and user satisfaction, thereby enhancing operators' competitive advantage.

[0197] Example 2:

[0198] like Figure 10 As shown, this embodiment provides a device for controlling the network quality of a fixed-line user, the device comprising:

[0199] An acquisition unit 10 is used to acquire problem data of fixed-line users in network operation;

[0200] Among them, problem data includes perceived problem data reported by users and abnormal performance data monitored by the network side;

[0201] The input unit 20 is connected to the acquisition unit 10 and is used to input the perceived problem data into the user perception poor quality model to obtain the first quality cause at the user perception level; and is also used to input the abnormal performance data into the network poor quality early warning model to obtain the second quality cause at the network performance level;

[0202] A fusion unit 30 is connected to the input unit 20 and is used to fuse the first quality reason and the second quality reason to obtain the root cause of poor quality of the fixed-line user network;

[0203] The implementation unit 40 is connected to the integration unit 30 and is used to implement poor quality dispatch and problem repair according to the root cause of the poor quality, thereby achieving network quality control for fixed-line users.

[0204] As a specific implementation, the fusion unit 30 includes:

[0205] The clustering module is used to cluster all historical quality reasons according to the network topology;

[0206] an identification module, connected to the clustering module, for identifying a potential association between the first quality cause and the second quality cause based on the same cluster grouping;

[0207] A matching module, connected to the identification module, configured to map and match the first quality cause and the second quality cause that have a potential correlation to form a fused quality cause;

[0208] The verification module is connected to the matching module and is used to verify the quality reasons of the fusion based on the two-way closed loop formed by the perception verification from the network side to the user and the feedback verification from the user to the network side, so as to obtain the root cause of the poor quality of the fixed network user network.

[0209] The device in this embodiment can execute the method in embodiment 1.

[0210] Example 3:

[0211] like Figure 11 As shown, this embodiment provides an electronic device, which includes a memory and a processor, wherein a computer program is stored in the memory. When the processor runs the computer program stored in the memory, the processor executes the method for controlling the network quality of fixed-line users according to Example 1.

[0212] It will be understood that the above embodiments are merely exemplary embodiments for illustrating the principles of the present invention, and the present invention is not limited thereto. Those skilled in the art will appreciate that various modifications and improvements can be made without departing from the spirit and substance of the present invention, and such modifications and improvements are also considered to be within the scope of protection of the present invention.

Claims

1. A method for controlling the network quality of fixed-line users, characterized in that: The method comprises the following steps: Obtain problem data on fixed-line users' network operations; The problem data includes the perceived problem data fed back by users and the abnormal performance data monitored by the network side; Inputting the perceived problem data into a user perceived poor quality model to obtain a first quality reason at the user perception level; and inputting the abnormal performance data into a network poor quality early warning model to obtain a second quality reason at the network side performance level; Combining the first quality reason and the second quality reason to obtain a root cause of poor quality of the fixed-line user network; Based on the root causes of poor quality, poor quality dispatch and problem repair are implemented, thereby achieving network quality control for fixed-line users.

2. The method for controlling the network quality of fixed-line users according to claim 1, characterized in that: Before inputting the perception problem data into the user perception quality difference model, the method further comprises the steps of: constructing the user perception quality difference model; The construction of the user perceived quality difference model specifically includes the following steps: Obtaining user perceived quality difference data of different platforms, different types, and different dimensions to obtain a first data set; performing data preprocessing on the first data set to obtain a second data set; Dividing the second data set into multiple category indicators according to network levels and business scenarios; Filter out key factors influencing user perception from the category indicators; According to the preset user perceived quality difference threshold, set the perceived quality difference judgment rule corresponding to the key influencing factors; All perceived quality difference judgment rules are associated with all user perceived quality differences to form a user perceived quality difference model.

3. The method for controlling the network quality of fixed-line users according to claim 2, characterized in that: The step of screening out key factors influencing user perception from the category indicators specifically includes the following steps: Standardizing the category indicators to obtain standardized data; Performing eigenvalue decomposition according to the covariance matrix of the standardized data to obtain eigenvectors and eigenvalues; Arrange the eigenvalues ​​in descending order, and select the top K eigenvalues ​​as principal components based on the cumulative contribution rate; Obtaining the main influencing indicators of user perception based on the eigenvectors corresponding to the principal components; The main influencing indicators of user perception are used as key influencing factors of user perception.

4. The method for controlling the network quality of fixed-line users according to claim 1, characterized in that: Before inputting the abnormal performance data into the network poor quality early warning model, the method further comprises the steps of: constructing a network poor quality early warning model; The construction of the network quality poor warning model specifically includes the following steps: Obtain network performance data; Dividing the network performance data into BRAS index performance data, SW index performance data, MER index performance data, MAR index performance data, OLT index performance data, PON index performance data, and ONU index performance data; According to a preset BRAS indicator threshold, a first network poor quality determination rule corresponding to the BRAS indicator performance data is set; and, according to a preset SW indicator threshold, a second network poor quality determination rule corresponding to the SW indicator performance data is set; and, according to a preset MER indicator threshold, a third network poor quality determination rule corresponding to the MER indicator performance data is set; and, according to a preset MAR indicator threshold, a fourth network poor quality determination rule corresponding to the MAR indicator performance data is set; and, according to a preset OLT indicator threshold, a fifth network poor quality determination rule corresponding to the OLT indicator performance data is set; and, according to a preset PON indicator threshold, a sixth network poor quality determination rule corresponding to the PON indicator performance data is set; and, according to a preset ONU indicator threshold, a seventh network poor quality determination rule corresponding to the ONU indicator performance data is set; a first network adjustment rule for intelligent dynamic adjustment is set based on the first network poor quality determination rule; a second network adjustment rule for intelligent dynamic adjustment is set based on the second network poor quality determination rule; a third network adjustment rule for intelligent dynamic adjustment is set based on the third network poor quality determination rule; a fourth network adjustment rule for intelligent dynamic adjustment is set based on the fourth network poor quality determination rule; a fifth network adjustment rule for intelligent dynamic adjustment is set based on the fifth network poor quality determination rule; a sixth network adjustment rule for intelligent dynamic adjustment is set based on the sixth network poor quality determination rule; and a seventh network adjustment rule for intelligent dynamic adjustment is set based on the seventh network poor quality determination rule; The BRAS indicator performance data, the first network poor quality determination rule, and the first network adjustment rule are associated to form a BRAS indicator network poor quality early warning model; and the SW indicator performance data, the second network poor quality determination rule, and the second network adjustment rule are associated to form a SW indicator network poor quality early warning model; the MER indicator performance data, the third network poor quality determination rule, and the third network adjustment rule are associated to form a MER indicator network poor quality early warning model; the MAR indicator performance data, the fourth network poor quality determination rule, and the fourth network adjustment rule are associated to form a MAR indicator network poor quality early warning model; the OLT indicator performance data, the fifth network poor quality determination rule, and the fifth network adjustment rule are associated to form an OLT indicator network poor quality early warning model; the PON indicator performance data, the sixth network poor quality determination rule, and the sixth network adjustment rule are associated to form a PON indicator network poor quality early warning model; the ONU indicator performance data, the seventh network poor quality determination rule, and the seventh network adjustment rule are associated to form an ONU indicator network poor quality early warning model; The BRAS indicator network poor quality warning model, the SW indicator network poor quality warning model, the MER indicator network poor quality warning model, the MAR indicator network poor quality warning model, the OLT indicator network poor quality warning model, the PON indicator network poor quality warning model, and the ONU indicator network poor quality warning model are summarized to obtain a network poor quality warning model.

5. The method for controlling the network quality of fixed-line users according to claim 1, characterized in that: After the poor quality dispatch and problem repair are implemented, the method further comprises the steps of: Verify the data of fixed-line user network operation based on the repair results after the problem is repaired.

6. The method for controlling the network quality of fixed-line users according to any one of claims 1 to 5, characterized in that: The fusing of the first quality reason and the second quality reason to obtain the root cause of poor quality of the fixed-line user network specifically includes the following steps: According to the network topology, all historical quality reasons are clustered and grouped; identifying a potential correlation between the first quality cause and the second quality cause based on the same clustering, and mapping and matching the first quality cause and the second quality cause with the potential correlation to form a fused quality cause; Based on the bidirectional closed loop formed by the perception verification from the network side to the user and the feedback verification from the user to the network side, the quality reasons of the fusion are verified to obtain the root cause of the poor quality of the fixed-line user network.

7. The method for controlling the network quality of fixed-line users according to claim 6, characterized in that: Before clustering all quality reasons according to the network topology, the method further includes the following steps: Build network topology; The construction of the network topology structure specifically includes the following steps: Determine the network element information of each network device; According to the home network element information, setting the network device as a topology node of a network topology structure; Identify the link relationship between the topology nodes using the neighbor information exchanged in the LLDP protocol; The topological nodes and the link relationships are integrated to obtain an end-to-end network topology structure.

8. A device for controlling the network quality of fixed-line users, characterized in that: include: An acquisition unit, used to acquire problem data of fixed-line users in network operation; The problem data includes the perceived problem data fed back by users and the abnormal performance data monitored by the network side; an input unit, connected to the acquisition unit, configured to input the perceived problem data into a user perceived poor quality model to obtain a first quality reason at the user perception level; and further configured to input the abnormal performance data into a network poor quality early warning model to obtain a second quality reason at the network performance level; a fusion unit connected to the input unit, configured to fuse the first quality reason and the second quality reason to obtain a root cause of poor quality of the fixed-line user network; The implementation unit is connected to the fusion unit and is used to implement poor quality dispatch and problem repair according to the root cause of the poor quality, thereby achieving the management and control of the network quality of fixed-line users.

9. The device for controlling the network quality of fixed-line users according to claim 8, characterized in that: The fusion unit comprises: The clustering module is used to cluster all historical quality reasons according to the network topology; an identification module, connected to the clustering module, configured to identify a potential association between the first quality cause and the second quality cause based on the same clustering; a matching module, connected to the identification module, configured to map and match the first quality reasons and the second quality reasons that have potential association to form a fused quality reason; The verification module is connected to the matching module and is used to verify the quality reasons of the fusion based on the two-way closed loop formed by the perception verification from the network side to the user and the feedback verification from the user to the network side, so as to obtain the root cause of the poor quality of the fixed-line user network.

10. An electronic device, characterized in that: The method comprises a memory and a processor, wherein a computer program is stored in the memory, and when the processor runs the computer program stored in the memory, the processor executes the method for controlling the network quality of fixed-line users according to any one of claims 1 to 7.

Images