A Method and System for Proactive Service of Home Broadband Users Based on Network Quality Labels
By generating root cause segments, positive capability and service experience tags for network card usage, and combining them with sliding time window statistics, the problem of insufficient identification of dynamic changes in network quality in existing technologies has been solved. This enables multi-dimensional characterization and dynamic perception of home broadband users, improving the efficiency of service resource allocation and the matching degree of user needs.
Patent Information
- Application Number
- CN202611577376.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-10-13
- Publication Date
- 2026-11-10
AI Technical Summary
Existing technologies cannot effectively identify dynamic changes in the network quality of home broadband users, lack the ability to identify healthy users, and result in a mismatch between service content and user needs, leading to low efficiency in service resource allocation and difficulty in establishing a unified proactive service system for all online users.
Based on multi-source data from home broadband users, root cause tags, positive capability tags, and service experience tags for network lag are generated. Lag events are statistically analyzed using a sliding time window to determine the network quality status. Service priorities are determined through comprehensive quality scoring and service scenario matching, achieving multi-dimensional characterization and dynamic perception.
It enables multi-dimensional profiling of home broadband users, timely detection of changes in network quality trends, accurate targeting of user groups requiring repair or upgrades, improved efficiency in service resource allocation, and the formation of a unified proactive service system covering all online users.
Smart Images

Figure CN123074947A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of network quality management and data-driven proactive service for home broadband users, and in particular to a method and system for proactive service for home broadband users based on network quality tags. Background Technology
[0002] In recent years, the number of home broadband users has continued to grow, and operators have shifted their focus in home broadband service operations from network construction to ensuring user experience and value creation. With tens of millions of users, proactively identifying user network problems, recognizing potential upgrade needs, and reaching users in the right way and at the right time have become key to improving user satisfaction and operational efficiency.
[0003] In existing technologies, some solutions rely on gateway-side device metrics (such as optical power, packet loss rate, and CPU utilization) for threshold judgment, only outputting a binary label indicating poor quality. This fails to distinguish the specific root cause of lag and makes it difficult to detect dynamic changes in user network quality. Other solutions use machine learning to output user experience scores, but these scores are not automatically correlated with service actions, and service decisions still depend on human experience. Furthermore, there are technical approaches in the industry that use experience scores to drive proactive maintenance loops, but these solutions only cover remedial, one-way scenarios and do not incorporate retention-upgrade marketing scenarios into the same technical system. Value mining for healthy users and service restoration remain two parallel systems.
[0004] The common shortcomings of the aforementioned solutions are: Firstly, they use a single dimension to characterize user network quality, focusing only on latency and neglecting positive dimensions such as terminal capabilities and service preferences. This fails to identify users with normal network performance but potential for upgrades, leaving healthy user groups in a service blind spot for extended periods. Secondly, they are slow to perceive changes in user network status, and the static threshold-based judgment method cannot promptly capture the transition from normal to deteriorating network quality. Furthermore, they lack an automatic mapping and prioritization mechanism between quality evaluation results and service scenarios, resulting in a mismatch between service content and actual user needs and low efficiency in service resource deployment. These shortcomings make it difficult for existing technologies to establish a unified proactive service system covering all online users and to achieve precise proactive service for a large-scale home broadband user base. Summary of the Invention
[0005] In view of the above problems, a proactive service method and system for home broadband users based on network quality labels is proposed to overcome or at least partially solve the above problems, including:
[0006] Based on multi-source data from home broadband users, generate root cause segment labels for network card usage, positive capability labels, and service experience labels;
[0007] Based on the root cause segment label, positive capability label, and service experience label of network lag, the first and second sliding time windows are used to count lag events and determine the network quality status of home broadband users. The network quality status includes no lag, deterioration, mitigation, stable lag, excellent and stable, and improving status.
[0008] The overall quality score is calculated based on the network card root cause segment label and positive capability label;
[0009] Based on the root cause segment label of network lag, positive capability label, service experience label and network quality status, the target service scenario is matched from the repair service scenario and the retention and upgrade service scenario.
[0010] Service priorities are determined based on network quality status, comprehensive quality score, and target service scenarios, and service instructions are sent to home broadband users according to these priorities.
[0011] Optionally, the network card lag root cause segment label includes lag indicators for six fault segments: user terminal segment, routing device segment, LAN side link segment, gateway WiFi segment, gateway segment, and Internet segment.
[0012] The lag metrics for the faulty segment include the number of lags, the number of lag days, the maximum number of consecutive lag days, the top terminal type, the top application name, and the top peak time period.
[0013] Optional positive capability labels include gigabit ready label, FTTR whole-house coverage label, large screen high definition label, 5G terminal ownership label, smart terminal ecosystem label, and device new / old label;
[0014] The gigabit-ready flag is determined based on the router's WiFi protocol version or LAN port speed;
[0015] The FTTR whole-house coverage label is determined based on the package order type or equipment model.
[0016] Optionally, the first sliding time window has a duration of 7 days, and the second sliding time window has a duration of 30 days.
[0017] Determine the network quality status of home broadband users, including:
[0018] The system records the number of buffering incidents, the number of buffering days, and the maximum number of consecutive buffering days for home broadband users within the first sliding time window, as well as the number of buffering incidents, the number of buffering days, and the maximum number of consecutive buffering days for home broadband users within the second sliding time window.
[0019] Calculate the ratio of the average daily number of stutters in the first sliding time window to the average number of stutters in the second sliding time window;
[0020] When the ratio is greater than the preset deterioration threshold and the maximum number of consecutive stuck days within the first sliding time window is greater than or equal to the preset consecutive days threshold, the network quality status is determined to be in a deteriorated state.
[0021] When the ratio is less than the preset mitigation threshold and there is a stuttering record within the second sliding time window, the network quality status is determined to be in a mitigation state.
[0022] When the number of stutters within the second sliding time window is zero, the number of tags with a value of 1 in the positive capability tag is greater than or equal to 2, and the household value index is greater than or equal to the preset index threshold, the network quality status is determined to be excellent and stable.
[0023] When the number of stutters within the second sliding time window is zero, and the number of tags with a value of 1 in the positive capability tag is less than 2 or the household value index is less than the preset index threshold, the network quality status is determined to be a stutter-free state.
[0024] If the number of stutters within the second sliding time window is not zero, the home broadband user does not meet the criteria for the deterioration state or the mitigation state. If the network quality status was a stable stuttering state or an improving state in the last status update, and the number of consecutive zero stuttering days reaches the preset confirmation number of days, the network quality status is determined to be an improving state.
[0025] When the network quality status does not fall into the categories of deterioration, mitigation, excellent and stable, no lag, or improving, the network quality status is determined to be a stable lag state.
[0026] Among them, the Family Value Index is a component of the Business Experience Label, used to characterize the value level of home equipment for home broadband users.
[0027] Optionally, determining the network quality status of home broadband users also includes a status anti-jitter mechanism:
[0028] When a home broadband user switches from the current state to the target state, the entry conditions for the target state must be met for a preset number of consecutive confirmation days.
[0029] After the state transition is completed, a cooldown period is set, during which it is prohibited to switch back to the state before the transition.
[0030] When a user complaint occurs or the overall quality score falls below the preset mandatory coverage threshold, the confirmation period and cooling-off period are ignored, and the network quality status is forcibly covered to a deteriorated state.
[0031] Optionally, the calculation of the overall quality score includes:
[0032] The experience score is calculated based on the lag index in the root cause segment label of network lag. The experience score is inversely related to the number of lags, the number of lag days, and the maximum number of consecutive lag days.
[0033] Ability scores are calculated based on the values of each identifier in the positive ability labels and preset weights;
[0034] A stable score is calculated based on the day penalty for exceeding a preset burst threshold for the number of times a single day experiences lag, and the difference between the first sliding time window and the second sliding time window in the experience score.
[0035] The overall quality score is obtained by weighting and summing the experience score, ability score, and stability score.
[0036] The weights of the experience score, the capability score, and the stability score are adjusted according to the network quality status.
[0037] Optionally, the method for determining the TOP1 root cause segment among the six fault segments is as follows:
[0038] Sort the number of stutters in the six fault segments;
[0039] When there are two or more fault segments with the same number of pauses, the maximum number of consecutive pause days, the significance during peak periods, and the repairability level of the fault segments within the corresponding sliding time window are compared in turn to determine the TOP1 root cause segment.
[0040] The repairability levels, from highest to lowest, are as follows: LAN-side link segment, gateway WiFi segment, and routing device segment are tied for the highest level, followed by user terminal segment, gateway segment, and Internet segment.
[0041] Optional, repair service scenarios include replacing old routers, adding FTTR sub-routers, adding Tianyi gateway sub-routers, and optimizing large-screen video stuttering.
[0042] Upgraded service scenarios include gigabit speed upgrade recommendation scenario, FTTR whole-house upgrade scenario, smart home product expansion scenario, and loyalty enhancement scenario.
[0043] Optionally, service priorities can be determined based on network quality status, overall quality score, and target service scenario, including:
[0044] When the network quality status is deteriorating, the overall quality score is lower than the first threshold, and the target service scenario is a repair service scenario, the service priority is determined to be the first priority.
[0045] When the network quality status is deteriorating, the comprehensive quality score is higher than or equal to the first threshold, and the target service scenario is a repair service scenario, the service priority is determined to be the second priority.
[0046] When the network quality status is stable but laggy, the overall quality score is lower than the second threshold, and the target service scenario is a repair service scenario, the service priority is determined to be the second priority.
[0047] When the network quality status is stable but laggy, the comprehensive quality score is higher than or equal to the second threshold, and the target service scenario is a repair service scenario or a retention upgrade service scenario, the service priority is determined to be the third priority.
[0048] When the network quality status is in a mitigated state, the overall quality score is higher than the third threshold, and no service scenario is hit, the service priority is determined to be the fourth priority.
[0049] When the network quality status is excellent and stable, the comprehensive quality score is higher than the fourth threshold, and the target service scenario belongs to the retention and upgrade service scenario, the service priority is determined to be the fifth priority.
[0050] When the network quality status is under improvement, the overall quality score is higher than the preset under improvement score threshold, and the target service scenario belongs to the retention-upgrade service scenario, the service priority is determined to be the fifth priority; where the preset under improvement score threshold is lower than the fourth threshold.
[0051] The urgency of services decreases sequentially from first priority to fifth priority.
[0052] A proactive service system for home broadband users based on network quality labels includes:
[0053] The tag calculation module is used to generate network lag root cause tags, positive capability tags, and service experience tags based on multi-source data of home broadband users; based on these tags, it performs lag event statistics using a first and second sliding time window to determine the network quality status of home broadband users, including no lag, deterioration, mitigation, stable lag, excellent and stable, and improving status; and it calculates a comprehensive quality score based on the network lag root cause tags and positive capability tags.
[0054] The scenario matching engine is used to match target service scenarios from repair service scenarios and retention and upgrade service scenarios based on network card failure root cause segment tags, positive capability tags, service experience tags, and network quality status.
[0055] The service decision module is used to determine service priorities based on network quality status, comprehensive quality score, and target service scenarios.
[0056] The reach execution module is used to send service instructions to home broadband users according to service priority.
[0057] This invention overcomes the limitations of existing technologies that only focus on the presence or absence of network lag by introducing three types of tags: root cause segment of network lag, positive capability, and service experience. Specifically, the positive capability tag identifies users with normal network status but upgrade potential, allowing for precise targeting of healthy user groups with upgrade-oriented services; the root cause segment of network lag tag pinpoints the specific fault segment and root cause of lag, providing a clear basis for remedial services; and the service experience tag further supplements user preferences and household characteristics, supporting differentiated matching of service scenarios. Thus, a multi-dimensional characterization of all household broadband users is achieved within a single technical system, covering both users experiencing lag requiring remediation and healthy users requiring maintenance, filling the gaps in existing technologies regarding positive capability identification and healthy user operation.
[0058] Based on statistical analysis using a dual sliding time window and six-state network quality status identification, the changing trends in user network quality can be perceived in a timely manner, including dynamic processes such as deterioration, mitigation, and improvement. This allows for proactive service triggering before users perceive significant deterioration, overcoming the lag problem caused by existing technologies relying on static threshold judgments. Furthermore, by combining network quality status, comprehensive quality score, and target service scenarios to determine service priorities, quality evaluation results can be automatically mapped to specific repair or preservation / upgrade service actions. This improves the matching degree between service content and actual user needs, and also improves the efficiency of service resource deployment, ultimately forming a unified proactive service system covering all online users. Attached Figure Description
[0059] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the description of the present invention will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0060] Figure 1 This is an overall architecture diagram of a home broadband user proactive service system based on network quality tags provided in an embodiment of the present invention;
[0061] Figure 2 This is a diagram of the rule structure of the scene matching engine provided in an embodiment of the present invention;
[0062] Figure 3 This is a flowchart of the label generation process provided in an embodiment of the present invention;
[0063] Figure 4 This is a flowchart of dual-window statistics and network quality status determination provided in an embodiment of the present invention;
[0064] Figure 5 This is a flowchart of the comprehensive quality score calculation provided in an embodiment of the present invention. Detailed Implementation
[0065] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.
[0066] Reference Figures 1 to 5 This invention provides a proactive service method for home broadband users based on network quality tags, which may specifically include:
[0067] Based on multi-source data from home broadband users, generate root cause segment labels for network card usage, positive capability labels, and service experience labels;
[0068] Based on the root cause segment label, positive capability label, and service experience label of network lag, the first and second sliding time windows are used to count lag events and determine the network quality status of home broadband users. The network quality status includes no lag, deterioration, mitigation, stable lag, excellent and stable, and improving status.
[0069] The overall quality score is calculated based on the network card root cause segment label and positive capability label;
[0070] Based on the root cause segment label of network lag, positive capability label, service experience label and network quality status, the target service scenario is matched from the repair service scenario and the retention and upgrade service scenario.
[0071] Service priorities are determined based on network quality status, comprehensive quality score, and target service scenarios, and service instructions are sent to home broadband users according to these priorities.
[0072] The method provided in this invention can be executed in a home broadband operation support system, involving multiple stages such as data collection, tag calculation, scenario matching, and service delivery. This method generates three types of tags based on multi-source data from home broadband users: root cause tags for network card usage issues, positive capability tags, and service experience tags. Based on these three types of tags, it dynamically identifies the network quality status and calculates a comprehensive quality score, thereby matching repair-oriented or retention-upgrade service scenarios, ultimately determining service priorities, and sending service instructions to users.
[0073] Specifically, the multi-source data originates from various data records accumulated by operators during the operation of home broadband services. This data includes lag event data, terminal capability data, and service experience data. Lag event data records lag events that occur when home broadband users use specific applications at specific times. Each event record includes the time of the lag, application name, terminal type and model, gateway media access control address, device media access control address, the fault segment to which the lag belongs, and the specific root cause. Terminal capability data includes information such as home gateway device model, wireless router protocol version, LAN port speed, set-top box resolution, terminal launch date, and the type and quantity of smart home devices. Service experience data includes users' recent application usage records, the service category of the application, the number of frequently used devices in the home, device activity periods, household spending level, and a household value index calculated by weighting device value. The above data can be stored in an offline data warehouse or a real-time data platform. The system collects and preprocesses the data according to a preset cycle. The preprocessing process includes field standardization, null value imputation, and outlier filtering to ensure the quality of the input data for subsequent label calculations.
[0074] Based on the collected multi-source data, the system generates three types of network quality labels.
[0075] The first type uses network lag root cause segment tags. These tags aggregate and statistically analyze lag event records in lag event data at the granularity of the broadband account of the home broadband user. They output information such as the number of lags, the number of lag days, the maximum number of consecutive lag days, the main terminal types and models involved, the main application names, and the peak time of the lag within a preset statistical period. At the same time, they retain the fault segment to which the lag belongs and the specific root cause dimension, so as to locate the cause of the lag in the future.
[0076] The second category is positive capability tags. These tags extract upgrade conditions for devices in the user's home network from terminal capability data. Examples include the router's supported wireless router protocol version, LAN port speed, whether whole-house fiber optic networking is possible, the TV's supported resolution, whether the user owns a terminal supporting 5G technology, the types and quantities of smart home devices, and the device's market launch date. This information is then processed according to rules and converted into an identifier field. The identifier field can have a value of 0 or 1, indicating whether the user possesses the corresponding positive capability.
[0077] The third category is business experience tags. These tags extract user business usage characteristics and family characteristics from business experience data, including business preference type, whether the family is a high-concurrency household, device activity time characteristics, whether the family is a large family with multiple devices, whether the family is a candidate for a large house, family value index, and consumption level. Among them, business preference type can be determined based on the user's usage frequency or distribution of lag events in each application category within a preset statistical period; high-concurrency households are judged by counting the number of devices that are frequently used in the family; and the family value index is obtained by weighting the value and level of the devices connected to the family.
[0078] After the three types of tags are generated, the system uses the first and second sliding time windows to statistically analyze the lag events to determine the network quality status of home broadband users. The first and second sliding time windows correspond to different time spans. For example, the first sliding time window can take a relatively short period of time to capture short-term changes in the user's network quality; the second sliding time window can take a longer period of time to depict the long-term trend of the user's network quality.
[0079] For each household broadband user, the system statistically analyzes metrics such as the number of buffering incidents, the number of buffering days, and the maximum number of consecutive buffering days within the first and second sliding time windows, and calculates the changes in these metrics between the two windows. For example, it can calculate the ratio of the average daily number of buffering incidents between the first and second sliding time windows, which reflects the direction of change in the user's recent buffering frequency relative to the long-term average level.
[0080] Simultaneously, the system also considers the values of the identifier fields in the positive capability label and the family value index in the business experience label to comprehensively determine which of the six network quality states the user belongs to. The six network quality states are: No Lag, Deteriorating, Alleviated, Stable Lag, Excellent Stable, and Improving. No Lag indicates that the user did not experience any lag events during the statistical period; Deteriorating indicates that the user's recent lag has significantly worsened compared to the long-term trend; Alleviated indicates that the user's lag is decreasing; Stable Lag indicates that the user has persistent lag issues but no significant recent changes; Excellent Stable indicates that the user has no lag issues and possesses good positive capabilities; Improving indicates that the user is transitioning from persistent lag to a better state. The system determines the state based on the definitions of each state and the user's statistical indicators, and uses the determination results as one of the inputs for subsequent scenario matching and priority determination.
[0081] The system also needs to calculate a comprehensive quality score based on the root cause segment label and positive capability label of the network card. The comprehensive quality score is used to comprehensively reflect the network quality level of home broadband users, and the calculation is evaluated from the experience dimension, capability dimension, and stability dimension.
[0082] The experience dimension is mapped based on metrics such as the number of times the network lag occurred, the number of days the network lag occurred, and the maximum number of consecutive days of lag in the root cause segment tag. The more times the network lag occurred, the more days the network lag occurred, or the longer the consecutive days of lag occurred, the lower the experience score will be.
[0083] The capability dimension is calculated by weighting the values of each identifier field in the positive capability label and the corresponding preset weights to obtain a capability score, which is used to reflect the user's potential for upgrading terminal capabilities.
[0084] The stability dimension is calculated by penalizing the changes in the experience score between the first and second sliding time windows, as well as the number of sudden lags, to obtain a stability score. Finally, the experience score, capability score, and stability score are weighted and summed according to preset weighting coefficients to obtain the comprehensive quality score. The comprehensive quality score ranges from 0 to 100, with higher values indicating better overall network quality for the user. The comprehensive quality score can be used to determine subsequent service priorities to differentiate the urgency and service needs of different users.
[0085] After completing the identification of network quality status and the calculation of comprehensive quality score, the system matches the target service scenario from the repair service scenario and the retention and upgrade service scenario based on the root cause segment label of network lag, positive capability label, service experience label and network quality status.
[0086] Repair-type service scenarios refer to network repair services provided to users experiencing lag issues, such as replacing old routers, adding whole-house fiber optic sub-routers, optimizing gateway wireless coverage, or resolving video buffering on large screens. Upgrade-type service scenarios refer to package upgrades or value-added services offered to users with good network quality but with upgrade potential, such as gigabit broadband speed increases, whole-house fiber optic network upgrades, expansion of smart home device offerings, or high-value user loyalty maintenance services.
[0087] The system pre-configures trigger conditions for each service scenario, defined based on the user's tag set and network quality status. When a user's tag value and status meet the trigger conditions for a scenario, that scenario is selected, and the system uses it as the target service scenario. If a user meets the trigger conditions for multiple scenarios simultaneously, the system can select a target service scenario according to a preset scenario priority mutual exclusion rule to ensure the targeting of service actions and the accuracy of resource delivery.
[0088] Finally, the system determines service priorities based on network quality status, overall quality score, and target service scenario, and sends service instructions to home broadband users according to these priorities. Service priorities are used to differentiate the urgency and processing order of different users during proactive service.
[0089] For example, users in a deteriorating state with a low overall quality score in a recovery scenario can be given the highest service priority, with outbound calls or on-site services prioritized; users in a stable but lagging state in a recovery scenario can be given a medium service priority, reaching them via app push or SMS; users who do not meet any scenario triggering conditions can be left unreached; and users in a good and stable state or an improving state who fall into the retention / upgrade scenario can be given a lower service priority, reaching them via app push or SMS for marketing purposes.
[0090] Based on service priority, the system generates corresponding service instructions and sends them to users through preset channels. These channels may include SMS, carrier app push notifications, customer service outbound calls, on-site installation and maintenance personnel work orders, and WeChat Work, with different priorities corresponding to different channels and delivery times.
[0091] Through the above steps, the system realizes multi-dimensional quality evaluation, status trend perception, automatic scenario matching, and tiered proactive service for home broadband users, thereby intervening and repairing before users' perceptions deteriorate significantly, and proactively managing the value of healthy users.
[0092] In one or more embodiments of the present invention, the network card lag root cause segment label includes lag indicators of six fault segments, namely user terminal segment, routing device segment, LAN side link segment, gateway WiFi segment, gateway segment and Internet segment;
[0093] The lag metrics for the faulty segment include the number of lags, the number of lag days, the maximum number of consecutive lag days, the top terminal type, the top application name, and the top peak time period.
[0094] Using root cause segment labels for network lag as a key basis for locating the root cause of lag issues in home broadband users requires a systematic division of lag events according to the location of the fault. This label divides potential lag faults during home broadband use into six fault segments: user terminal segment, routing device segment, LAN-side link segment, gateway WiFi segment, gateway segment, and internet segment.
[0095] The user terminal segment covers user-side terminal devices connected to the home network, such as mobile phones, tablets, laptops, desktop computers, smart TVs, and set-top boxes, whose own issues can cause lag. The routing device segment covers faults related to independently deployed wireless routers within the home, such as routers only supporting older wireless protocols or malfunctions causing lag. The LAN link segment covers transmission quality issues within the home LAN, such as high jitter, high packet loss, and high latency causing lag.
[0096] The gateway WiFi segment covers faults related to the built-in wireless LAN function of the home gateway device, such as lag caused by insufficient WiFi signal coverage, channel interference, or lack of dual-band support. The gateway segment also covers faults related to the operational status of the home gateway device itself, such as lag caused by full bandwidth usage, high memory consumption, high processor temperature, abnormal optical module power, abnormal gateway disconnection, gateway speed limiting, excessive gateway operation time, too many connected devices, or high optical module temperature. The internet segment covers lag caused by internet-side services or links after the home broadband connection is established, such as lag caused by abnormal response of domestic application services, internet link problems, abnormal domain name resolution service response, or access to foreign applications.
[0097] The six fault segments together constitute a complete root cause localization dimension for the lag. Each fault segment corresponds to a field prefix, and the statistical indicators of different fault segments are distinguished by the field prefix in the user root cause segment lag details wide table.
[0098] The generation process of the root cause segment label for network lag is based on the original lag event records in the network lag event log. Each lag event record includes the time of lag occurrence, application name, terminal type, terminal model, gateway media access control address, device media access control address, fault segment identifier to which the lag belongs, and specific root cause identifier.
[0099] The system aggregates data based on home broadband users' broadband accounts, groups and statistically analyzes raw lag event records according to fault segments, and generates lag metrics for each fault segment. The lag metrics for each of the six fault segments include the number of lags, the number of lag days, the maximum number of consecutive lag days, the top terminal type, the top application name, and the top peak time period.
[0100] The number of stutters represents the total number of stutter events that occurred within a certain fault segment during a preset statistical period. This is obtained by counting the number of stutter event records corresponding to the fault segment identifier. The number of stutter days represents the number of natural days during which stutter events occurred within a certain fault segment during a preset statistical period. This is obtained by extracting the date from the timestamp information in the stutter event records and then deduplicating the count. The maximum number of consecutive stutter days represents the maximum number of natural days during which stuttering occurred consecutively within a certain fault segment during a preset statistical period. This requires sorting stutter events by date, identifying consecutive stutter intervals, and then taking the maximum value among the lengths of each consecutive interval.
[0101] When aggregating the above three metrics at the user level, window functions and row number difference methods from offline data warehouses can be used for calculation. The principle of the row number difference method is as follows: after grouping the stuttering dates of each fault segment by user and sorting them, each stuttering date record corresponds to a row number. The difference is obtained by subtracting the row number from the date sequence number. Records with consecutive dates correspond to the same difference, thus identifying consecutive stuttering intervals. The maximum consecutive stuttering days can then be obtained by taking the maximum value of the interval length.
[0102] The above calculations can generate corresponding fields in the user root cause segment lag details wide table. For example, the field name for the number of lags is formatted as fault segment prefix plus lag number field name, the field name for the number of lag days is formatted as fault segment prefix plus day statistics field name, and the field name for the maximum consecutive lag days is formatted as fault segment prefix plus consecutive days segment name.
[0103] The top-ranked device type, top-ranked application name, and top-ranked peak time period in the lag metrics for a fault segment are used to characterize the most prevalent lag issues within that segment. The top-ranked device type is determined by statistically analyzing the distribution of device types experiencing lag events within that fault segment. Device types include categories such as mobile phones, tablets, laptops, desktop computers, smart TVs, and set-top boxes. If multiple device types experience the same number of lag events within the same fault segment, the top-ranked device type can be determined based on the most recent lag occurrence time, the device type's activity weight within the user's household, or a preset device priority order.
[0104] The top application name is obtained by statistically analyzing the distribution of application names that experienced stuttering events within the fault segment, such as specific application names like video players, game applications, office software, and educational applications. If multiple application names have the same number of stuttering events, the top application name can be determined based on the application's most recent stuttering time or the application's usage frequency weight.
[0105] The TOP1 peak period is obtained by statistically analyzing the distribution of time periods during which lag events occur within the fault segment. A day can be divided into early morning, morning, afternoon, evening peak, and nighttime periods. If multiple time periods have the same number of lag events, the TOP1 peak period can be determined according to the degree of impact of the time period or the preset time period priority order.
[0106] In addition, the lag metrics for the fault segment can also include detailed information such as terminal type distribution, application name distribution, and time period distribution, stored in JSON format in the user root cause segment lag detail wide table for more granular judgment during subsequent scenario matching. The above statistical process is executed in the system's daily offline tag calculation task, recalculating the metrics within the most recent preset statistical period daily to ensure the timeliness of the tags.
[0107] By generating six fault segment lag indicators, the system can identify whether a user has experienced lag, and further pinpoint the specific location of the lag within the home network and the main lag characteristics at that location. This provides a clear root cause basis for subsequent service scenario matching and proactive service.
[0108] In one or more embodiments of the present invention, the positive capability label includes a gigabit ready identifier, an FTTR whole-house coverage identifier, a large-screen high-definition identifier, a 5G terminal ownership identifier, a smart terminal ecosystem identifier, and a device new / old file identifier.
[0109] The gigabit-ready flag is determined based on the router's WiFi protocol version or LAN port speed;
[0110] The FTTR whole-house coverage label is determined based on the package order type or equipment model.
[0111] Positive capability tags are used to characterize the upgrade conditions or potential capabilities of home broadband users in terms of terminals and network equipment. The generation of positive capability tags is based on the home terminal capability dimension table and the downstream device detail wide table.
[0112] The Home Terminal Capability Dimension Table records attributes such as device model, device type, wireless protocol version, LAN port speed, TV resolution, release date, and whether it supports 5G mobile communication networks for each home broadband account. The Connected Device Details Wide Table records detailed information about connected devices in the user's home, including device type, device model, number of devices, and specific types of smart home devices. The system associates these two types of data at the broadband account level to generate values for each identifier field in the positive capability label. Each identifier field can have a value of 0 or 1; a value of 1 indicates that the user possesses the corresponding positive capability, and a value of 0 indicates that the user does not. The specific determination process for each identifier is as follows.
[0113] The gigabit-ready flag is determined based on the router's WiFi protocol version or LAN port speed. In this embodiment, the wireless router device record in the home terminal capability dimension table includes a `wifi_protocol` field, which indicates the wireless protocol version supported by the router. A value of 6 indicates support for WiFi 6, and a value of 7 indicates support for WiFi 7. The home terminal capability dimension table also records a `router_lan_rate` field, which indicates the wired transmission rate of the router's LAN port, in megabits per second.
[0114] The gigabit ready flag is set to 1 when at least one router in the user's home has a wifi_protocol field value of 6 or 7. The gigabit ready flag is also set to 1 when at least one router in the user's home has a router_lan_rate field value greater than or equal to 1000. If neither of these conditions is met, the gigabit ready flag is set to 0.
[0115] It's important to note that a value of 6 or 7 in the `wifi_protocol` field indicates that the router can support gigabit or higher broadband access, while a value greater than or equal to 1000 in the `router_lan_rate` field indicates that the router has gigabit wired access capability. Therefore, meeting either condition is sufficient to determine that the user's home network is gigabit-ready, which can be used as a preliminary judgment for gigabit speed-up recommendation scenarios. For users with multiple routers, the system checks each router separately, and as long as one meets the conditions, the user is considered gigabit-ready.
[0116] The FTTR whole-home coverage identifier is determined based on the subscription package type or equipment model. The FTTR whole-home coverage identifier is used to identify whether a user already has or is using whole-home fiber optic networking capabilities. The subscription package type is recorded in the user's subscription package information table. When the value of the subscription package type field falls within the preset FTTR subscription package type set, such as FTTR or FTTRJB, the FTTR whole-home coverage identifier is set to 1.
[0117] In addition, the system can also determine the model based on the user's home gateway device. The system maintains a database of FTTR main optical modem device models, containing multiple models such as HG3142F, HN8156XR, HN8145XR, G7615V2, G7615TV2, G7615, TYFZA1, F4607P, F7607P, V173, V271E, V176, and TY-F802G. The system matches the user's home gateway device model field against models in the database. The matching method involves converting the device model field to uppercase and then using string inclusion relationships for comparison. When the user's home gateway device model matches any model in the database, the FTTR whole-house coverage flag is set to 1.
[0118] Meanwhile, Tianyi Gateway 3.0 and Tianyi Gateway 4.0 devices are identified separately. When the Tianyi Gateway 3.0 and Tianyi Gateway 4.0 device models appear in the user's home gateway device model field, they are not directly identified as FTTR whole-house coverage. Instead, they are used as the basis for determining the gateway type in subsequent FTTR sub-router installation scenarios. The value of the FTTR whole-house coverage identifier is ultimately determined by the package subscription type or the device model matching result; as long as either condition is met, the value is 1.
[0119] The large-screen high-definition identifier is determined based on the resolution supported by the television device. The home terminal capability dimension table records a `tv_resolution` field, indicating the resolution specifications supported by the television devices in the user's home. When the value of the `tv_resolution` field is 4K or 8K, the value of the large-screen high-definition identifier is 1. In addition, the home terminal capability dimension table also records a `tv_4k_cnt` field, indicating the number of television devices in the user's home that support 4K resolution. When the value of the `tv_4k_cnt` field is greater than or equal to 1, the value of the large-screen high-definition identifier is also 1. When neither of the above two conditions is met, the value of the large-screen high-definition identifier is 0.
[0120] This identifier is used to identify whether a user's home television device has high-definition display capabilities, and can be used for preliminary judgment in scenarios where large-screen video experience optimization is required. In cases where multiple television devices exist in a user's home, the system checks the tv_resolution field or calculates the tv_4k_cnt field for each television device. If any television device meets the 4K or 8K resolution condition, it is determined to have large-screen high-definition capabilities.
[0121] The 5G terminal holding identifier is determined based on whether the user's household has a terminal device that supports 5G mobile communication networks. The household terminal capability dimension table records a field called `phone_5g_support`, which indicates whether the user's mobile device supports 5G mobile communication networks. A value of 1 indicates support, and a value of 0 indicates no support. When at least one mobile device in the user's household has a `phone_5g_support` value of 1, the 5G terminal holding identifier is set to 1. When no mobile device in the user's household supports 5G mobile communication networks, the 5G terminal holding identifier is set to 0. This identifier is used as a preliminary judgment for subsequent 5G mobile communication package recommendation scenarios. The system iterates through all mobile devices in the user's household; if even one mobile device meets the condition, it is determined to hold a 5G terminal.
[0122] The smart terminal ecosystem identifier is determined based on the number and types of smart home devices in a user's home. The wide table detailing connected devices records the device type field for various smart home devices in the user's home, including cameras, door locks, doorbells, and speakers. The system counts the types of smart home devices present in the user's home; different device types are considered different categories, and multiple devices of the same type are counted as only one category.
[0123] When the number of smart home devices in a user's home is greater than or equal to a preset threshold, the smart terminal ecosystem identifier is set to 1. The preset threshold can be set in the system configuration; for example, setting it to 3 types means that a user's home is considered to have a smart terminal ecosystem only if at least three types of smart home devices—cameras, door locks, and doorbells—are present. When the number of smart home devices in a user's home is less than the preset threshold, the smart terminal ecosystem identifier is set to 0. This identifier is used to identify whether a user's home has initially formed a smart home ecosystem and can be used for preliminary judgment in smart home expansion scenarios. When counting device types, the system uses the device type field value as the sole classification criterion; multiple devices of the same type are not counted repeatedly.
[0124] The new / old device classification is determined based on the market age of the main network devices in the user's home. The home terminal capability dimension table records the market age field for each device. The system calculates the difference in years between the market age and the current date for each device. Based on this difference, devices are categorized into multiple new / old device classes.
[0125] In this embodiment, devices launched within 2 years of the current date belong to the first tier; devices launched more than 2 years but no more than 5 years ago belong to the second tier; and devices launched more than 5 years ago belong to the third tier. The value of the device's new / old tier identifier is determined based on the highest tier of the main network devices in the user's home. When at least one main network device in the user's home belongs to the first tier, the device's new / old tier identifier is set to 1, indicating the presence of a relatively new device in the user's home. When none of the main network devices in the user's home belong to the first tier, the device's new / old tier identifier is set to 0, indicating that all the main network devices in the user's home are in an older tier. Main network devices may include a home gateway and a router.
[0126] This identifier is used for preliminary judgment in subsequent equipment replacement recommendation scenarios. When calculating the year difference, the system uses a date function to calculate the difference in days between the listing date field and the current date field, and then converts it into a year difference. The listing date field may be in month granularity. For example, if the listing date is January 2023, the year difference is calculated based on the first day of that month.
[0127] After generation, the six positive capability identifiers mentioned above are uniformly stored in the user tag wide table, with the broadband account as the associated primary key. Each identifier field has a value of 0 or 1, and the fields are independent of each other. When calculating capability scores, the system reads the values of each positive capability identifier field in the user tag wide table and performs a weighted calculation according to preset weights. When determining network quality status, the system reads the number of tags with a value of 1 in the positive capability identifiers, serving as one of the auxiliary judgment conditions for excellent and stable status and no lag status. By generating multi-dimensional positive capability tags, the system can identify users with normal network quality but with upgrade potential, thus providing data support for retaining upgrade-oriented service scenarios.
[0128] In one or more embodiments of the present invention, the duration of the first sliding time window is 7 days, and the duration of the second sliding time window is 30 days.
[0129] Determine the network quality status of home broadband users, including:
[0130] The system records the number of buffering incidents, the number of buffering days, and the maximum number of consecutive buffering days for home broadband users within the first sliding time window, as well as the number of buffering incidents, the number of buffering days, and the maximum number of consecutive buffering days for home broadband users within the second sliding time window.
[0131] Calculate the ratio of the average daily number of stutters in the first sliding time window to the average number of stutters in the second sliding time window;
[0132] When the ratio is greater than the preset deterioration threshold and the maximum number of consecutive stuck days within the first sliding time window is greater than or equal to the preset consecutive days threshold, the network quality status is determined to be in a deteriorated state.
[0133] When the ratio is less than the preset mitigation threshold and there is a stuttering record within the second sliding time window, the network quality status is determined to be in a mitigation state.
[0134] When the number of stutters within the second sliding time window is zero, the number of tags with a value of 1 in the positive capability tag is greater than or equal to 2, and the household value index is greater than or equal to the preset index threshold, the network quality status is determined to be excellent and stable.
[0135] When the number of stutters within the second sliding time window is zero, and the number of tags with a value of 1 in the positive capability tag is less than 2 or the household value index is less than the preset index threshold, the network quality status is determined to be a stutter-free state.
[0136] If the number of stutters within the second sliding time window is not zero, the home broadband user does not meet the criteria for the deterioration state or the mitigation state. If the network quality status was a stable stuttering state or an improving state in the last status update, and the number of consecutive zero stuttering days reaches the preset confirmation number of days, the network quality status is determined to be an improving state.
[0137] When the network quality status does not fall into the categories of deterioration, mitigation, excellent and stable, no lag, or improving, the network quality status is determined to be a stable lag state.
[0138] Among them, the Family Value Index is a component of the Business Experience Label, used to characterize the value level of home equipment for home broadband users.
[0139] Specifically, the first sliding time window is set to a duration of 7 days, and the second sliding time window is set to a duration of 30 days. Both windows are based on the current label calculation date and trace back the corresponding number of days. The first sliding time window covers 6 days prior to the current calculation date, totaling 7 calendar days; the second sliding time window covers 29 days prior to the current calculation date, totaling 30 calendar days.
[0140] The system performs batch tag calculation tasks daily. During each calculation, the time range covered by the window is redefined according to the aforementioned window definition, and the lag event data within the window is statistically analyzed. This dual-window design allows both short-term fluctuations and long-term trends to be reflected simultaneously. The 7-day window captures sudden changes in network quality within the past week, while the 30-day window characterizes the overall lag level and persistent characteristics of the user over a one-month period. The statistical indicators from the two windows work together to provide data for determining the subsequent six network quality states.
[0141] In this embodiment, the determination of network quality status is based on the statistical results of lag events for users within two sliding time windows. The system extracts lag event records for each user within the window coverage time range from the user root cause segment lag detail wide table, using the broadband account as the granularity. For each user, the system counts the number of lags, the number of lag days, and the maximum number of consecutive lag days within the first sliding time window, and simultaneously counts the number of lags, the number of lag days, and the maximum number of consecutive lag days within the second sliding time window.
[0142] The number of stutters is the total number of stutter events recorded within the user's window time range; the number of stutter days is the number of calendar days within the user's window time range where stutter events occurred, specifically calculated by extracting date information from stutter event records and counting by deduplicated dates; the maximum number of consecutive stutter days is the maximum number of calendar days within the user's window time range where stutter events occurred consecutively, specifically calculated using the row number difference method. Each user's stutter dates are sorted by calendar day to obtain a date sequence number, and the difference is obtained by subtracting the row number from the date sequence number. Consecutive dates have the same difference, and the number of dates with the same difference is the length of a consecutive stutter interval. The maximum value among all consecutive interval lengths is taken as the maximum number of consecutive stutter days. For users with no stutter event records within the window time range, the values of the above three statistical indicators are all 0, and the system can fill the statistical results table with 0 values to avoid null values interfering with subsequent judgments.
[0143] After obtaining the lag statistics for the two windows, the system calculates the ratio of the average daily lag counts for the first sliding time window to the second sliding time window. This ratio measures the direction of change in the user's short-term lag frequency relative to the long-term average level. Specifically, the calculation method is as follows: divide the number of lags within the first sliding time window by the length of the first sliding time window (7 days) to obtain the average daily lag count for the first sliding time window; divide the number of lags within the second sliding time window by the length of the second sliding time window (30 days) to obtain the average daily lag count for the second sliding time window; then divide the average daily lag count for the first sliding time window by the average daily lag count for the second sliding time window to obtain the ratio of the average daily lag counts.
[0144] When the average number of stutters per day in the second sliding time window is 0, this ratio cannot be calculated mathematically. In this case, the user experiences no stutters within the second sliding time window and is not included in the ratio determination of the deteriorating state and the mitigating state. The user directly enters the determination of the excellent stable state or the stutter-free state. The calculation of this ratio can be completed using arithmetic operation functions in the offline data warehouse, and the generated results are stored in the user tag wide table.
[0145] When the ratio of the average daily number of stutters exceeds a preset deterioration threshold, and the maximum number of consecutive stutter days within the first sliding time window is greater than or equal to a preset consecutive days threshold, the system determines the user's network quality status as deteriorated. The preset deterioration threshold can be calibrated and set based on actual production data. For example, a value of 1.5 indicates that the user's average daily stutter count over the past 7 days is more than 1.5 times the average daily stutter count over the past 30 days, showing a clear deteriorating trend. The preset consecutive days threshold can be set to 3 days, indicating that the user has experienced at least 3 consecutive days of stuttering within the past 7 days, suggesting that the stuttering is not an occasional, isolated event but rather a persistent issue. When both conditions are met simultaneously, the system determines that the user is in a deteriorated state, indicating that the user is experiencing a rapid decline in network quality and is a priority for proactive service.
[0146] When the ratio of daily average lag occurrences to the preset mitigation threshold is less than the threshold, and lag records exist within the second sliding time window, the system classifies the user's network quality status as "mitigated." The preset mitigation threshold can be set to 0.8, meaning that the user's average daily lag occurrences over the past 7 days have decreased to less than 0.8 times the average daily lag occurrences over the past 30 days, indicating a significant decrease in lag frequency. The condition of lag records existing within the second sliding time window is used to exclude users who have not experienced lag for a long period from being mistakenly classified as remitted. Only if a user has actually experienced lag in the past 30 days and the lag has significantly decreased recently is the user considered to be in a remitted state. A remitted state indicates that the user's problem is improving, and the system can accordingly reduce the intensity of proactive service interventions to avoid unduly disturbing the user.
[0147] When the number of stutters within the second sliding time window is zero, the number of tags with a value of 1 in the positive capability label is greater than or equal to two, and the home value index is greater than or equal to a preset index threshold, the system determines the user's network quality status as excellent and stable. Zero stutters within the second sliding time window means the user has not experienced any stuttering events in the past 30 days. A number of tags with a value of 1 in the positive capability label is greater than or equal to two, indicating the user possesses at least two positive capabilities, such as simultaneously having gigabit readiness and a 5G terminal, or simultaneously having FTTR whole-house coverage and large-screen high-definition display. A home value index greater than or equal to a preset index threshold indicates that the user's home equipment value level is at a relatively high level; the preset index threshold can be set to 70 points. The excellent and stable status is used to identify users with consistently stable network quality, strong equipment capabilities, and high home value; these users are the optimal target group for retaining upgraded service scenarios.
[0148] When the number of stutters within the second sliding time window is zero, and the number of tags with a value of 1 in the positive capability label is less than 2, or the family value index is less than the preset index threshold, the system determines the user's network quality status as a stutter-free state. A stutter-free state indicates that although the user has no stuttering records, similar to the excellent / stable state, their positive capability or family value level has not yet reached the excellent / stable standard. For example, the user may only possess one positive capability or have a family value index below 70. These users still need to be monitored regularly, but they can be temporarily excluded from priority access in the upgrade service. The difference between the two states lies in the number of positive capability tags and the family value index. When making a judgment, the system first checks whether all the conditions for excellent / stable are met simultaneously; if not, the user falls into the stutter-free state.
[0149] If the number of stutters within the second sliding time window is not zero, and the home broadband user does not meet the criteria for deterioration or mitigation, and the user's network quality status at the time of the last status update was either stable stuttering or improving, and the number of consecutive zero stuttering days reaches the preset confirmation period, the system will determine the user's network quality status as improving. The last status update refers to the user's status record at the time the system most recently completed status determination before the current tag calculation date. The number of consecutive zero stuttering days refers to the number of consecutive natural days without a stuttering event, traced back from the current tag calculation date. The system can check for stuttering records day by day from the current calculation date based on the date information of the stuttering event records until a stuttering record is found; the number of days without a stuttering record is the number of consecutive zero stuttering days. The preset confirmation period can be set to 3 days, meaning that the status will only be confirmed as improving after the user has had 3 consecutive days without stuttering. The "Improving" status indicates that the user's network quality has begun to improve from continuous lag, but has not yet reached the criteria for a "good, stable" or "lag-free" status. The system will continue to observe in subsequent calculations and switch to the "Relief" status when the ratio is less than the preset relief threshold, or switch to the corresponding status after the complete criteria for a "good, stable" or "lag-free" status are met.
[0150] When a user does not meet the criteria for deterioration, mitigation, excellent / stable, no lag, or improvement, the system will classify the user's network quality status as stable lag. Stable lag is a fallback status, describing situations where a user has lag records within the second sliding time window, but the daily average lag frequency ratio has not reached the deterioration or mitigation boundary conditions, or where the ratio meets the conditions but other additional conditions such as the number of consecutive lag days are not met. The system can determine status in the following order: deterioration and mitigation first, excellent / stable and no lag second, improvement third, and stable lag last. Each status determination outputs only one status label, which is written to the network quality status field in the user's label wide table.
[0151] The Home Value Index, as a component of the service experience label, is used to characterize the value level of home equipment for home broadband users. In this embodiment, the calculation process of the Home Value Index is as follows: the system obtains the equipment type, model, and unit price information of various types of downstream equipment in the user's home from the downstream equipment detail wide table, and performs a weighted summation of all equipment in the user's home according to preset equipment type weights and equipment grade weights to obtain the total home equipment value score. Then, the total home equipment value score is converted into a Home Value Index within the range of 0 to 100 according to preset mapping rules.
[0152] Device type weights can be set based on the device's contribution to the home network experience; for example, core network devices such as routers and gateways have higher weights than sensor devices. Device tier weights can be set based on the device's release date and brand positioning; more recently released, high-end devices have higher tier weights. These weighted summation and mapping rules can be maintained in the system configuration and executed in batches during tag calculation tasks. The home value index is not only used to distinguish between excellent and stable conditions and smooth operation, but can also serve as a reference dimension for identifying high-value users in scenarios requiring upgraded services.
[0153] In one or more embodiments of the present invention, determining the network quality status of a home broadband user further includes a status anti-jitter mechanism:
[0154] When a home broadband user switches from the current state to the target state, the entry conditions for the target state must be met for a preset number of consecutive confirmation days.
[0155] After the state transition is completed, a cooldown period is set, during which it is prohibited to switch back to the state before the transition.
[0156] When a user complaint occurs or the overall quality score falls below the preset mandatory coverage threshold, the confirmation period and cooling-off period are ignored, and the network quality status is forcibly covered to a deteriorated state.
[0157] In addition to relying on dual-window statistical indicators and label values, the determination of network quality status also requires constraints on the state switching process through a state de-jitter mechanism. The purpose of the state de-jitter mechanism is to prevent users from frequently switching between adjacent states due to occasional daily lag events, which would cause the service system to repeatedly trigger meaningless service actions.
[0158] For example, if a user experiences lag for several consecutive days due to a temporary network outage, and the system immediately switches the user from a lag-free state to a stable lag state, and then immediately switches back to a lag-free state the next day after the user returns to lag-free status, the status label will fluctuate between the two states. This jitter not only increases the system's processing load but may also cause the service decision module to misjudge the user's true trend, resulting in unnecessary service outreach. A state jitter prevention mechanism needs to introduce two time constraints—an acknowledgment period and a cooling-off period—in the state switching process, and use forced overwrite to ensure that the state can be corrected promptly in extreme cases.
[0159] The complete process of state switching can be described as follows: the system determines the user's network quality status on each tag calculation day. During the determination process, the user's original network quality status is recorded in the status field generated on the previous tag calculation day; this status is called the current status. The system calculates the user's appropriate status for the day based on the dual-window statistical results and status determination conditions; this is called the target status.
[0160] When the target state is the same as the current state, the system maintains the current state without entering the anti-jitter judgment process. When the target state is different from the current state, the system initiates the state switching process and performs an acknowledgment period judgment according to the requirements of the anti-jitter mechanism.
[0161] The confirmation period mechanism ensures that users have indeed entered the network quality change phase corresponding to the target state, rather than the target state conditions being met instantaneously due to accidental factors. The system maintains a state candidate record within the confirmation period mechanism, which includes the candidate target state identifier and the number of consecutive days that the state entry conditions have been met.
[0162] Specifically, when the target state differs from the current state, the system first checks if the same candidate target state already exists in the candidate state record. If the candidate target state does not exist in the candidate state record, the system creates a new candidate state record, writes the candidate target state identifier into the record, and sets the number of consecutive days it meets the requirement to 1. At this time, the user's network quality status remains the current state, and no actual switch occurs.
[0163] In subsequent tag calculation days, if the target state is still different from the current state but the same as the candidate target state in the state candidate record, the system increments the number of consecutive days that the candidate target state is satisfied by 1. If the target state changes on a later day, for example, if the target state returns to the same as the current state or becomes a different target state, the system deletes the original state candidate record and restarts the counting.
[0164] The system will only officially switch the user's network quality status from the current state to the candidate target state and delete the candidate status record when the number of consecutive days that a candidate target state in the status candidate record has been met reaches the preset confirmation period. The preset confirmation period can be set to 3 days, meaning that the entry condition for the target state needs to be met for 3 consecutive days based on the daily average of the tags before the system confirms the status switch.
[0165] The cooldown mechanism suppresses reverse state transitions for a period of time after a state switch is completed, preventing users from oscillating between two states. After completing a state switch, the system records the completion time of the switch and the state identifier before the switch. The system also sets the cooldown period length, which can be set to 7 days.
[0166] During the cooldown period, if the user's target state is determined to be the state before the switch, meaning the user's state has regressed, the system will not execute the reverse switch and will continue to maintain the post-switch state. The cooldown period is calculated from the date the tag was calculated after the state switch was completed.
[0167] For example, if a user switches from a stable, lagging state to an improving state, and the user's statistical indicators again meet the criteria for a stable, lagging state the following day (i.e., the target state is stable, lagging), the system will not execute a reverse switch because this is the state before the switch and is still within the 7-day cooldown period. The user will remain in the improving state. After the cooldown period ends, if the user's target state is still stable, lagging, the system will determine whether to execute a reverse switch according to the normal confirmation period mechanism. The cooldown period mechanism temporarily freezes the reverse switch path, providing a stable observation window for the new state and ensuring the sustainability of the state switch results.
[0168] Candidate status records and cooldown period records can be stored in extended fields of the user tag wide table, or they can be stored independently in the system's status management data table. Specifically, the storage method involves adding a candidate target status field, a field for the number of consecutive days the candidate status has been met, a cooldown period expiration date field, and a pre-cooldown status field to the user tag wide table.
[0169] After each tag calculation task is completed, the system writes the updated status fields and dejitter-related fields back to the user tag wide table. At the start of the next tag calculation day, the system reads the status fields and dejitter-related fields written in the previous cycle from the user tag wide table, using them as the current status and dejitter auxiliary information in this status switching process. This data storage method ensures the continuity and consistency of the status switching process across batch processing tasks spanning multiple days.
[0170] The mandatory coverage mechanism handles two special cases: user complaints and abnormally low overall quality scores. User complaint data comes from the customer service system and complaint ticket records. The system correlates user complaint data on each tag calculation day. When it detects that a user has a complaint record within a preset time range, such as a complaint record within the last 24 hours, the system determines that the user has triggered the complaint conditions. An abnormally low overall quality score means that the user's overall quality score obtained in the current tag calculation is lower than a preset mandatory coverage threshold, which can be set to 30 points.
[0171] When a user triggers a complaint or their overall quality score falls below a preset mandatory coverage threshold, the system ignores the confirmation period and cooling-off period mechanisms, directly overwriting the user's network quality status to a deteriorated state, and simultaneously clearing the user's status candidate record and cooling-off period record. After mandatory coverage, the user's status switching process on the next tag calculation day will restart with the deteriorated state as the new current state. The mandatory coverage mechanism ensures that the system can respond immediately when a user has expressed strong dissatisfaction through a complaint or when the user's network quality has severely declined, without waiting for the confirmation period or cooling-off period window.
[0172] In this embodiment, the anti-jitter mechanism, dual-window statistics, and status determination are executed in the same batch processing task. Following the batch processing scheduling order of the offline data warehouse, the system first completes data collection, stuttering event statistics, and tag generation, then performs status determination, subsequently calculates the comprehensive quality score, and finally performs a forced coverage check based on complaint records and the comprehensive quality score, generating a network quality status field and writing it back to the user tag wide table.
[0173] The preset confirmation days, cooldown period length, and preset forced coverage threshold during the jitter reduction process are all stored as system configuration parameters in the knowledge base management module. The business side can adjust these parameters based on actual operational results, and the adjusted parameters will take effect in the next batch of processing tasks. By introducing a state jitter reduction mechanism, the system ensures the sensitivity of state determination while suppressing state oscillations caused by occasional events, enabling proactive service decisions to be based on more stable network quality state identification results.
[0174] In one or more embodiments of the present invention, calculating the overall quality score includes:
[0175] The experience score is calculated based on the lag index in the root cause segment label of network lag. The experience score is inversely related to the number of lags, the number of lag days, and the maximum number of consecutive lag days.
[0176] Ability scores are calculated based on the values of each identifier in the positive ability labels and preset weights;
[0177] A stable score is calculated based on the day penalty for exceeding a preset burst threshold for the number of times a single day experiences lag, and the difference between the first sliding time window and the second sliding time window in the experience score.
[0178] The overall quality score is obtained by weighting and summing the experience score, ability score, and stability score.
[0179] The weights of the experience score, the capability score, and the stability score are adjusted according to the network quality status.
[0180] The overall quality score is used to convert a user's network usage quality level into a comparable numerical value, ranging from 0 to 100. A higher value indicates better overall network usage quality. The overall quality score consists of three parts: experience score, capability score, and stability score. Each part describes the user's network usage quality from a different perspective, and the final result is obtained by weighted summation.
[0181] After completing daily tag calculation and network quality status assessment, the system performs a comprehensive quality score calculation. The input data required for the calculation includes the lag indicators in the network lag root cause segment tag, the values of various identifiers in the positive capability tag, and the lag statistics for the first and second sliding time windows. The calculation process is completed in batch processing in an offline data warehouse. A comprehensive quality score is generated for each household broadband user and stored in the user tag wide table.
[0182] The experience score is calculated based on the lag index in the root cause segment label of network lag, and is used to reflect the quality of the user's actual network experience. In this embodiment, the system uses three indicators as inputs for the experience score: the total number of lags across all fault segments, the number of lag days, and the maximum number of consecutive lag days. All three indicators are obtained from the user's root cause segment lag detail wide table. The number of lags is the total number of lag event records across all fault segments within the second sliding time window; the number of lag days is the number of natural days in which lag events occurred across all fault segments within the second sliding time window; and the maximum number of consecutive lag days is the maximum number of natural days in which lag events occurred consecutively across all fault segments within the second sliding time window.
[0183] The system uses preset rating mapping rules to map the three indicators to scores ranging from 0 to 100. These three mapped scores are then weighted and averaged to obtain the experience score. The rating mapping rules can be configured as a segmented mapping table; the following example illustrates this.
[0184] The mapping rule for the number of stutters is as follows: when the number of stutters within the second sliding time window is 0, the corresponding stutter score is 100 points; when the number of stutters is 1 to 5, the corresponding stutter score is 80 points; when the number of stutters is 6 to 10, the corresponding stutter score is 60 points; when the number of stutters is 11 to 20, the corresponding stutter score is 40 points; and when the number of stutters is 21 or more, the corresponding stutter score is 20 points.
[0185] The mapping rules for the number of lag days are as follows: when the number of lag days in the second sliding time window is 0, the corresponding lag day score is 100 points; when the number of lag days is 1 to 3, the corresponding lag day score is 80 points; when the number of lag days is 4 to 7, the corresponding lag day score is 60 points; when the number of lag days is 8 to 15, the corresponding lag day score is 40 points; and when the number of lag days is 16 days or more, the corresponding lag day score is 20 points.
[0186] The mapping rule for the maximum consecutive lag days is as follows: when the maximum consecutive lag days within the second sliding time window is 0 days, the corresponding consecutive days score is 100 points; when the maximum consecutive lag days is 1 to 2 days, the corresponding consecutive days score is 80 points; when the maximum consecutive lag days is 3 to 5 days, the corresponding consecutive days score is 60 points; when the maximum consecutive lag days is 6 to 10 days, the corresponding consecutive days score is 40 points; and when the maximum consecutive lag days is 11 days or more, the corresponding consecutive days score is 20 points.
[0187] The scores for the three indicators are averaged to obtain the experience score, which is equal to the sum of the scores for the number of times the system lagged, the number of days the system lagged, and the number of consecutive days the system lagged, divided by 3. The above mapping table and weighting method are only one implementation method in this embodiment. In actual deployment, the boundaries of each interval and the corresponding scores of each interval can be adjusted through system configuration.
[0188] The capability score is calculated based on the values of each identifier in the positive capability tags and their preset weights. The system reads six identifier fields from the positive capability tags in the user tag wide table, including gigabit readiness, FTTR whole-house coverage, large-screen high-definition, 5G terminal ownership, smart terminal ecosystem, and device newness / oldness. Each identifier field has a value of 0 or 1. The system assigns a weight coefficient to each identifier, which represents the degree to which the positive capability contributes to the user's overall capability.
[0189] In this embodiment, the default weighting coefficients are configured as follows: Gigabit Ready identifier weight is 0.25, FTTR Whole-House Coverage identifier weight is 0.20, Large Screen High Definition identifier weight is 0.15, 5G Terminal Holding identifier weight is 0.15, Smart Terminal Ecosystem identifier weight is 0.15, and Device New / Old Profile identifier weight is 0.10. The capability score is calculated by multiplying the value of each identifier by its corresponding weighting coefficient to obtain a weighted value for that identifier. The weighted values of the six identifiers are then summed to obtain a total weighted value. This total weighted value is divided by the sum of the six weighting coefficients and then multiplied by 100 to obtain the capability score. Since the sum of the six weighting coefficients is 1, in actual calculations, the total weighted value multiplied by 100 can be directly used as the capability score.
[0190] For example, if a user has a gigabit-ready flag of 1, an FTTR whole-house coverage flag of 1, a 5G terminal ownership flag of 1, and all other flags of 0, then the user's capability score is 0.25 + 0.20 + 0.15, totaling 0.60, which multiplied by 100 gives 60 points. The capability score reflects the user's upgrade potential in terms of terminal capabilities; a higher capability score indicates a higher degree of adaptability for scenarios requiring upgrades.
[0191] The stability score is calculated based on the penalty for the number of days the number of stutters exceeds a preset burst threshold and the difference between the experience score and the first sliding time window. The system first calculates the difference in experience score across the two sliding time windows. Specifically, using the same scoring mapping rules as the experience score, the experience score for the first sliding time window and the experience score for the second sliding time window are calculated respectively. The experience score for the first sliding time window is mapped and averaged using three indicators: the number of stutters, the number of stutter days, and the maximum number of consecutive stutter days within a 7-day window. The experience score for the second sliding time window is mapped and averaged using the same indicators within a 30-day window. After calculating the two experience scores, the absolute value of the difference is taken. This value is used to measure the consistency between the user's recent and long-term experience. The larger the difference, the more significant the fluctuation in the user's network quality in the short term.
[0192] The calculation process for the number of bursts is as follows: the system counts the total number of lags for each day within the second sliding time window, forming a daily lag count sequence. A preset burst threshold is used to define the criteria for judging an abnormally high number of lags on a single day; in this embodiment, the preset burst threshold can be configured to 8 times. When the total number of lags on a certain day exceeds the preset burst threshold, that day is counted as a burst day. The number of burst days within the second sliding time window is accumulated to obtain the burst count.
[0193] The system multiplies the number of sudden incidents by a preset penalty coefficient to obtain the sudden incident penalty value, and multiplies the absolute value of the difference between the experience scores of two windows by a preset difference penalty coefficient to obtain the difference penalty value. The stability score is calculated by subtracting the sudden incident penalty value and the difference penalty value from a maximum score of 100; that is, the stability score equals 100 minus the sudden incident penalty and then minus the window experience score difference penalty. In this embodiment, the preset penalty coefficient is configured as 5, meaning 5 points are deducted for each sudden incident day; the preset difference penalty coefficient is configured as 0.8, meaning 0.8 points are deducted for every 1-point difference in the window experience scores. If the calculated stability score is lower than 0, it is treated as 0 to ensure that the stability score remains within the range of 0 to 100. A higher stability score indicates better consistency and stability of the user's network quality across short and long windows.
[0194] The overall quality score is a weighted sum of the experience score, ability score, and stability score. The default weighting is 0.5 for the experience score, 0.3 for the ability score, and 0.2 for the stability score, with a sum of 1. Therefore, the overall quality score equals the experience score multiplied by 0.5, the ability score multiplied by 0.3, and the stability score multiplied by 0.2.
[0195] The system adjusts weighting coefficients according to preset rules based on the user's current network quality status. When the user's network quality is deteriorating, user experience becomes the primary concern. The system increases the experience score weight to 0.7, while the remaining 0.3 is allocated to the capability score and stability score according to the original ratio. After adjustment, the capability score weight is 0.18, and the stability score weight is 0.12. When the user's network quality is good and stable, lag issues are not significant, and capability management becomes the focus. The system increases the capability score weight to 0.5, while the remaining 0.5 is allocated to the experience score and stability score according to a 3:2 ratio. After adjustment, the experience score weight is 0.3, and the stability score weight is 0.2. When the user has a complaint record within a preset time range, the experience score weight increases to 0.6, while the remaining 0.4 is allocated to the capability score and stability score according to the original ratio. After adjustment, the capability score weight is 0.24, and the stability score weight is 0.16. When the user's network quality status is other and there are no complaints, the default weights are used. The weight adjustment rules are stored in the system configuration and can be adjusted according to actual operational needs. When multiple weight adjustment conditions are met simultaneously, the weight configuration that is hit first is executed in the order of worsening state rule, complaint record rule, and good / stable state rule.
[0196] After the comprehensive quality score is calculated, the system writes the score value into the comprehensive quality score field of the user tag wide table, which serves as an important input for determining subsequent service priorities. By weighting and combining the experience score, capability score, and stability score, the comprehensive quality score can simultaneously reflect the user's network experience level, terminal capability level, and status stability level, providing a quantitative basis for distinguishing the service priorities of different users.
[0197] In one or more embodiments of the present invention, the method for determining the TOP1 root cause segment among the six fault segments is as follows:
[0198] Sort the number of stutters in the six fault segments;
[0199] When there are two or more fault segments with the same number of pauses, the maximum number of consecutive pause days, the significance of peak periods, and the repairability level of the fault segment are compared in turn to determine the TOP1 root cause segment. Among them, the significance of peak periods is determined based on the number of pauses corresponding to the TOP1 peak period and its proportion to the total number of pauses in the fault segment.
[0200] The repairability levels, from highest to lowest, are as follows: LAN-side link segment, gateway WiFi segment, and routing device segment are tied for the highest level, followed by user terminal segment, gateway segment, and Internet segment.
[0201] The process of identifying the TOP1 root cause segment among six fault segments is used to identify the fault segment that has the greatest impact on the user's network experience at the current stage from multiple potential fault locations in the user's home network that may experience network lag. The TOP1 root cause segment provides direct root cause basis for subsequent remedial service scenario matching, installation and maintenance dispatch direction, and service script selection. The system generates TOP1 root cause segment results for the first and second sliding time windows respectively, and writes them into the corresponding TOP1 root cause segment field in the user tag wide table.
[0202] The core criterion for determining the TOP1 root cause segment is the number of times each of the six fault segments experienced buffering within the window statistical period. The system reads the buffering frequency index of the six fault segments under the same broadband account from the user root cause segment buffering details wide table. The six fault segments are the user terminal segment, routing device segment, LAN-side link segment, gateway WiFi segment, gateway segment, and Internet segment. First, the fault segments are sorted in descending order according to the number of buffering times. When the number of buffering times of a certain fault segment is different from that of other fault segments, that fault segment is determined as the TOP1 root cause segment. When two or more fault segments have the same number of buffering times, a tie-breaking process is initiated, using three dimensions—continuous buffering days, peak-hour significance, and installation / maintenance repairability level—for a step-by-step comparison.
[0203] The specific steps of the tie-breaking process are as follows: The system first compares the maximum consecutive lag days for each fault segment with the same number of lags. The maximum consecutive lag days is the maximum number of natural days in which lag events occur consecutively within the window statistical period for that fault segment. This value has already been calculated using the row number difference method and stored in the corresponding field during the generation of the aforementioned user root cause segment lag detail wide table. The system then filters out the fault segment with the highest maximum consecutive lag days from these fault segments. If only one fault segment remains after filtering, that fault segment is determined as the TOP1 root cause segment. If multiple fault segments still have the same maximum consecutive lag days, the system continues to compare the significance of peak periods for these fault segments.
[0204] Comparing the significance of peak periods requires data on the distribution of stuttering times for each fault segment. In this embodiment, the peak period refers to the period with the highest proportion of stuttering events, which typically corresponds to the evening peak hours. The system reads the TOP1 peak period field and the number of stutters during the peak period from the user root cause segment stuttering detail wide table for each fault segment. During the comparison, the system first looks at the number of stutters corresponding to the TOP1 peak period for each fault segment; the larger the value, the more significant the peak period. When the number of stutters during the TOP1 peak period is also the same, the system then compares the proportion of the number of stutters during the peak period to the total number of stutters in that fault segment. This proportion is obtained by dividing the number of stutters during the peak period by the total number of stutters in the fault segment; the higher the proportion, the more concentrated the stuttering is during the peak period, and the higher the significance of the peak period. If, after the above comparison, multiple fault segments still have the same significance of peak periods, the system proceeds to the comparison of installation and maintenance repairability levels.
[0205] The repairability rating measures the ease with which problems corresponding to different fault segments can be located and resolved during on-site installation and maintenance. Fault segments with higher repairability ratings are prioritized as the top root cause segment. The order of repairability ratings from highest to lowest is as follows: LAN-side link segments, gateway WiFi segments, and routing device segments are tied for the highest level; user terminal segments are at the second highest level; gateway segments are at the medium level; and internet segments are at the lowest level.
[0206] The LAN link segment involves the quality of the internal links of the home LAN, which can be resolved through on-site network optimization, replacing network cables, or adjusting the equipment layout. The gateway WiFi segment involves the gateway's built-in wireless LAN function, which can be resolved through channel optimization or adding coverage devices. The router segment involves the user-purchased wireless router, which can be resolved by replacing the router. All three of these are issues that can be handled on-site within the home, and their repairability is roughly equal, therefore ranking highest. The user terminal segment involves the user's own terminal equipment; installation and maintenance personnel can only offer suggestions but cannot directly replace it, making its repairability lower than the previous three. The gateway segment involves the home gateway equipment provided by the operator, requiring equipment replacement or backend configuration modifications, making on-site troubleshooting more difficult than the user terminal segment. The internet segment involves internet-side service or link issues, which installation and maintenance personnel cannot resolve on-site, requiring backend coordination or link upgrades, making its repairability the lowest.
[0207] During comparison, the system sorts the fault segments according to their corresponding repairability levels, with the fault segment ranked higher being prioritized as the TOP1 root cause segment. If two fault segments are at the same level, the system determines the TOP1 root cause segment according to a preset order of six fault segments: LAN-side link segment, gateway WiFi segment, routing device segment, user terminal segment, gateway segment, and Internet segment.
[0208] Following the aforementioned tie-breaking process, the system ultimately determines the top-1 root cause segment for each user within the statistical window period. The first and second sliding time windows execute the above process independently, calculating the corresponding fields for the top-1 root cause segment. The system then writes both the top-1 root cause segments from the first and second sliding time windows into a user tag wide table as one of the inputs for subsequent scenario matching. Using this method, the system can, when multiple fault segments experience similar numbers of glitches, select the fault segment with the most significant impact on users and the most favorable conditions for on-site resolution based on three dimensions: continuous glitches characteristics, peak period concentration, and on-site repairability. This improves the accuracy of top-1 root cause location and the effectiveness of subsequent order dispatch.
[0209] In one or more embodiments of the present invention, the repair service scenarios include old router replacement scenarios, FTTR sub-router installation scenarios, Tianyi gateway sub-router installation scenarios, and large-screen video lag optimization scenarios.
[0210] Upgraded service scenarios include gigabit speed upgrade recommendation scenario, FTTR whole-house upgrade scenario, smart home product expansion scenario, and loyalty enhancement scenario.
[0211] The service scenario matching engine is configured with eight specific scenarios, including repair service scenarios and retention / upgrade service scenarios. Repair service scenarios include old router replacement, FTTR sub-router installation, China Telecom gateway sub-router installation, and large-screen video buffering optimization. Retention / upgrade service scenarios include gigabit speed-up recommendation, FTTR whole-house upgrade, smart home expansion, and loyalty enhancement. The triggering conditions for all scenarios are defined based on one or more conditions from the following categories: network buffering root cause segment label, positive capability label, service experience label, and network quality status. On each label calculation day, the scenario matching engine reads the relevant fields from the user's wide label table, matches them one by one according to the scenario triggering conditions, and outputs the identifier of the matched target service scenario. If a user matches multiple scenarios simultaneously, the system selects the highest priority scenario as the target service scenario according to the scenario priority mutual exclusion rule, which is described in later paragraphs of this embodiment. The specific triggering conditions and corresponding service actions for each of the eight scenarios are explained below.
[0212] The old router replacement scenario falls under the category of repair-type service. This scenario targets home broadband users using older wireless routers who are experiencing lag issues. Triggering conditions include: the user's home router uses WiFi 4 or WiFi 5, and no routers supporting WiFi 6 or WiFi 7 are present; the number of lag occurrences for the corresponding router segment is greater than zero, indicating that the user's router is involved in the lag event; the total number of lag occurrences across all fault segments exceeds a preset high-frequency threshold (which can be set to 22 times, indicating the user is a frequent lag user); the user has not changed their service plan within the last preset number of months (which can be set to 3 months); and the user is not an FTTR user. Users meeting all of the above conditions qualify for the old router replacement scenario. Service actions include dispatching a work order to the installation and maintenance channel, where installation and maintenance personnel will visit the user's home to inspect the wireless router and recommend replacing it with a router supporting the WiFi 6 protocol. For users with higher spending power, a router replacement plan with a monthly fee increase of at least 60 yuan can be recommended; for users with moderate spending power, a router replacement plan with a monthly fee increase of at least 30 yuan can be recommended; for users with lower spending power, users can be guided to experience free network optimization services. The root cause of this scenario is determined based on the lag indicators of the router device segment.
[0213] The FTTR sub-router installation scenario is a remedial service scenario. This scenario is for users who have subscribed to an FTTR plan but whose actual network has insufficient sub-modems. Triggering conditions include: the user's plan type is an FTTR plan, or the user's home gateway device model matches the FTTR main modem device model database; the user's actual number of sub-modems does not match the number specified in the plan, specifically, the number of sub-modems included in the plan is greater than the actual number installed; the number of stutters on the user's gateway WiFi segment is greater than zero, indicating a problem with wireless coverage; the total number of stutters across all faulty segments exceeds a preset high-frequency threshold, which can be set to 22 times. Users meeting all of the above conditions qualify for the FTTR sub-router installation scenario. Service actions include dispatching an on-site survey work order, with installation and maintenance personnel using wireless signal testing equipment to collect a heat map of the user's entire home's wireless coverage, showing the user the locations of coverage blind spots, and recommending the installation of FTTR sub-modems to achieve whole-home wireless coverage. For users whose packages include a sub-modem but which are not actually installed, it can be installed free of charge as per the package terms; for users who need to add a sub-modem, it can be recommended as an added-value service.
[0214] The Tianyi Gateway sub-router installation scenario falls under the category of remedial service. This scenario targets users who haven't subscribed to an FTTR plan and are using Tianyi gateway devices, but experience insufficient wireless coverage. Triggering conditions include: the user's home gateway device model is Tianyi Gateway 3.0 or Tianyi Gateway 4.0; the number of Wi-Fi signal interruptions on the user's gateway is greater than zero, indicating a coverage problem with the gateway's built-in wireless function; and the total number of interruptions across all faulty segments exceeds a preset high-frequency threshold, which can be set to 22 times. Users meeting all of the above conditions qualify for the Tianyi Gateway sub-router installation scenario. Service actions include dispatching an on-site inspection work order, having installation and maintenance personnel check the gateway's wireless signal strength, recommending the installation of a sub-router device to extend wireless coverage, or recommending an upgrade to an FTTR whole-house networking solution.
[0215] The large-screen video stuttering optimization scenario is a remedial service scenario. This scenario targets users experiencing stuttering in large-screen video applications, including two sub-scenarios: non-ITV users and inactive ITV users. The triggering conditions for non-ITV users include: stuttering records in the user's TV or set-top box applications; the total number of stutters across all faulty segments exceeding a preset high-frequency threshold (which can be set to 22 times); and the user not having subscribed to ITV services. The triggering conditions for inactive ITV users include: stuttering records in the user's TV or set-top box applications; the total number of stutters across all faulty segments exceeding a preset high-frequency threshold (which can be set to 22 times); and the user having subscribed to ITV services but with the ITV service inactive. Service actions include recommending the ITV Premium Package and a 4K gigabit Ethernet set-top box upgrade to non-ITV users; and pushing the ITV wake-up service to inactive ITV users, recommending the Premium Package and bundled membership benefits. In this scenario, when identifying stuttering records of TV applications, the application name field in the stuttering event data is matched with a preset whitelist of large-screen video applications. For example, the whitelist includes applications such as "TV version of [application name]", "TV of [application name]", and "Video of [application name]". The terminal type field is also used for auxiliary judgment.
[0216] The gigabit speed upgrade recommendation scenario belongs to the "retain and upgrade" service scenario. This scenario is for users whose home networks already have gigabit capacity but whose current package speed is lower than gigabit. Triggering conditions include: the user's gigabit ready flag is set to 1, meaning the user's home router supports WiFi 6 or WiFi 7 protocols or the router's LAN port speed is greater than or equal to 1000 Mbps; the user's current package speed is lower than 1000 Mbps; the user's monthly data usage exceeds a preset data usage threshold (which can be set to 300 gigabytes); the user's network quality status is excellent and stable; the overall quality score exceeds a preset score threshold (which can be set to 85 points); the user is a high-value bundled user or their average monthly spending exceeds or equals a preset amount threshold (which can be set to 99 yuan). High-value bundled users refer to users whose average income over the past 3 months is greater than or equal to 129 yuan. Service actions include pushing gigabit package upgrade discount information through the operator's application, such as a 20% discount for upgrading to a gigabit package, and following up on the user's intention to upgrade through WeChat customer service. The core objective of this scenario is to identify user groups whose network conditions are ready but whose service plans are not yet matched, and to accelerate value enhancement through proactive recommendations.
[0217] The FTTR whole-house upgrade scenario is a retention-upgrade service scenario. This scenario is for large households with multiple devices and insufficient wireless coverage, as well as users experiencing lag on the gateway WiFi segment or LAN side link segment. Triggering conditions include: the number of resident devices in the user's home is greater than or equal to a preset device threshold (which can be set to 15 devices), and the number of mobile devices is greater than or equal to a preset mobile phone threshold (which can be set to 7 mobile phones); the user's router does not have Mesh networking capabilities or there are multiple routers in the user's home (Mesh networking refers to multiple wireless routers forming a mesh network through wireless or wired connections to achieve whole-house WiFi coverage); the number of lags on the user's gateway WiFi segment or LAN side link segment is greater than or equal to a preset lag threshold (which can be set to 5 lags); the user is not an FTTR user; and the user's consumption level is high or their household value index is greater than or equal to a preset index threshold (which can be set to 70 points). Service actions include outbound calls recommending whole-house WiFi upgrade solutions and arranging on-site surveys. For users with higher spending power, an FTTR networking solution with a monthly fee increase of no less than 90 yuan is recommended; for users with moderate spending power, an FTTR networking solution with a monthly fee increase of no less than 60 yuan is recommended.
[0218] The smart home expansion scenario falls under the category of "retaining and upgrading" service scenarios. This scenario targets high-value users whose smart home devices are experiencing lag and who have networking needs. Triggering conditions include: the number of times a user's smart home devices experience lag is greater than or equal to a preset threshold (which can be set to 3 times); the number of times a user experiences lag on their LAN link segment, gateway WiFi segment, or router segment is greater than or equal to a preset segment lag threshold (which can be set to 5 times); the user has a smart home need identifier or their top five lag device types include door locks, doorbells, and cameras; the user's spending level is high; and the user's service preference type is comprehensive. Service actions include pushing smart home package information, such as doorbell, camera, and speaker bundles, via SMS and WeChat Work, while simultaneously arranging for installation and maintenance personnel to perform on-site network optimization and improve the connection stability of smart home devices. This scenario can identify users who already have a basic understanding of smart homes but experience network bottlenecks, expanding the penetration rate of home ecosystem products.
[0219] Loyalty enhancement is a retention / upgrade service scenario. This scenario targets high-value users who have been with the network for a long time, hold multiple services, have high star ratings, and show signs of cross-network contact. Triggering conditions include: the user's network quality is excellent and stable with a comprehensive quality score greater than a preset score threshold (which can be set to 85); the user's network tenure is greater than or equal to 3 years; the user holds more than or equal to 3 services, including broadband, IPTV, mobile communication, smart home, and home security services; the user's star rating is greater than or equal to 4 stars; and the user shows signs of cross-network contact, which can be determined based on the user's connected devices on other networks or information about secondary SIM cards on other networks. Service actions include a dedicated follow-up call from an account manager, offering complimentary value-added service benefits, such as a one-year trial of 2000 Mbps broadband, a premium package trial, or a discount on a second TV service within the same city. This scenario aims to solidify the loyalty of high-value users before they churn, reducing the risk of churn.
[0220] The triggering conditions and service actions for the eight service scenarios mentioned above are all configured in the rule base of the scenario matching engine during system deployment. Scenario rules can be maintained in a structured configuration format. The business side can adjust the threshold parameters in the triggering conditions according to operational strategies. After parameter adjustment, the new rules take effect in the next batch of processing tasks through hot updates, without requiring system downtime and restart. By configuring both repair-type and retention-upgrade-type scenarios, the system achieves unified identification and differentiated services for users experiencing lag and healthy users, effectively integrating the two business logics of network quality assurance and value management.
[0221] In one or more embodiments of the present invention, determining service priority based on network quality status, comprehensive quality score, and target service scenario includes:
[0222] When the network quality status is deteriorating, the overall quality score is lower than the first threshold, and the target service scenario is a repair service scenario, the service priority is determined to be the first priority.
[0223] When the network quality status is deteriorating, the comprehensive quality score is higher than or equal to the first threshold, and the target service scenario is a repair service scenario, the service priority is determined to be the second priority.
[0224] When the network quality status is stable but laggy, the overall quality score is lower than the second threshold, and the target service scenario is a repair service scenario, the service priority is determined to be the second priority.
[0225] When the network quality status is stable but laggy, the comprehensive quality score is higher than or equal to the second threshold, and the target service scenario is a repair service scenario or a retention upgrade service scenario, the service priority is determined to be the third priority.
[0226] When the network quality status is in a mitigated state, the overall quality score is higher than the third threshold, and no service scenario is hit, the service priority is determined to be the fourth priority.
[0227] When the network quality status is excellent and stable, the comprehensive quality score is higher than the fourth threshold, and the target service scenario belongs to the retention and upgrade service scenario, the service priority is determined to be the fifth priority.
[0228] When the network quality status is under improvement, the overall quality score is higher than the preset score threshold for improvement, and the target service scenario belongs to the retention-upgrade service scenario, the service priority is determined to be the fifth priority; where the preset score threshold for improvement is lower than the fourth threshold.
[0229] The urgency of services decreases sequentially from first priority to fifth priority.
[0230] The service priority determination process is executed in the service decision module. The input data for the service decision module includes the network quality status field and the comprehensive quality score field from the user tag wide table, as well as the target service scenario identifier output by the scenario matching engine. Service priority is used to classify the urgency and processing order of proactive service outreach, with the urgency decreasing sequentially from first to fifth priority. The system sequentially determines whether a user meets all the conditions for each priority level according to pre-configured priority determination rules. When a user meets all the conditions for a certain priority, that priority is assigned to that user, and subsequent priority determination stops. If a user does not meet any of the conditions from first to fifth priority, no proactive service instructions will be generated for that user within the current tag calculation period.
[0231] The criteria for determining the first priority are: the user's network quality status is deteriorating, the overall quality score is below the first threshold, and the target service scenario is a repair-type service scenario. The first threshold is used to identify the boundary value where the overall quality score is in the severely low range. In this embodiment, the first threshold can be configured to 40 points. When the user's status is deteriorating, it means that their network quality has declined significantly recently, accompanied by persistent lag. An overall quality score below 40 points means that the user's experience score, capability score, and stability score are all at a low level after weighting, indicating a high degree of dissatisfaction with the current network service. A target service scenario being a repair-type service scenario means that the system has already matched specific repair actions for the user, such as replacing old routers, adding FTTR sub-routers, adding Tianyi gateway sub-routers, or optimizing large-screen video lag. When all three conditions are met, the user's service priority is determined to be the first priority. The contact method corresponding to the first priority is outbound calling plus dispatching an on-site work order, with the contact time being the completion of the first contact within 24 hours. The system sends outbound call tasks to the customer service outbound call system through the reach execution module, and at the same time sends on-site service work orders to the installation and maintenance work order system. The service instructions carry user identifier, target service scenario identifier and corresponding service script template number.
[0232] The second priority is determined as follows: if a user's network quality is deteriorating, their overall quality score is higher than or equal to the first threshold, and the target service scenario is a repair-type service scenario, then the service priority is determined as the second priority. Alternatively, if a user's network quality is stable but experiencing lag, their overall quality score is lower than the second threshold, and the target service scenario is a repair-type service scenario, then the service priority is also determined as the second priority. In this embodiment, the first threshold can be configured to 40 points, and the second threshold can be configured to 60 points. When a user is in a deteriorating state and their overall quality score is higher than or equal to 40 points, their network quality problem is already obvious, and the system determines the service priority of such users as the second priority. When a user is in a stable but experiencing lag and their overall quality score is lower than 60 points, their persistent lag problem is already quite serious, and the system also determines them as the second priority. The contact method corresponding to the second priority is SMS plus outbound call suggestion. The system sends a service reminder message to the user via SMS and generates an outbound call suggestion task for customer service personnel to follow up. The outbound call time limit is completed within 24 hours. With this tiered approach, users in a deteriorating state are categorized by the first threshold to determine their urgency, while users in a stable, stagnant state are categorized by the second threshold to determine whether they require escalation. This ensures that the allocation of service resources matches the severity of the problem.
[0233] The criteria for determining the third priority are: the user's network quality status is stable with occasional lag, the overall quality score is higher than or equal to the second threshold, and the target service scenario belongs to either a repair-type service scenario or a retention-upgrade-type service scenario. The second threshold is used to define the boundary between the second and third priorities for users in a stable lag state; in this embodiment, the second threshold can be configured to 60 points. When a user's status is stable with occasional lag and the overall quality score is higher than or equal to 60 points, it indicates that the user has a persistent lag problem but has not yet reached the severity level requiring upgrade processing, and belongs to the category of users who can be guided to service through regular channels. The system determines the service priority of such users as the third priority.
[0234] The third priority corresponds to the delivery method of operator application push plus SMS. The system pushes service activity information in the application and sends maintenance reminders via SMS. No emergency outbound calls or on-site visits are arranged to achieve batch delivery of routine repair scenarios at a lower cost. When the service action of the target service scenario itself includes on-site survey or installation and maintenance operations (such as FTTR sub-router installation scenario, smart home product expansion scenario), installation and maintenance work orders are dispatched according to the scenario service action.
[0235] The fourth priority is determined when the user's network quality status is in a mitigated state, the overall quality score is higher than the third threshold, and no service scenario is matched. The third threshold is used to identify the boundary value when the user's network quality recovers to a good level; in this embodiment, the third threshold can be configured to 80 points. When the user's status is in a mitigated state and the overall quality score is higher than 80 points, it indicates that the user's lag problem is significantly improving, and the overall network quality has recovered to a good level. At this time, if the system does not match any repair or upgrade service scenario (i.e., the target service scenario identifier is empty), the system determines the user's service priority as the fourth priority. The fourth priority means that the system keeps monitoring the user but does not actively send service instructions to avoid unnecessary disturbance to users in the recovery period. This priority introduces the condition that the target service scenario is not matched, ensuring logical consistency between the scenario matching result and the service priority determination; that is, only users who actually need service will be assigned proactive outreach actions.
[0236] When a user's network quality status is in a mitigated state, the system will not proactively send service commands regardless of whether the target service scenario is matched. Matched service scenario identifiers will be suspended, and the triggering conditions for that scenario will be continuously evaluated during the suspension period. If the triggering conditions are no longer met, the suspension record will be automatically cleared. If the triggering conditions are still met and the user's network quality status is updated, scenario matching will be performed again. When a user's network quality status is in an improving state and the target service scenario is a repair service scenario, the repair service scenario will also be suspended, and no proactive service commands will be generated. Matching will be performed again after the user's network quality status is updated.
[0237] Users identified as fourth priority are placed on the recovery observation list for continuous tracking by the effect feedback module. Once the lag is detected to recur, the deterioration status review process is triggered directly. After the service scenario for repair is re-matched, it is processed according to the first or second priority without having to go through the full scenario matching cycle again.
[0238] The criteria for determining the fifth priority are as follows: if the user's network quality status is excellent and stable, the overall quality score is higher than the fourth threshold, and the target service scenario belongs to the retention / upgrade service scenario, then the service priority is determined to be the fifth priority; or, if the user's network quality status is improving, the overall quality score is higher than the preset improvement score threshold, and the target service scenario belongs to the retention / upgrade service scenario, then the service priority is also determined to be the fifth priority. The fourth threshold is higher than the third threshold; in this embodiment, the fourth threshold can be configured to 85 points. The preset improvement score threshold can be configured to 70 points, lower than the fourth threshold, to adapt to the scoring characteristics of users in the improvement stage who have not yet fully recovered but already possess business value. When the user's status is excellent and stable and the overall quality score is higher than 85 points, it indicates that the user's current network experience is good and they possess high positive capabilities and family value, and the system determines it to be the fifth priority; when the user's status is improving and the overall quality score is higher than 70 points, it indicates that the user's network quality is recovering and they still possess business value, and the system also determines it to be the fifth priority. If the scenario matched by the system is a service scenario that retains and upgrades services, such as gigabit speed upgrade recommendation, FTTR whole-house upgrade, smart home product expansion or loyalty reinforcement, the system will execute the corresponding scenario-based outreach.
[0239] The fifth priority level corresponds to contextualized push notifications, including push notifications from carrier apps, SMS, or WeChat messages, without outbound calls or in-person visits. Service instructions include specific scenario identifiers and corresponding marketing script templates, with push content configured differently based on the scenario type. While the fifth priority level has the lowest urgency among the five, it targets high-quality user groups and undertakes proactive value management functions.
[0240] The five priority levels are determined in a fixed order from first to fifth priority. The system checks the first priority condition first, then the second priority, and so on. This order ensures that users in a deteriorating state and those with low scores receive service resources first. All threshold parameters are stored in the configuration table of the knowledge base management module, with default configurations of 40 points for the first threshold, 60 points for the second, 80 points for the third, 85 points for the fourth, and a preset improvement-oriented score threshold of 70 points. The second threshold of 60 points also serves as the boundary between the second and third priorities for users in a stable, lagging state. The business side can adjust these parameters based on city, package tier, or operational strategy. Adjustments take effect in the next processing cycle through a hot update mechanism. After service priorities are determined, the system writes the results into the service priority field of the user tag wide table, and the outreach execution module generates and sends service instructions according to the outreach channels and timelines corresponding to the priorities. Through this method, service priorities link network quality status, comprehensive quality score, and target service scenarios, achieving a differentiated service rhythm that prioritizes remedial users and proactively manages healthy users.
[0241] When a user has a complaint record within a preset time frame and the network quality status is deteriorating, regardless of whether the overall quality score is lower than the second threshold, the complaint will be processed with a priority no lower than the second priority, and the first contact will be completed within 24 hours. If the target service scenario is empty, a complaint response ticket will be generated, and the first contact will be completed within 24 hours using a complaint reassurance script.
[0242] In the complete service chain of this invention embodiment, the data collection and multi-source data association stage also includes logical association of multiple data tables and supplementary processing for data quality governance. The logical association sequence of multi-source data unfolds layer by layer within the system according to the user identifier. The first layer associates the user's broadband account with the package subscription information table to obtain the user's package type, package speed, and subscription service identifier. The second layer associates the complaint record table to obtain the number of complaints and the most recent complaint time within a preset time range. The third layer associates the home terminal device dimension table to obtain the model information of the user's main devices such as home gateways, routers, and set-top boxes. The fourth layer associates the downstream device detail wide table to obtain the quantity and device type distribution of various downstream devices in the user's home. The fifth layer associates the application service usage record table to obtain the frequency distribution of user usage under each application category.
[0243] After completing the above five-layer association, the system proceeds to execute the sixth-layer association, connecting to the terminal capability dimension table. This adds attribute fields such as router wireless protocol version, LAN port speed, TV resolution, mobile phone 5G network support, and device launch date to the subsequent calculation of positive capability tags. The seventh layer associates the application service category dimension table, supplementing the calculation of service experience tags with the mapping relationship between application names and application categories. After completing the above seven-layer association, the system generates an initial user wide table with broadband account as the unique primary key, for the tag calculation module to call.
[0244] During data association, if a table has no corresponding record for a specific user, the system uses an empty table sentinel value to fill the gap. An empty table sentinel value is a preset placeholder value; different field types use different sentinel values. For example, numeric fields are filled with 0, identifier fields are filled with 0, and text fields are filled with an empty string. Using an empty table sentinel value avoids null values in the association results, preventing them from interfering with subsequent statistical calculations.
[0245] In addition to null value filling, the system also performs volatility filtering on the associated results. Volatility filtering targets time-series indicators such as the number of stutters per day. When a user's stutter count on a particular day shows an abnormal surge compared to the previous day, and then returns to its original level the following day, the system marks this abnormal record as jittery data and excludes it from the consecutive stutter count calculation. Volatility filtering is achieved by calculating the rate of change of stutter counts between adjacent days. When the rate of change exceeds a preset volatility threshold and falls back to a preset percentage of the original level the following day, that day is determined to be jittery data.
[0246] The tolerance self-correction mechanism works in conjunction with volatility filtering. When a particular day with a lag is identified as jittery data, the consecutive lag intervals preceding that day are skipped during counting, but the interval continuity judgment is maintained. In other words, the jittery day does not interrupt the accumulation of consecutive intervals. The above data processing mechanism ensures that the raw data upon which subsequent dual-window statistics and status determination rely has reasonable stability and continuity.
[0247] In this embodiment, the scene matching process includes not only the triggering conditions of each scene itself, but also cross-scene mutual exclusion rules and data correction logic. When the same user simultaneously meets the triggering conditions of multiple service scenes, the system determines a unique target service scene according to the scene priority mutual exclusion rules. The scene priority mutual exclusion rules include multiple principles.
[0248] The first principle is that repair-oriented service scenarios take precedence over retention-upgrade service scenarios. When a user simultaneously triggers both a repair-oriented scenario and a retention-upgrade scenario, the target service scenario is determined to be the repair-oriented scenario, and the retention-upgrade scenario is temporarily suspended and not involved in the generation of this service instruction. This design avoids pushing marketing content simultaneously when the user has network quality issues, thus preventing conflicting user experiences.
[0249] The second principle is the "prohibition of push during deterioration period." When a user's network quality is in a deteriorated state, the system prohibits any "reserved upgrade" service scenario from being used as the target service scenario. Even if the user meets all the triggering conditions for a "reserved upgrade" scenario, the system will still set its target service scenario to null and only process it based on the matching result of its "repair" scenario. If the user matches both a "repair" scenario and a "reserved upgrade" scenario in a deteriorated state, according to the combined effect of the first and second principles, the system will output the "repair" scenario as the target service scenario.
[0250] The third principle is the overdue payment prohibition principle. When the system detects that the value of the user's overdue amount field is greater than zero, the system prohibits any repair service scenario and the retention upgrade service scenario as the target service scenario. The service process is converted to the billing processing path and no active service instruction is generated.
[0251] The fourth principle is the high-value priority principle. When a user's high-value integration identifier is 1 or the customer star rating is greater than or equal to 4 stars, the user's retention and upgrade scenario will be prioritized in the outreach queue. The outreach channel can be upgraded to outbound calling or door-to-door service based on the default channel of the corresponding service priority. Under the same conditions, it will be given priority to enter the proactive outreach queue.
[0252] The fifth principle is the "pre-subscribed prohibition principle." When a user has already subscribed to an FTTR package or has an FTTR package subscription record in the system, the system will prohibit matching that user to the FTTR whole-house upgrade scenario. Instead, it will attempt to match the user to the FTTR sub-router installation scenario based on their actual lag situation. The above mutual exclusion rules are configured as rule conditions in the scenario matching engine. The matching engine executes the rule judgments in a fixed order and finally outputs the target service scenario identifier and the scenario parameters required for service instruction generation.
[0253] In the service instruction generation and delivery execution phases, the system introduces a customer segmentation mechanism to re-allocate delivery channels. Customer segmentation is based on a comprehensive assessment of users' social, commercial, and value attributes, classifying users into Category A, Category B, and Category C customers.
[0254] Category A customers refer to the home broadband accounts of key personnel or other important related personnel in government and enterprise groups. Identification criteria include whether the user is associated with the group's customer network, whether the user is on the account manager's service list, and whether the person handling the business is a key personnel of the group. Category A customers are identified in the system by customer codes, and their service is primarily accessed through a dedicated channel with account managers, emphasizing exclusive protection and a high-quality experience.
[0255] Category B customers refer to home broadband users in scenarios such as business customers or street-front shops. Identification criteria include whether the user's address belongs to an office building or street-front shop, and whether the user has a corresponding record in the business customer management system. Contacting Category B customers primarily involves telephone communication with account managers, with on-site visits arranged when necessary. The sales pitch emphasizes the value of operational efficiency and a stable network to shop operations.
[0256] Category C customers refer to high-value public users in ordinary residential households. Identification criteria include a user ARPU exceeding a preset high threshold, a package tier of high or ultra-high level, and a star rating of 4 stars or above. Reaching Category C customers is primarily handled by the installation and maintenance management team, with telephone communication emphasizing free health checks and the best experience.
[0257] For ordinary users not categorized into A, B, or C classes, the system does not perform additional customer classification processing and directly executes the default outreach channel corresponding to the service priority. Customer classification information is used in conjunction with service priority in the service decision module. When the number of users within the same priority level exceeds the daily outreach capacity, the system prioritizes users with higher customer classifications. This customer classification mechanism runs parallel to the aforementioned service priority, does not change the calculation result of the service priority, and only affects the selection of outreach channels and the adjustment of the outreach order.
[0258] Differentiated script templates are configured for both repair-type and upgrade-type service scenarios. The script templates for repair-type service scenarios focus on free inspection, equipment replacement, and network optimization. Example script content includes identifying the service personnel, explaining that the system has detected recent network fluctuations in the user's home broadband, inviting the user to undergo a free on-site network health check service. The check includes a Wi-Fi environment assessment, equipment testing, and network optimization; the check is free of charge, and the user is asked for a convenient time slot for the visit. This script template is applicable to all repair-type service scenarios and can be embedded in SMS, app push notifications, or outbound call scripts within the service instruction, using the script template number associated with the scenario identifier.
[0259] The script template for retaining upgraded services focuses on meeting existing standards, discounted upgrades, and exclusive benefits. Example scripts include explaining to users that the system has detected that their home network has gigabit capacity or is ready for whole-house coverage upgrades, inviting them to enjoy package discounts or free on-site inspections through a dedicated upgrade channel, and suggesting they contact their account manager or customer service hotline for details. This template is applicable to all retained upgraded service scenarios. For example, gigabit speed upgrade recommendations use gigabit package discount scripts, FTTR whole-house upgrades use whole-house wireless upgrade scripts, smart home expansion uses smart home bundle scripts, and loyalty enhancement uses dedicated follow-up and value-added experience scripts. The specific text of the script template can be edited by operations personnel in the knowledge base management module and then synchronized to the execution module via hot update without requiring a system restart.
[0260] The omnichannel outreach mechanism is configured internally by service priority and scenario type. SMS is used for notifications and confirmations for all service priorities; SMS messages have length limitations and are suitable for sending brief service requests and appointment confirmations. Carrier app push notifications are used for scenario-based pushes for second, third, and fifth priority services; in-app details with images and links can be displayed, making them suitable for retaining information on upgraded plans and application guidance.
[0261] The outbound calling channel is used for emergency communication of first and second priority levels. The customer service system dials the user's reserved number sequentially according to the outbound calling task list in the service instruction, and the outbound calling script calls the corresponding scenario's script template. The on-site work order channel is used for repair scenarios and surveying scenarios of first and second priority levels. The work order carries the user's address, target service scenario, recommended solution, and value-added threshold information, and is dispatched to the task queue of the installation and maintenance management system.
[0262] The WeChat Work channel is used for dedicated follow-up with Category A and Category B customers. Account managers send service reminders and processing links to users via WeChat Work. A dedicated follow-up channel for Category A customers and loyalty enhancement scenarios involves account managers conducting telephone follow-ups within 72 hours. Outreach data from all channels is collected centrally by the outreach execution module, including SMS sending status, app push click status, outbound call connection status, work order completion status, and WeChat Work conversation status. The collected results are written to the outreach log table for subsequent tag correction by the performance feedback module.
[0263] Scene rule DSL configuration and threshold parameter management are implemented in the knowledge base management module. System administrators can define scene triggering conditions and priority determination rules through a visual interface or text editing. A complete scene configuration includes scene identifier, scene name, scene priority, triggering condition list, service channel, script template number, and service value enhancement threshold fields. The triggering condition list is organized in a structured format, with each condition item including a field name, operator, and comparison value. Multiple condition items are connected by AND or OR logic. The system supports referencing any field from the user tag wide table in the configuration, such as the router wireless protocol version field, total number of lags field, positive capability identifier field, and network quality status field. After configuration, the system automatically generates an executable rule engine script without modifying the system code.
[0264] The threshold parameter management function centrally stores all thresholds used in the judgment process, including the deterioration threshold, mitigation threshold, consecutive lag days threshold, high-frequency lag threshold, comprehensive score interval endpoints, adaptive adjustment parameters for comprehensive score weights, device quantity threshold, preset traffic threshold, and value increase thresholds corresponding to each consumption level. Each threshold has a default value, which is calibrated and set based on actual production environment data in Jiangsu Province.
[0265] For example, the default threshold for high-frequency lag is 22 times per month, with a deterioration ratio of 1.5 and a mitigation ratio of 0.8. The threshold for consecutive lag days is 3 days, the threshold for segmented lag pain points is 5 times, the excellent score threshold is 85 points, the deterioration score threshold is 40 points, the mitigation / recovery score threshold is 80 points, the second / third priority dividing line for stable lag is 60 points, the improvement score threshold is 70 points, and the mandatory coverage threshold is 30 points. These default values are written into the parameter configuration table during system initialization. During operation, authorized personnel can adjust these settings by city, season, or package type, and adjustment records and effective times are recorded in the system. Through scenario rule configuration and threshold parameter management, the business side can adjust service strategies without changing the code. The system automatically loads new configuration parameters in daily batch processing tasks, ensuring the dynamic adaptability of service strategies.
[0266] In one or more embodiments of the present invention, the determination condition for the multi-device family identifier in the service experience tag is as follows: the number of router devices in the user's home is greater than or equal to 2, or the number of TV devices is greater than or equal to 2, or the number of robot vacuum cleaner devices is greater than or equal to 1, or the number of camera devices is greater than or equal to 2, or the number of air conditioner devices is greater than or equal to 3, and the number of mobile phone devices is greater than or equal to 7, and the total number of all connected devices is greater than or equal to 15. When all the above conditions are met, the multi-device family identifier is set to 1, indicating that the user's home has a large number of devices and is suitable for recommending a whole-house networking solution. When determining the condition, the system counts the number of each type of device from the connected device details wide table, compares the statistical results with the above thresholds one by one, and sets the multi-device family identifier to 1 when all conditions are met, otherwise it is set to 0.
[0267] The criteria for determining a large-family home candidate identifier are as follows: the user simultaneously meets the condition that the multi-device large-family identifier value is 1, the total number of lags in the past 30 days exceeds the preset severe lag threshold, and the user's family apartment type data meets the preset large-family home criteria. The preset severe lag threshold can be set to 22 times, consistent with the high-frequency lag threshold. The user's family apartment type data can be obtained from external data sources, such as apartment area data or apartment structure data obtained through web scraping. When the apartment area is greater than or equal to 110 square meters, or the apartment structure is three bedrooms or more, it is determined to meet the large-family home criteria. For users who meet all the above conditions, the large-family home candidate identifier value is 1, indicating that the user belongs to a large-family home and experiences severe network lag, making the FTTR whole-house networking solution the preferred recommendation.
[0268] The determination of the time period characteristic type is based on the distribution data of users' lag times over the past 7 days. The system counts the occurrence time of lag events for each user every day, dividing the day into the early morning, morning, afternoon, evening peak, and nighttime periods. The early morning period can be defined as 0:00 to 6:00, the morning period as 6:00 to 12:00, the afternoon period as 12:00 to 18:00, the evening peak period as 18:00 to 23:00, and the nighttime period as 23:00 to 24:00. The system calculates the percentage of lag occurrences in each time period. When the percentage of lag occurrences in a certain time period exceeds a preset threshold, for example, exceeding 50%, the user is marked as the concentrated type for that time period. When the time period distribution is relatively even, and the percentage of lag occurrences in any time period is below the preset threshold, it is marked as the evenly distributed type throughout the day. This time period characteristic type is written into the time period characteristic field in the business experience tag.
[0269] The consumption level is determined based on the user's historical consumption data. The system obtains the user's ARPU (Average Revenue Per User) and household consumption data for the past three months from the user's subscription information or billing system. The ARPU reflects the user's communication consumption level, while household consumption data can include the user's household spending on value-added services, smart home devices, etc. The system weights the ARPU and household consumption data according to a preset ratio to obtain a comprehensive consumption level score. When the comprehensive consumption level score is higher than the preset high consumption threshold, the consumption level is judged as high; when the comprehensive consumption level score is lower than the preset low consumption threshold, it is judged as low; otherwise, it is judged as medium. The preset high and low consumption thresholds can be configured according to the local economic level and operational strategies.
[0270] In one or more embodiments of the present invention, the identification of the actual number of sub-modems in the FTTR sub-router installation scenario is accomplished through the modem topology table. The modem topology table records the topology relationship between the home gateway and its downstream devices, specifically named `ldapd_gateway_down_trmnl_topo_day`. This table contains a device hostname field, which records the name of each downstream device. When performing FTTR sub-router installation scenario matching, the system reads all downstream device records from the modem topology table and filters out records whose device hostname field contains the string "FTTRSUB". The string "FTTRSUB" is the naming convention for FTTR sub-modem devices in the topology table, and fuzzy matching is performed using the condition `LIKE '%FTTRSUB%'`. The system counts the number of matched records, which represents the actual number of FTTR sub-modems installed in the user's home.
[0271] The system compares the actual number of sub-optical modems with the agreed-upon number of sub-optical modems in the user's FTTR subscription information. The subscription information records the agreed-upon number of sub-optical modems when the user subscribed to the FTTR package. When the actual number of sub-optical modems is 0 and the agreed-upon number is greater than or equal to 1, it is determined that the package includes sub-optical modems but none are actually installed; when both the actual and agreed-upon number of sub-optical modems are 0, it is determined that the package does not include sub-optical modems; when the actual number of sub-optical modems is greater than 0 but less than the agreed-upon number, it is determined that the actual number installed is insufficient. The results of this comparison serve as part of the triggering conditions for FTTR sub-router installation scenarios, identifying users who need to install sub-optical modems. Through this identification method, the system can accurately grasp the completeness of a user's home FTTR network, providing an accurate basis for installation and maintenance dispatch.
[0272] In one or more embodiments of the present invention, the large-screen video application whitelist includes video playback applications running on large-screen smart terminals (e.g., the whitelist includes applications such as "XX TV Version," "XX TV," and "XX Video"). The applications in the whitelist are used to identify stuttering records of large-screen video applications in the stuttering event data. During matching, the system compares the application name field in the stuttering event record with the names in the whitelist one by one; a match is counted as a stuttering event in the large-screen video count. Furthermore, the system also uses the terminal type field for auxiliary filtering; when the terminal type is a TV or set-top box, it is prioritized for inclusion in the large-screen video stuttering statistics to improve identification accuracy.
[0273] In one or more embodiments of the present invention, the determination of the business preference type is based on the distribution of user usage frequency under each application category in the past 30 days, combined with the distribution of lag events under each application category. The application categories are pre-divided into nine categories, specifically including video, games, social, office, education, finance, e-commerce, application distribution, and tools. The system extracts the number of times the user has used each application in the past 30 days from the application business usage record table, and groups the usage counts into the nine categories according to the mapping relationship between application names and application categories, calculating the usage count percentage of each category. When the usage count percentage of a certain category is higher than a preset preference threshold, that category is determined as the user's business preference type. When the usage count percentages of multiple categories are all higher than the preset preference threshold, the category with the highest usage count percentage is selected as the business preference type. When the usage count percentages of all nine categories are lower than the preset preference threshold, the system uses the distribution of lag events for auxiliary judgment, selecting the application category with the highest percentage of lag events as the business preference type.
[0274] The criteria for identifying a high-concurrency family is that the number of resident devices in the user's family is greater than or equal to 15. The number of resident devices refers to the total number of downstream devices appearing in the user's family network within a preset statistical period. The system obtains the number of resident devices by counting the number of device records for each user from the downstream device details wide table. When the number of resident devices is greater than or equal to 15, the high-concurrency family identifier is set to 1, indicating that the probability of devices in the user's family accessing the network simultaneously is relatively high, classifying it as a high-concurrency family. When the number of resident devices is less than 15, the high-concurrency family identifier is set to 0. The difference between the high-concurrency family identifier and the multi-device family identifier is that the high-concurrency family identifier relies solely on the total number of resident devices for determination, while the multi-device family identifier requires the simultaneous fulfillment of multiple conditions, such as a specific number of devices and a number of mobile phones.
[0275] The method for determining ITV inactivity is as follows: the system retrieves the number of ITV power-ons and the duration of ITV usage within the past 30 days from the ITV service usage records. If a user has subscribed to ITV service, but the number of ITV power-ons in the past 30 days is zero or the ITV usage duration is less than a preset inactivity threshold, the user is considered inactive. The preset inactivity threshold can be set to 30 minutes, meaning that a user's cumulative ITV usage time within the past 30 days is less than 30 minutes, indicating inactivity. The system writes the ITV inactivity status into the ITV inactivity flag field in the pre-defined basic tag, with a value of 1 indicating inactivity and a value of 0 indicating activity. In scenarios involving large-screen video buffering optimization, the system reads the value of this flag field to distinguish between inactive ITV users and non-ITV users, and then selects the corresponding service action. The number of ITV power-ons can be obtained from the heartbeat data reported by the set-top box or the power-on records of the service platform, and the ITV usage duration can be obtained by summarizing the on-demand and live streaming usage records of the service platform.
[0276] The data source for the number of sub-optical modems (SAMs) specified in the package plan comes from the "Scheduled Sub-optical Modem Quantity" field in the package subscription information table. During the FTTR sub-router installation scenario matching process, the system reads the value of the "Scheduled Sub-optical Modem Quantity" field from the package subscription information table as the number of SAMs specified in the package. This field records the number of SAMs specified when the user subscribed to the FTTR package; a value of 0 indicates that the package does not include any SAMs, a value of 1 indicates that the package includes one SAM, and so on. The system compares the number of SAMs specified in the package with the actual number of SAMs obtained from the optical modem topology table. When the actual number of SAMs is less than the number specified in the package, it is determined that there is a SAM shortage, triggering the FTTR sub-router installation scenario's condition check.
[0277] Regarding system interface specifications, modules interact through a standardized message bus, with data transmitted in JSON string format. When sending messages to the message bus, each module organizes data using agreed-upon JSON key names, and when receiving messages, it parses the data using the same set of JSON key names. For example, when the tag calculation module sends user tag set data to the scene matching engine, the message body includes broadband account key, status key, rating key, and tag key, with various tag field names and values nested under the tag key. The JSON format ensures compatibility with data interaction needs between different modules. The knowledge base management module sends configuration parameters to other modules via a hot update interface. The hot update interface uses the HTTP protocol, with both the request and response bodies in JSON format. During system deployment, each module registers its own queue name in the message bus, enabling asynchronous decoupled communication between modules through message queues, thus supporting daily batch processing tasks for tens of millions of users.
[0278] In terms of production verification, the method has been deployed and verified in a real production environment for home broadband users in Jiangsu Province. The system begins collecting multi-source data at 2:00 AM daily, completing full tag calculation, status determination, scenario matching, and service priority generation for tens of millions of users within 6 hours, and begins reaching and scheduling users at 8:00 AM. The daily data processing volume is approximately 5TB, covering details of home broadband outage events, device maintenance tables, application usage records, etc. Approximately 2 million users are reached daily. The success rate for repair-related scenarios is approximately 75%, and the conversion rate for upgrade-related scenarios is approximately 15%.
[0279] This invention also provides a proactive service system for home broadband users based on network quality tags, comprising:
[0280] The tag calculation module is used to generate network lag root cause tags, positive capability tags, and service experience tags based on multi-source data of home broadband users; based on these tags, it performs lag event statistics using a first and second sliding time window to determine the network quality status of home broadband users, including no lag, deterioration, mitigation, stable lag, excellent and stable, and improving status; and it calculates a comprehensive quality score based on the network lag root cause tags and positive capability tags.
[0281] The scenario matching engine is used to match target service scenarios from repair service scenarios and retention and upgrade service scenarios based on network card failure root cause segment tags, positive capability tags, service experience tags, and network quality status.
[0282] The service decision module is used to determine service priorities based on network quality status, comprehensive quality score, and target service scenarios.
[0283] The reach execution module is used to send service instructions to home broadband users according to service priority.
[0284] The present invention also provides the following embodiments:
[0285] Example 1
[0286] In one embodiment of the present invention, the entire process of replacing an old router is illustrated using a bundled broadband user in Nanjing as an example. This user's broadband account identifier is 320100XXXXXX, and their package is a 129 yuan bundled package, classifying them as a high-value user. System statistics show that this user experienced a total of 38 instances of buffering within 30 days, meeting the criteria for a high-frequency buffering user. Device capability identification results show that this user's home only has a WiFi 4 router, with no WiFi 5, WiFi 6, or WiFi 7 routers. The number of buffering incidents on the router device segment is 12. The user has not changed their package in the past 3 months, is not an FTTR user, and the top root cause segment in the first sliding time window is the router device segment. The network quality status is deteriorated, with a comprehensive quality score of 52 points. The service experience tag indicates that the user's buffering periods are mainly concentrated during the evening peak hours.
[0287] Following the end-to-end process, at 02:00 on day T, the data acquisition module performs a data acquisition task, pulling detailed lag data, device dimension table data, and package data from the Hive offline wide table. From 03:00 to 05:00 on day T, the tag calculation module performs L1 tag generation, outputting the user's Class A network lag root cause segment tag, where the routing device segment is the main fault segment; the Class B positive capability tag indicates that the user only has WiFi4 router capabilities; and the Class C service experience tag indicates that the time period is concentrated during the evening peak. At 05:00 on day T, the tag calculation module performs L2 diagnostic tag generation, outputting that the user's network quality status is deteriorated, with a comprehensive quality score of 52 points, and the TOP1 root cause segment is the routing device segment. At 05:30 on day T, the scenario matching engine performs scenario matching, and the user matches the scenario of replacing an old router. At 06:00 on day T, the service decision module outputs that the user's service priority is the second priority, corresponding to the deteriorated status and a comprehensive quality score higher than or equal to the first threshold. At 08:00 on day T, the outreach module, based on customer segmentation rules, assigned this high-value user to the customer service specialist channel. The order was then dispatched through the high-end installation and maintenance team, with the service instruction including the template number for replacing the WiFi 6 router and the details of free on-site installation. At 14:00 on day T, the installation and maintenance personnel contacted the user to schedule a visit. The script included reminding the user of recent network fluctuations in their home broadband and informing them of the company's arrangement for a free on-site network health check, including WiFi environment assessment, equipment testing, and network optimization. The personnel also inquired about the user's convenience, asking if Saturday morning or Sunday was a convenient time. On day T+1, the installation and maintenance personnel conducted an on-site inspection using an eight-step standardized process. The eight-step standardized troubleshooting process includes: first, checking the status of the optical modem's indicator lights and whether the line connection is normal; second, using testing equipment to test the incoming optical power and link attenuation; third, checking the number of downstream terminals and bandwidth usage of the gateway device; fourth, checking the model, protocol version, and operating status of the wireless router; fifth, using wireless signal testing equipment to measure the wireless coverage strength in all rooms of the house; sixth, checking the connection speed and network configuration of the user's terminal devices; seventh, testing the jitter, packet loss, and latency indicators of the home LAN link; and eighth, summarizing the troubleshooting conclusions and generating a test report. The user's troubleshooting conclusion was that the old router was causing network lag. The installation and maintenance personnel recommended a WiFi 6 router replacement plan, with a monthly fee increase starting at 60 yuan. The user accepted the replacement plan on the spot. On T+7, the effect feedback module conducted a follow-up phone call. The user rated the service as 5 stars and was satisfied. The system generated a successful repair feedback tag and wrote it back to the tag calculation module. On subsequent tag calculation days, the user's network quality status was switched to the alleviated state due to a significant decrease in the frequency of lag. After the number of lags returned to zero within the second sliding time window, it was switched to the no-lag state or the excellent and stable state.
[0288] Example 2
[0289] In one embodiment of the present invention, the entire process of adding an FTTR sub-router is illustrated using a Suzhou bundled broadband user as an example. This user's broadband account identifier is 320500XXXXXX, and their package is a 199 yuan bundled package, classifying them as a high-value user. Device identification results show that the user's home gateway is the FTTR main optical modem, model HN8156XR. The package includes one sub-optical modem, but the actual number of sub-optical modems is 0, indicating an abnormal state of FTTR1+0 with no actual sub-optical modems. The user's gateway WiFi segment experienced 8 instances of lag, with a total of 27 lags over 30 days. There are 22 resident devices, and the network quality is described as stable with occasional lag, resulting in a comprehensive quality score of 63.
[0290] The tag calculation module outputs L1 tags. The Class B positive capability tag indicates the user has installed the FTTR main optical modem but has no sub-gateway. The Class A network lag root cause tag indicates poor quality of the gateway WiFi segment is the main root cause of lag for this user. The tag calculation module outputs L2 tags, indicating a stable but laggy network quality status with a comprehensive quality score of 63 points. The top root cause segment is the gateway WiFi segment. The scenario matching engine matches the FTTR sub-router installation scenario. The service decision module outputs the user's service priority as the third priority. This user is a high-value user, and the contact channel is upgraded to on-site installation and maintenance service through the customer classification mechanism. The contact execution module dispatches the order to the installation and maintenance channel, and installation and maintenance personnel conduct on-site surveys of fiber optic supplementation points throughout the house. During the survey, the installation and maintenance personnel carry a PDA device, a portable handheld terminal used daily by the installation and maintenance personnel, capable of receiving data collected by the wireless signal detection module. The installation and maintenance personnel used a PDA device with a wireless signal detection device to collect on-site WiFi coverage heatmaps for each room in the user's home. The coverage strength data was displayed on the PDA screen as a heatmap, visually showing the user the locations of coverage blind spots. As per the package agreement, the personnel installed a sub-optical modem free of charge for the user, upgrading the FTTR network from 1+0 to 1+1. They also introduced a whole-house WiFi upgrade plan to the user, with monthly fee increases starting at 90 yuan, which the user could choose voluntarily. On T+7, the feedback module conducted a follow-up visit, and the user confirmed a significant improvement in wireless coverage and a decrease in the number of buffering incidents. The system generated a successful repair feedback tag and wrote it back.
[0291] Example 3
[0292] In one embodiment of the present invention, the gigabit speed-up recommendation process is illustrated using a bundled broadband user in Nantong as an example. This user's broadband account identifier is 320600XXXXXX, their current package is a 99 yuan bundled package with a speed of 500 Mbps. Device capability data shows that the user's router supports the WiFi 6 protocol, the router's LAN port speed is 1000 Mbps, the monthly data usage is 380 gigabytes, and the gigabit-ready flag is set to 1. This user is not a high-value bundled user, but their average monthly spending is 108 yuan, classifying them as a mid-value user (a mid-value user refers to a user with an average monthly spending greater than or equal to 99 yuan and less than 129 yuan). The network quality status is excellent and stable, with a comprehensive quality score of 88 points, including a capability score of 85 points, an experience score of 95 points, and a stability score of 85 points.
[0293] The tag calculation module outputs the L1 tag, where the Gigabit Ready flag for the Class B positive capability tag is set to 1. The tag calculation module also outputs the L2 tag, indicating a good and stable state with a comprehensive quality score of 88. The scenario matching engine matches the recommended Gigabit speed-up scenario, triggering conditions that simultaneously meet the following criteria: Gigabit Ready, non-Gigabit plan, monthly data usage exceeding 300 gigabytes, good and stable state, comprehensive quality score greater than 85, and median user identity. The service decision module outputs that the user's service priority is the fifth priority, belonging to the reserved operation category. The outreach execution module pushes a message through the operator's application, informing the user that the network has reached gigabit capability, and that upgrading to a gigabit plan will enjoy a 20% discount, improving video and gaming experiences. After clicking the in-app push message, the user is redirected to the gigabit plan upgrade page, completes the upgrade, and subscribes to the 199 yuan gigabit bundled plan. On T+7 days, the effect feedback module performs a follow-up visit, and the user reports improved video and gaming experiences; the system generates a successful upgrade feedback tag.
[0294] Example 4
[0295] In one embodiment of the present invention, the entire process of recommending smart home products is illustrated using a single broadband user in Taizhou as an example. This user's broadband account is identified as 321200XXXXXX, the package is single broadband, and the monthly fee is 139 yuan. Within 30 days, this user experienced 9 instances of smart home device-related lag. The top five types of devices experiencing lag included cameras, door locks, and doorbells. There were 11 lag instances on the LAN link segment. The number of resident devices was 18. The smart home demand identifier was 1, the "elderly family member" identifier was 1, the consumption level was high, the service preference type was comprehensive, there was no record of smart home package revenue, the network quality status was stable with occasional lag, and the overall quality score was 62 points.
[0296] The tag calculation module outputs L1 tags, where the Class A network lag root cause tag indicates poor LAN link quality, and the Class C service experience tag indicates a comprehensive service preference type. The tag calculation module also outputs L2 tags, indicating a stable lag state with a comprehensive quality score of 62, and the top root cause segment is the LAN link. The scenario matching engine matches the smart home expansion scenario, triggering conditions that simultaneously meet the following criteria: smart home device quality issues greater than or equal to 3 times, LAN link lag issues greater than or equal to 5 times, smart home demand identifier of 1, high consumption level, and comprehensive service preference type. The service decision module outputs that the user's service priority is the third priority, corresponding to a stable lag state and a comprehensive quality score higher than or equal to the second threshold. This user is a high-value user, and the customer grading mechanism upgrades the outreach channel to on-site installation and maintenance service. The outreach execution module dispatches the order to the installation and maintenance channel, where installation and maintenance personnel perform on-site network optimization and simultaneously recommend doorbell and camera kits. The installation and maintenance personnel's network optimization operations included checking the quality of the network cable connections on the LAN side, optimizing the configuration of the gateway's WiFi channel, and adjusting the network parameters of the smart home devices to re-stabilize the connection between each smart home device and the home network. The user accepted the recommended doorbell and camera set, completing the smart home expansion. On T+7 days, the feedback module performed a follow-up visit; the user confirmed that the smart home devices were stably online, and the system generated a successful expansion feedback label.
[0297] Example 5
[0298] In one embodiment of the present invention, the daily end-to-end process for millions of users is illustrated using a real production environment in Jiangsu Province as an example. The system covers approximately 20.47 million household broadband users and executes a complete batch processing flow once daily. The data acquisition module starts at 02:00 on day T, pulling multi-source wide table data from Hive. Each user involves more than 200 fields, and the data sources include the lag details table, equipment maintenance table, package table, complaint table, and application table. From 02:00 to 03:00 on day T, the data acquisition module performs data preprocessing, including multi-source table association, field standardization, empty table sentinel filling, and volatility filtering, outputting a standardized event stream for use by the tag calculation module. From 03:00 to 04:30 on day T, the tag calculation module performs L1 tag generation, covering approximately 20.47 million users, generating more than 80 L1 tag fields per user. From 04:30 to 05:00 on day T, the tag calculation module performs L2 diagnostic tag generation, covering approximately 20.47 million users, generating more than 30 L2 tag fields per user. From 05:00 to 05:30 on day T, the scenario matching engine executed rule matching, with an average of 0.3 service scenarios matched per user. From 05:30 to 06:00 on day T, the service decision module output the service priority distribution, where first-priority users accounted for approximately 0.5% of all users, second-priority users accounted for approximately 2%, third-priority users accounted for approximately 5%, fourth-priority users accounted for a small proportion (these users were only monitored and not actively contacted), fifth-priority users accounted for approximately 8%, and the remaining users did not generate proactive service instructions. The above percentages are based on the total number of users; the actual reach is constrained by the daily reach capacity limit, and any excess will be processed on the next day. From 06:00 to 08:00 on day T, the reach execution module completed reach scheduling, reaching approximately 2 million users in a single day. Scheduling methods included batch SMS delivery, outbound call queue generation, and installation and maintenance work order dispatch. From 08:00 to 23:00 on day T, each channel executed actual reach operations, with a reach success rate of approximately 75%. On day T+1, the performance feedback module collects outreach results. The success rate for corrective scenarios is approximately 75%, while the conversion rate for upgrade scenarios is approximately 15%. Data statistics are based on Jiangsu Province's production data for the 202603 payment period. On day T+7, the performance feedback module performs telephone follow-ups and satisfaction surveys, completing tag corrections and status updates. The knowledge base management module continues to run, allowing the business side to adjust thresholds, script templates, and priority parameters at any time according to operational needs. Adjusted parameters take effect in the next batch of processing tasks via hot updates. The entire end-to-end process, from data collection at 02:00 on day T to outreach at 08:00 on day T, completes full tag calculation and scenario matching for millions of users within 6 hours.
[0299] This invention overcomes the limitations of existing technologies that only focus on the presence or absence of network lag by introducing three types of tags: root cause segment of network lag, positive capability, and service experience. Specifically, the positive capability tag identifies users with normal network status but upgrade potential, allowing for precise targeting of healthy user groups with upgrade-oriented services; the root cause segment of network lag tag pinpoints the specific fault segment and root cause of lag, providing a clear basis for remedial services; and the service experience tag further supplements user preferences and household characteristics, supporting differentiated matching of service scenarios. Thus, a multi-dimensional characterization of all household broadband users is achieved within a single technical system, covering both users experiencing lag requiring remediation and healthy users requiring maintenance, filling the gaps in existing technologies regarding positive capability identification and healthy user operation.
[0300] Based on statistical analysis using a dual sliding time window and six-state network quality status identification, the changing trends in user network quality can be perceived in a timely manner, including dynamic processes such as deterioration, mitigation, and improvement. This allows for proactive service triggering before users perceive significant deterioration, overcoming the lag problem caused by existing technologies relying on static threshold judgments. Furthermore, by combining network quality status, comprehensive quality score, and target service scenarios to determine service priorities, quality evaluation results can be automatically mapped to specific repair or preservation / upgrade service actions. This improves the matching degree between service content and actual user needs, and also improves the efficiency of service resource deployment, ultimately forming a unified proactive service system covering all online users.
[0301] The above provides a detailed description of a proactive service method and system for home broadband users based on network quality tags. Specific examples have been used to illustrate the principles and implementation methods of this invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this invention. At the same time, those skilled in the art will recognize that there will be changes in the specific implementation methods and application scope based on the ideas of this invention. Therefore, the content of this specification should not be construed as a limitation of this invention.
Claims
1. A proactive service method for home broadband users based on network quality tags, characterized in that, The method includes: Based on multi-source data from home broadband users, generate root cause segment labels for network card usage, positive capability labels, and service experience labels; Based on the root cause segment label of network lag, the positive capability label, and the service experience label, the first sliding time window and the second sliding time window are used to perform lag event statistics to determine the network quality status of the home broadband user; the network quality status includes no lag, deterioration, mitigation, stable lag, excellent and stable, and improving status. A comprehensive quality score is calculated based on the network card root cause segment label and the positive capability label; Based on the network lag root cause segment label, the positive capability label, the service experience label, and the network quality status, the target service scenario is matched from the repair service scenario and the retention upgrade service scenario; Service priorities are determined based on the network quality status, the comprehensive quality score, and the target service scenario, and service instructions are sent to the home broadband users according to the service priorities.
2. The method according to claim 1, characterized in that, The network lag root cause segment label includes lag indicators for six fault segments: user terminal segment, routing device segment, LAN side link segment, gateway WiFi segment, gateway segment, and Internet segment. The lag indicators for the fault segment include the number of lags, the number of lag days, the maximum number of consecutive lag days, the top terminal type, the top application name, and the top peak time period.
3. The method according to claim 2, characterized in that, The positive capability labels include gigabit ready label, FTTR whole-house coverage label, large screen high definition label, 5G terminal ownership label, smart terminal ecosystem label, and device new / old file label. The gigabit-ready flag is determined based on the router's WiFi protocol version or LAN port speed. The FTTR whole-house coverage identifier is determined based on the package subscription type or equipment model.
4. The method according to claim 3, characterized in that, The first sliding time window has a duration of 7 days, and the second sliding time window has a duration of 30 days. Determining the network quality status of the home broadband user includes: The system records the number of times the home broadband user experienced lag, the number of days of lag, and the maximum number of consecutive lag days within the first sliding time window, as well as the number of times the home broadband user experienced lag, the number of days of lag, and the maximum number of consecutive lag days within the second sliding time window. Calculate the ratio of the average daily number of stutters in the first sliding time window to the average number of stutters in the second sliding time window; When the ratio is greater than a preset deterioration threshold and the maximum number of consecutive stuck days within the first sliding time window is greater than or equal to a preset consecutive days threshold, the network quality status is determined to be in a deteriorated state. When the ratio is less than the preset mitigation threshold and there is a stuttering record within the second sliding time window, the network quality status is determined to be in a mitigation state. When the number of stutters within the second sliding time window is zero, and the number of tags with a value of 1 in the positive capability tags is greater than or equal to 2, and the household value index is greater than or equal to the preset index threshold, the network quality status is determined to be excellent and stable. When the number of stutters within the second sliding time window is zero, and the number of tags with a value of 1 in the positive capability tag is less than 2 or the family value index is less than the preset index threshold, the network quality status is determined to be a stutter-free state. When the number of stutters within the second sliding time window is not zero, the home broadband user does not meet the determination conditions for deterioration and mitigation, and the network quality status was a stable stuttering state or an improving state during the last status update, and the number of consecutive zero stuttering days reaches the preset confirmation number of days, the network quality status is determined to be an improving state. When the network quality status does not belong to the deterioration state, the mitigation state, the excellent and stable state, the no-stutter state, or the improving state, the network quality status is determined to be the stable stutter state. The family value index is a component of the service experience label and is used to characterize the value level of the family equipment of the family broadband user.
5. The method according to claim 4, characterized in that, The process of determining the network quality status of the home broadband user also includes a status anti-jitter mechanism: When the home broadband user switches from the current state to the target state, the entry conditions for the target state must be met for a preset number of consecutive confirmation days. After the state transition is completed, a cooldown period is set, during which it is prohibited to switch back to the state before the transition. When a user complaint occurs or the overall quality score is lower than the preset mandatory coverage threshold, the confirmation day condition and the cooling-off period are ignored, and the network quality status is forcibly covered to a deteriorated state.
6. The method according to claim 5, characterized in that, The calculation of the overall quality score includes: The experience score is calculated based on the lag index in the network lag root cause segment label. The experience score is inversely related to the number of lags, the number of lag days, and the maximum number of consecutive lag days. Ability scores are calculated based on the values of each identifier in the positive ability labels and preset weights. A stable score is calculated based on the number of days for which the number of times the lag occurs in a single day exceeds a preset burst threshold and the difference between the experience score and the first sliding time window and the second sliding time window. The overall quality score is obtained by weighted summing of the experience score, the ability score, and the stability score. The weights of the experience score, the capability score, and the stability score are adjusted according to the network quality status.
7. The method according to claim 6, characterized in that, The method for determining the TOP1 root cause segment among the six fault segments is as follows: The number of pauses in the six fault segments are sorted. When there are two or more fault segments with the same number of pauses, the maximum number of consecutive pause days, the significance during peak periods, and the repairability level of the fault segment within the corresponding sliding time window are compared sequentially to determine the TOP1 root cause segment. The repairability levels, from highest to lowest, are as follows: LAN-side link segment, gateway WiFi segment, and routing device segment are tied for the highest level, followed by user terminal segment, gateway segment, and Internet segment.
8. The method according to claim 7, characterized in that, The repair service scenarios include replacing old routers, adding FTTR sub-routers, adding Tianyi gateway sub-routers, and optimizing large-screen video stuttering. The retained upgrade service scenarios include gigabit speed enhancement recommendation scenario, FTTR whole-house upgrade scenario, smart home product expansion scenario, and loyalty reinforcement scenario.
9. The method according to claim 8, characterized in that, The step of determining service priority based on the network quality status, the comprehensive quality score, and the target service scenario includes: When the network quality status is deteriorating, the overall quality score is lower than the first threshold, and the target service scenario is a repair service scenario, the service priority is determined to be the first priority. When the network quality status is deteriorating, the comprehensive quality score is higher than or equal to the first threshold, and the target service scenario is a repair service scenario, the service priority is determined to be the second priority. When the network quality status is stable but laggy, the overall quality score is lower than the second threshold, and the target service scenario is a repair service scenario, the service priority is determined to be the second priority. When the network quality status is stable but laggy, the comprehensive quality score is higher than or equal to the second threshold, and the target service scenario is a repair service scenario or a retention upgrade service scenario, the service priority is determined to be the third priority. When the network quality status is in a mitigated state, the overall quality score is higher than the third threshold, and no service scenario is hit, the service priority is determined to be the fourth priority. When the network quality status is excellent and stable, the comprehensive quality score is higher than the fourth threshold, and the target service scenario belongs to the retention and upgrade service scenario, the service priority is determined to be the fifth priority. When the network quality status is in the "improving" state, the overall quality score is higher than the preset "improving" score threshold, and the target service scenario belongs to the "retain and upgrade" service scenario, the service priority is determined to be the fifth priority; wherein, the preset "improving" score threshold is lower than the fourth threshold; The urgency of services decreases sequentially from the first priority to the fifth priority.
10. A proactive service system for home broadband users based on network quality tags, characterized in that, include: The tag calculation module is used to generate root cause tags, positive capability tags, and service experience tags for network cards based on multi-source data of home broadband users. Based on the root cause segment label of network lag, the positive capability label and the service experience label, the first sliding time window and the second sliding time window are used to perform lag event statistics to determine the network quality status of the home broadband user. The network quality status includes no lag, deterioration, mitigation, stable lag, excellent and stable, and improving status. And calculate the overall quality score based on the network card root cause segment label and the positive capability label; The scenario matching engine is used to match the target service scenario from the repair service scenario and the retention and upgrade service scenario based on the network card failure root cause segment label, the positive capability label, the service experience label and the network quality status. The service decision module is used to determine service priorities based on the network quality status, the comprehensive quality score, and the target service scenario. The reach execution module is used to send service instructions to the home broadband user according to the service priority.
