Mobile communication network quality monitoring system

Abstract

The invention relates to a mobile communication network quality monitoring system. The system mainly comprises a multi-source scene binding type acquisition module, a data storage and processing module, a linkage type business management and control module and a double-linkage situation visualization module, the method comprises the following steps: constructing a multi-index collaborative network quality amplitude estimation model by acquiring network quality data sampled by a multi-source scene, and acquiring an estimated value of a network quality amplitude; the linkage type service management and control module is used for acquiring monitoring information of network quality through a dynamic mapping mechanism, predicting an abnormal duration and associated abnormal indexes, and pushing classified information through a hierarchical permission display mechanism; the double-linkage situation visualization module obtains network quality monitoring data of a corresponding test site through a front-end interface scenarized layout design of the management system; real-time situation monitoring and multi-dimensional data interaction are utilized to obtain visual presentation and structured traceability of data. And the quality monitoring of the mobile communication network is realized.

Description

Technical Field

[0001] This invention belongs to the field of communication technology, and specifically relates to a mobile communication network quality monitoring system. Background Technology

[0002] In the current context of accelerated digital transformation, mobile networks have become a core infrastructure for social operation, economic development, and public services. Their quality directly impacts user experience, enterprise operational efficiency, and the digitalization process of various industries. However, the field of mobile communication network quality monitoring currently faces multiple bottlenecks, severely restricting network optimization efficiency and user experience improvement. These bottlenecks manifest themselves in the following ways: Data fragmentation: Network test data is scattered across multiple channels such as special tests, crowdsourced testing, OTT platform feedback, and user complaints and reports. The multi-source data is stored in a scattered manner and lacks a unified association mechanism. The field naming systems of different data sources are very different, the timestamps are not synchronized, and the spatial location association is vague, making it difficult to combine data value and lacking complete data support for the whole-link quality assessment.

[0003] Insufficient accuracy in scene binding: The binding of data collection with physical scenes and terminal devices lacks deep correlation; the division of administrative regions and the classification of scene types are not clear enough; the lack of on-site environmental data collection in special tests makes it difficult to accurately locate regional quality bottlenecks; there are blind spots in fault diagnosis; and the problem resolution cycle is long.

[0004] The reliability of data processing is lacking: the raw data contains outliers, missing values, and timestamp asynchrony, making it difficult for traditional data processing methods to balance data smoothness and feature retention. This affects the accuracy of subsequent analysis results and fails to provide a stable and reliable data foundation for network quality assessment.

[0005] Inefficient anomaly warning and push notifications: There is a lack of dynamic anomaly judgment criteria adapted to different scenarios, the prediction accuracy of anomaly duration is low, and the identification of associated anomaly indicators is incomplete; the push notification of warning information does not take into account user permissions and management needs, resulting in information overload or insufficient targeting, making it difficult for front-line maintenance personnel to quickly obtain key handling information.

[0006] Limited Filtering Dimensions: Traditional monitoring tools' filtering functions are limited to a single dimension or a few core indicators, making it difficult to meet the needs for accurate data retrieval across multiple scenarios and conditions. For example, they cannot simultaneously combine multiple dimensions such as region, operator, network type, test time period, speed range, and signal strength for filtering, resulting in difficulty in quickly locating network problems in specific scenarios.

[0007] Poor visualization and traceability experience: The visualization of monitoring results lacks hierarchical and multi-dimensional interactive design, making it impossible to achieve in-depth drilling from the overall situation to local test points; the data traceability process is cumbersome, the operation of switching between multiple systems is complicated, and it is difficult to quickly complete the full-link data traceability and problem location. Summary of the Invention

[0008] To address the aforementioned problems in the existing technology, this invention provides a mobile communication network quality monitoring system.

[0009] The objective of this invention can be achieved through the following technical solutions: a multi-source scene-binding acquisition module, a data storage and processing module, a linkage business management module, and a dual-linkage situation visualization module; The multi-source scene-bound acquisition module constructs a three-dimensional data acquisition system through preset data test points to obtain network quality data sampled from multiple source scenes; The data storage and processing module performs field-level mapping and alignment on the associated database based on the network quality data, normalizes and stores multi-dimensional fields according to spatiotemporal labels, and constructs a network quality prediction associated dataset. Based on the associated dataset, the data is normalized in multiple orders to generate a normalized steady-state data stream. The network quality feature vector is obtained by using sliding window energy extraction and time-series trend extrapolation. The estimated value of the network quality amplitude is obtained by constructing a multi-indicator collaborative network quality amplitude estimation model. The linkage-based business management module obtains network quality monitoring information based on the estimated value of the network quality amplitude through a dynamic mapping mechanism; it manages and binds the network quality monitoring information with multi-dimensional data, predicts the duration of abnormalities and associated abnormal indicators, and obtains early warning information packages; based on the early warning information packages, it pushes classified information through a hierarchical permission display mechanism. The dual-linkage situation visualization module acquires network quality monitoring data for the corresponding test locations through the scenario-based layout design of the management system's front-end interface; by utilizing real-time situation monitoring and multi-dimensional data interaction, it constructs a network quality visualization system with dual-view linkage, obtaining an intuitive presentation and structured traceability of the data.

[0010] As a preferred technical solution of the present invention, the construction of a three-dimensional data acquisition system through preset data test points includes collecting multiple indicators of multiple preset data test points, associating them with administrative regions and scene types, and requiring the collection of on-site environmental data in special test scenarios. The deep binding of indicator data, physical scenes and terminal devices constructs a three-dimensional data acquisition system.

[0011] Specifically, the process of performing field-level mapping and alignment on the associated database includes converting network quality data fields and target fields in the associated database into high-dimensional semantic vectors through a pre-trained network data semantic encoder, and obtaining automatic mapping across naming systems based on vector cosine similarity. The field-level alignment adopts an adaptive time granularity alignment algorithm, which dynamically adjusts the time window according to the data density, corrects asynchronous timestamps through linear interpolation, and associates multi-source field data at the same time node; based on the network topology digital twin model, the multi-source field data is bound to the digital twin of the physical network node, and spatial position alignment across devices is obtained through spatial coordinate encoding.

[0012] Specifically, the construction process of the network quality prediction associated dataset includes: assigning time granularity labels and spatial topology labels to multidimensional fields through a spatiotemporal label dynamic binding mechanism to anchor the fields to the network physical scenario; and using the spatiotemporal labels and field features to construct a two-layer index through an adaptive storage architecture for the associated dataset to obtain the associated dataset for network quality monitoring that enables fast source tracing and multi-dimensional associated queries. Specifically, the multi-order regularization method involves using the associated dataset of network quality monitoring as initial input data, and using the anomaly-free data obtained after processing by the spatiotemporal outlier detection model as the first-order regularization value. Based on the first-order regularization value, a multi-scale dynamic smoothing algorithm is used to automatically adjust the smoothing window size according to the data fluctuation frequency to balance data smoothness and feature retention. The smoothed intermediate data stream obtained after processing is used as the second-order regularization value. Based on the second-order regularization value, a time-series feature synchronization calibration engine is used to unify the asynchronous timestamps of multiple devices to a standard time axis using a timestamp calibration algorithm, and an attention mechanism time-series interpolation model is used to complete the missing data of multi-dimensional fields, resulting in a regularized steady-state data stream that is time-continuous and feature-complete.

[0013] Specifically, the process of obtaining the network quality feature vector includes: using a fluctuation response sliding window mechanism, taking the index mutation threshold in the steady-state data stream as the trigger condition and dynamically adjusting the window, extracting stable energy features through the moving average energy value and energy decay rate, and obtaining a dual-state energy feature set of sudden energy parameters and stable energy baseline; A multi-source anchoring trend inference model is constructed. Based on the spatiotemporal labels of the steady-state data stream, three types of benchmark data are selected from the associated dataset and linear fitting of the three anchor points is used to obtain the trend feature set of trend direction and fluctuation amplitude. Using a feature selection and fusion algorithm, the mutual information value between the energy feature set and the trend feature set is calculated, and features with strong correlation in mutual information are retained. The selected features are sorted according to the two-dimensional matrix of energy intensity and trend stability, and feature weights are assigned to generate a network quality feature vector with the bi-state energy parameter, ternary anchored trend parameter and spatiotemporal label feature.

[0014] Specifically, the process of obtaining the estimated value of the network quality amplitude includes taking the dual-state energy parameter, ternary anchoring trend parameter and spatiotemporal label feature in the network quality feature vector as input, using the spatiotemporal label to retrieve samples of the same period and scene in the historical database, calculating the contribution ratio of the two types of parameters in the event, and allocating dynamic weights according to the ratio; and obtaining the calibrated estimated value of the network quality amplitude through weighted summation and residual correction.

[0015] Specifically, the process of obtaining abnormal information about network quality includes, firstly, establishing a two-dimensional dynamic mapping system between abnormal amplitude quantization values ​​and scene adaptation thresholds: using spatiotemporal label features to retrieve historical data from the same period and scene, and generating dynamic thresholds for the corresponding scene through a quantile dynamic calibration algorithm; A three-dimensional association rule is constructed for abnormal amplitude, impact range and indicator synergy. Based on the quantified abnormal amplitude value, combined with the number of affected network nodes and coverage scale associated with spatiotemporal labels, as well as the multi-indicator synergistic anomaly reflected by bi-state energy parameters and ternary anchoring trend parameters, a mapping score is obtained through weighted matrix operation to obtain abnormal information of network quality.

[0016] Specifically, the process of acquiring the early warning information package includes: based on the spatiotemporal label features of the abnormal early warning information, retrieving historical abnormal data of the same period and scenario from the network quality prediction association dataset; extracting the correspondence between historical abnormalities, amplitude change curves and duration; coordinating abnormal indicators and occurrence frequency statistics; associating the dual-state energy and ternary anchoring trend parameters of the steady-state data stream; calculating the similarity between current and historical abnormal features; combining the real-time trend slope change rate; and using a scenario-adaptive piecewise linear interpolation algorithm to obtain the minute-level abnormal duration; and based on the historical coordinating abnormal indicator probability and the current parameter coordinating fluctuation, using the early warning information and the multi-dimensional data bound to four levels and three categories to acquire the structured early warning information package.

[0017] Specifically, the categorized information push process is as follows: based on the aforementioned linkage business control, and combined with the network quality area data bound to the warning information, a multi-dimensional identity graph is dynamically generated for each user; the network quality monitoring system will atomically decompose the warning content according to the bound network quality area information, filter out the information modules that match the permissions of different areas, encapsulate them at the role level according to the user's multi-dimensional identity graph, and push the network quality warning information.

[0018] Specifically, the process of obtaining network quality monitoring data for the corresponding test location includes: the management system front-end interface pre-configures test location information through the location management function, presents the speed test interface with a scenario-based layout design, displays the location information of the current test location, provides a drop-down option for location selection to associate with the test location pre-configured by the management system, and supports the location photo upload function to supplement on-site information, thereby obtaining the binding between the test location and the network quality monitoring data.

[0019] Specifically, the construction of the network quality visualization system includes using a real-time situational monitoring dashboard for regional positioning, which can be selected down to the third-level administrative region. After filtering, the network quality data of the target region is displayed, and the data statistics information is updated synchronously to the target region data. Data interaction supports preset filtering conditions, and a multi-dimensional network quality visualization system is constructed.

[0020] The beneficial effects of this invention are as follows: By deeply associating preset data test points with three levels of administrative regions and two levels of scene types, combined with the standardized configuration of the management system's location management function, and the linkage design of location selection and location photo upload in the APP speed test interface, the accuracy of binding collected data with physical scenes is improved. For key scenes such as transportation hubs and universities, regional quality bottlenecks can be quickly located, avoiding blind spots in problem investigation caused by ambiguous location associations, and shortening fault location time from hours to minutes. Through field-level mapping and alignment technology, the field consistency rate of scattered data such as special tests, crowdsourcing tests, and complaints is improved, and the correlation between complaint data problem descriptions and measured indicators is established, transforming multi-source data from isolated storage into value superposition, providing complete data support for end-to-end quality assessment.

[0021] Based on dynamic mapping of abnormal information, scenario-adaptive, piecewise linear interpolation, prediction of duration and hierarchical permission push, and precise binding of early warning information with the four-level three-category management system, the timeliness of front-line maintenance personnel receiving targeted early warnings is improved, the duration of abnormalities is shortened, and the early warning effect feedback mechanism can dynamically optimize monitoring strategies, forming a closed loop of monitoring, early warning, handling, and optimization.

[0022] The interactive view supports drilling down from the provincial level to district / county test points, combining precise location filtering across three levels of administrative regions. Data display and statistical information for the target area are updated synchronously, enabling end-to-end tracing of overall trends, localized issues, and individual data points without switching between multiple systems. Operational steps are simplified, while tracing efficiency is improved. The APP speed test interface and the pre-configured location management information in the management system are linked in real time, automatically associating test data with standard test points, avoiding identification errors caused by manual location input. Data synchronization latency is reduced, data collection utilization is improved, and data resource waste is minimized. Attached Figure Description

[0023] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0024] Figure 1 This is a flowchart illustrating the mobile communication network quality monitoring system of the present invention. Figure 2 This is a flowchart of the data processing for network quality detection in this invention. Detailed Implementation

[0025] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided.

[0026] Please see Figure 1-2 A mobile communication network quality monitoring system includes: a multi-source scenario-bound acquisition module, a data storage and processing module, a linkage service management module, and a dual-linkage situation visualization module; The multi-source scene-bound acquisition module constructs a three-dimensional data acquisition system through preset data test points to obtain network quality data sampled from multiple source scenes; The data storage and processing module performs field-level mapping and alignment on the associated database based on the network quality data, normalizes and stores multi-dimensional fields according to spatiotemporal labels, and constructs a network quality prediction associated dataset. Based on the associated dataset, the data is normalized in multiple orders to generate a normalized steady-state data stream. The network quality feature vector is obtained by using sliding window energy extraction and time-series trend extrapolation. The estimated value of the network quality amplitude is obtained by constructing a multi-indicator collaborative network quality amplitude estimation model. The linkage-based business management module obtains network quality monitoring information based on the estimated value of the network quality amplitude through a dynamic mapping mechanism; it manages and binds the network quality monitoring information with multi-dimensional data, predicts the duration of abnormalities and associated abnormal indicators, and obtains early warning information packages; based on the early warning information packages, it pushes classified information through a hierarchical permission display mechanism. The dual-linkage situation visualization module obtains network quality monitoring data for the corresponding test locations through the scenario-based layout design of the management system's front-end interface APP; by utilizing real-time situation monitoring and multi-dimensional data interaction, it constructs a network quality visualization system with dual-view linkage, obtaining an intuitive presentation and structured traceability of the data.

[0027] As a preferred technical solution of the present invention, the construction of a three-dimensional data acquisition system through preset data test points includes collecting multiple indicators of multiple preset data test points, associating them with administrative regions and scene types, and requiring the collection of on-site environmental data in special test scenarios. The deep binding of indicator data, physical scenes and terminal devices constructs a three-dimensional data acquisition system.

[0028] Specifically, the process of performing field-level mapping and alignment on the associated database includes converting network quality data fields and target fields in the associated database into high-dimensional semantic vectors through a pre-trained network data semantic encoder, and obtaining automatic mapping across naming systems based on vector cosine similarity. The field-level alignment adopts an adaptive time granularity alignment algorithm, which dynamically adjusts the time window according to the data density, corrects asynchronous timestamps through linear interpolation, and associates multi-source field data at the same time node; based on the network topology digital twin model, the multi-source field data is bound to the digital twin of the physical network node, and spatial position alignment across devices is obtained through spatial coordinate encoding.

[0029] Specifically, the construction process of the network quality prediction associated dataset includes: assigning time granularity labels and spatial topology labels to multidimensional fields through a spatiotemporal label dynamic binding mechanism to anchor the fields to the network physical scenario; and using the spatiotemporal labels and field features to construct a two-layer index through an adaptive storage architecture for the associated dataset to obtain the associated dataset for network quality monitoring that enables fast source tracing and multi-dimensional associated queries.

[0030] Based on the network coverage outlined in the document, a network topology digital twin model is constructed. This model is divided into multiple levels: provincial core network, municipal aggregation nodes, district / county base stations, and terminal test points. Each level's nodes are assigned a spatial topology code containing the level, region code, and node ID. Time-aligned multi-source data is then bound to the corresponding node's digital twin. Subsequently, a dynamic spatiotemporal label binding mechanism assigns time-granularity labels matching the data type and spatial topology labels to the corresponding nodes to the multi-source data. Utilizing an adaptive storage architecture for the associated dataset, a two-layer index is constructed, using spatiotemporal labels as the primary index and data field features as the secondary index. This index enables rapid data tracing and multi-dimensional relational queries, forming a network quality prediction associated dataset.

[0031] Specifically, the multi-order regularization method involves using the associated dataset of network quality monitoring as initial input data, and using the anomaly-free data obtained after processing by the spatiotemporal outlier detection model as the first-order regularization value. Based on the first-order regularization value, a multi-scale dynamic smoothing algorithm is used to automatically adjust the smoothing window size according to the data fluctuation frequency to balance data smoothness and feature retention. The smoothed intermediate data stream obtained after processing is used as the second-order regularization value. Based on the second-order regularization value, a time-series feature synchronization calibration engine is used to unify the asynchronous timestamps of multiple devices to a standard time axis using a timestamp calibration algorithm, and an attention mechanism time-series interpolation model is used to complete the missing data of multi-dimensional fields, resulting in a regularized steady-state data stream that is time-continuous and feature-complete.

[0032] In this embodiment, the associated dataset of network quality monitoring in the document (including special test data and crowdsourced testing data) is used as the initial input. It is first processed by the spatiotemporal outlier detection model: a density-based clustering algorithm is used, with the node range corresponding to the spatial topology label as the spatial domain and the time window corresponding to the temporal granularity label as the time domain, to calculate the local density of data points in the spatiotemporal domain. The formula is as follows: , Where, d ij Let i be the spatiotemporal distance between data points i and j. Using the neighborhood radius, outlier data with local density below a threshold are removed to obtain the first-order normalized value for anomaly removal. Based on the first-order normalized value, a multi-scale dynamic smoothing algorithm is employed: first, the data fluctuation frequency is calculated using the following formula: , Where T is the length of the time series, x t The data value at time t is used as the basis for dynamic adjustment of the smoothing window size based on the fluctuation frequency. The smoothing value is calculated by moving average to balance the smoothness of the data and the feature retention, and the smoothed intermediate data stream is obtained as the second-order normalized value.

[0033] Based on the second-order warp value, the timing feature synchronization calibration engine processes the data: First, using a timestamp calibration algorithm, with the system's default standard timeline in the document as the benchmark, the deviation between the timestamps of each device and the standard timeline is calculated. The formula is as follows: =t dev -t std , t dev For the device's original timestamp, t std To standardize the timestamps, all data timestamps are uniformly corrected to standard time (formula: t). cal =t dev -Δt); and then use the attention mechanism temporal interpolation model to complete the missing data in the multi-dimensional fields: calculate the attention weight between the missing data point and the surrounding valid data points, as shown in the following formula: , sim(x i x j Let be the feature similarity between data points i and j. By weighted summation to complete the missing values, a regular steady-state data stream with continuous time and complete features can be obtained.

[0034] Specifically, the process of obtaining the network quality feature vector includes: using a fluctuation response sliding window mechanism, taking the index mutation threshold in the steady-state data stream as the trigger condition and dynamically adjusting the window, extracting stable energy features through the moving average energy value and energy decay rate, and obtaining a dual-state energy feature set of sudden energy parameters and stable energy baseline; A multi-source anchoring trend inference model is constructed. Based on the spatiotemporal labels of the steady-state data stream, three types of benchmark data are selected from the associated dataset and linear fitting of the three anchor points is used to obtain the trend feature set of trend direction and fluctuation amplitude. Using a feature selection and fusion algorithm, the mutual information value between the energy feature set and the trend feature set is calculated, and features with strong correlation in mutual information are retained. The selected features are sorted according to the two-dimensional matrix of energy intensity and trend stability, and feature weights are assigned to generate a network quality feature vector with the bi-state energy parameter, ternary anchored trend parameter and spatiotemporal label feature.

[0035] Specifically, the process of obtaining the estimated value of the network quality amplitude includes taking the dual-state energy parameter, ternary anchoring trend parameter and spatiotemporal label feature in the network quality feature vector as input, using the spatiotemporal label to retrieve samples of the same period and scene in the historical database, calculating the contribution ratio of the two types of parameters in the event, and allocating dynamic weights according to the ratio; and obtaining the calibrated estimated value of the network quality amplitude through weighted summation and residual correction.

[0036] In this embodiment, the multi-index collaborative network quality amplitude estimation model is a multi-source parameter fusion model based on spatiotemporal label matching, dynamic weight allocation, and residual correction. Its core is to combine the bi-state energy parameter and ternary anchoring trend parameter in the network quality feature vector with historical sample data corresponding to the spatiotemporal labels to achieve accurate estimation of the network quality amplitude. Samples from the same period and scenario are retrieved from the historical database using spatiotemporal labels, and the bi-state energy parameter (denoted as E, including the burst energy parameter E) is calculated. b Stable energy baseline E s ) and the three-element anchoring trend parameter (denoted as T, including the trend direction parameter T) d Fluctuation amplitude parameter T v The contribution percentage of the network quality amplitude events in history is then used to assign dynamic weights according to the contribution percentage, and finally the estimated value is obtained through weighted summation and residual correction.

[0037] The key formulas of the model are as follows: Dynamic weight calculation: Let ω be the contribution ratio of the two-state energy parameter in the historical samples. E The contribution ratio of the ternary anchoring trend parameter is ω. T , satisfying ω E +ω t =1, where, , ω T =1−ω E cov() represents covariance, A hist Historical network quality amplitude; The initial amplitude estimation formula is as follows: , Where 'a' represents the weight of the burst energy parameter in the two-state energy, and 'b' represents the weight of the trend direction parameter in the ternary anchored trend; residual correction: = init +ϵ,ϵ=A hist - init,hist , init,hist The initial estimate is obtained from the historical sample. The initial estimate is then calibrated using the residuals ε, resulting in the final value. This is an estimate of the network quality amplitude.

[0038] Specifically, the process of obtaining abnormal information about network quality is as follows: establish a two-dimensional dynamic mapping system between abnormal amplitude quantization value and scene adaptation threshold: retrieve historical data of the same period and scene using spatiotemporal label features, and generate dynamic thresholds for the corresponding scene through quantile dynamic calibration algorithm. A three-dimensional association rule is constructed for abnormal amplitude, impact range and indicator synergy. Based on the quantified abnormal amplitude value, combined with the number of affected network nodes and coverage scale associated with spatiotemporal labels, as well as the multi-indicator synergistic anomaly reflected by bi-state energy parameters and ternary anchoring trend parameters, a mapping score is obtained through weighted matrix operation to obtain abnormal information of network quality.

[0039] Specifically, the process of acquiring the early warning information package includes: based on the spatiotemporal label features of the abnormal early warning information, retrieving historical abnormal data of the same period and scenario from the network quality prediction association dataset; extracting the correspondence between historical abnormalities, amplitude change curves and duration; coordinating abnormal indicators and occurrence frequency statistics; associating the dual-state energy and ternary anchoring trend parameters of the steady-state data stream; calculating the similarity between current and historical abnormal features; combining the real-time trend slope change rate; and using a scenario-adaptive piecewise linear interpolation algorithm to obtain the minute-level abnormal duration; and based on the historical probability of coordinated abnormal indicators and the current parameter coordinated fluctuation, using the early warning information and management-bound multi-dimensional data to acquire the structured early warning information package.

[0040] Specifically, the categorized information push process involves dynamically generating a multi-dimensional identity graph for each user based on the linked business control and the network quality area data bound to the warning information. The network quality monitoring system will atomically decompose the warning content according to the bound network quality area information, filter out information modules that match the permissions of different areas, encapsulate them at the role level according to the user's multi-dimensional identity graph, and push network quality warning information.

[0041] In this embodiment, network quality testing is conducted using the Caiyun Network Testing System: First, a test task is created based on the system's backend specialized test management module. The system selects the government affairs center scenario type and the 5G network type, sets test indicators including download speed and signal strength, and assigns testers. Testers then initiate speed tests in this scenario via an app. The system collects specialized test data and crowdsourced testing data to form a correlated dataset. After obtaining a steady-state data stream through multi-order regularization, a network quality feature vector is generated using sliding window energy extraction and time-series trend extrapolation. This vector is then input into a multi-indicator collaborative amplitude estimation model to obtain the quality amplitude, which is then obtained through a dynamic mapping mechanism. The system retrieves historical data from the same period in the government service center to generate dynamic thresholds. It then calculates mapping scores based on the number of affected nodes and the collaborative anomalies of multiple indicators to identify abnormal information such as a sudden drop in signal strength. Subsequently, it correlates historical abnormal data, calculates feature similarity and real-time trend slope, and uses a scenario-adaptive piecewise linear interpolation algorithm to obtain the duration of the anomaly. Combined with the management system, it pushes complete early warning information containing the abnormal amplitude, duration, and coverage of the 5G signal in the government service center to provincial industry management users, and pushes information modules that only involve the responsibility for anomaly repair in the region to operator district and county users, thereby realizing network quality monitoring and hierarchical early warning.

[0042] Specifically, the process of obtaining network quality monitoring data for the corresponding test location includes: the management system front-end interface pre-configures test location information through the location management function, presents the speed test interface with a scenario-based layout design, displays the location information of the current test location, provides a drop-down option for location selection to associate with the test location pre-configured by the management system, and supports the location photo upload function to supplement on-site information, thereby obtaining the binding between the test location and the network quality monitoring data.

[0043] Specifically, the construction of the network quality visualization system includes using a real-time situational monitoring dashboard for regional positioning, selecting corresponding administrative regions, filtering and displaying network quality data within the target region, and synchronously updating the data statistics to the target region data; data interaction supports preset filtering conditions, and constructs a network quality visualization system from multiple dimensions.

[0044] In this embodiment, when staff need to obtain network quality monitoring data for a commercial building on Beijing Road in Guandu District, Kunming City, Yunnan Province, they first add the test location information for the commercial building through the location management function module of the system's backend management system: Enter the location name: Beijing Road, commercial building, detailed address: Beijing Road, Guandu District, Kunming City, select the scenario type as commercial building and hotel, Grade A office building, fill in the latitude and longitude, and save to complete the pre-configuration of the test location; then the tester opens the system APP... The speed test interface on the homepage adopts a scenario-based layout design, automatically locating and displaying the approximate location information of the current test location, such as Beijing Road, Guandu District, Kunming City. The interface also provides a drop-down menu for location selection, allowing testers to choose from pre-configured locations such as Beijing Road or office buildings in the management system, thus linking the APP with the management system's test locations. If additional details about the on-site environment are needed, testers can use the location photo upload function to take photos of the test location within the office building, such as the distribution of indoor signal coverage equipment and the surrounding environment, and upload them. The APP then binds this test location information with real-time collected network quality monitoring data, such as download speed, upload speed, signal strength, and Ping value, and synchronously uploads it to the backend management system. The backend can then view the complete network quality monitoring data bound to the test location in the data management and specialized test data modules by filtering by location name, Beijing Road, or office building.

[0045] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A mobile communication network quality monitoring system, characterized in that, include: Multi-source scenario-binding acquisition module, data storage and processing module, linkage business management and control module, and dual-linkage situation visualization module; The multi-source scene-bound acquisition module constructs a three-dimensional data acquisition system through preset data test points to obtain network quality data sampled from multiple source scenes; The data storage and processing module performs field-level mapping and alignment on the associated database based on the network quality data, normalizes and stores multidimensional fields according to spatiotemporal labels, and constructs a network quality prediction associated dataset. Based on the aforementioned associated dataset, the data is normalized in multiple orders to generate a normalized steady-state data stream. The network quality feature vector is obtained by using sliding window energy extraction and temporal trend extrapolation. By constructing a multi-index collaborative network quality amplitude estimation model, the estimated value of network quality amplitude is obtained; The linkage-based business management module obtains network quality monitoring information based on the estimated value of the network quality amplitude through a dynamic mapping mechanism; it manages and binds the network quality monitoring information with multi-dimensional data, predicts the duration of abnormalities and associated abnormal indicators, and obtains early warning information packages; based on the early warning information packages, it pushes classified information through a hierarchical permission display mechanism. The dual-linkage situation visualization module obtains network quality monitoring data for the corresponding test locations through the scenario-based layout design of the management system's front-end interface. By leveraging real-time situational monitoring and multi-dimensional data interaction, a network quality visualization system with dual-view linkage is constructed, providing an intuitive presentation of data and structured traceability.

2. The system according to claim 1, characterized in that, The construction of a three-dimensional data acquisition system through preset data test points includes collecting multiple indicators from multiple preset data test points, associating them with administrative regions and scenario types, and requiring the collection of on-site environmental data in specific test scenarios. This involves the deep binding of indicator data, physical scenarios, and terminal devices to construct a three-dimensional data acquisition system.

3. The system according to claim 1, characterized in that, The specific process of performing field-level mapping and alignment of the associated database includes converting network quality data fields and target fields of the associated database into high-dimensional semantic vectors through a pre-trained network data semantic encoder, and obtaining automatic mapping across naming systems based on vector cosine similarity. The field-level alignment adopts an adaptive time granularity alignment algorithm, which dynamically adjusts the time window according to the data density, corrects asynchronous timestamps through linear interpolation, and associates multi-source field data at the same time node; based on the network topology digital twin model, the multi-source field data is bound to the digital twin of the physical network node, and spatial position alignment across devices is obtained through spatial coordinate encoding.

4. The system according to claim 1, characterized in that, The construction process of the network quality prediction association dataset includes assigning time granularity labels and spatial topology labels to multidimensional fields through a spatiotemporal label dynamic binding mechanism, and anchoring the fields to the network physical scenario. By using an adaptive storage architecture for associated datasets, and leveraging the spatiotemporal labels and field features to construct a two-layer index, we can obtain associated datasets for network quality monitoring that enable rapid tracing and multi-dimensional related queries.

5. The system according to claim 1, characterized in that, The multi-order warping method involves using the associated dataset of network quality monitoring as initial input data, and using the anomaly-removed data obtained after processing by the spatiotemporal outlier detection model as the first-order warping value. Based on the first-order warping value, a multi-scale dynamic smoothing algorithm is used to automatically adjust the smoothing window size according to the data fluctuation frequency to balance data smoothness and feature retention. The smoothed intermediate data stream obtained after processing is used as the second-order warping value. Based on the second-order warping value, a time-series feature synchronization calibration engine is used to unify the asynchronous timestamps of multiple devices to a standard time axis using a timestamp calibration algorithm, and an attention mechanism time-series interpolation model is used to complete the missing data in multi-dimensional fields, resulting in a time-continuous and feature-complete warped steady-state data stream.

6. The system according to claim 1, characterized in that, The process of obtaining the network quality feature vector includes: using a fluctuation response sliding window mechanism, taking the index mutation threshold in the steady-state data stream as the trigger condition and dynamically adjusting the window, extracting stable energy features through the moving average energy value and energy decay rate, and obtaining a dual-state energy feature set of sudden energy parameters and stable energy baseline; A multi-source anchoring trend inference model is constructed. Based on the spatiotemporal labels of the steady-state data stream, three types of benchmark data are selected from the associated dataset and linear fitting of the three anchor points is used to obtain the trend feature set of trend direction and fluctuation amplitude. Using a feature filtering and fusion algorithm, the mutual information value between the energy feature set and the trend feature set is calculated, and features with strong correlation in mutual information are retained. The filtered features are sorted according to a two-dimensional matrix of energy intensity and trend stability, and weights are assigned to the features to generate a network quality feature vector with the dual-state energy parameters, ternary anchoring trend parameters, and spatiotemporal label features.

7. The system according to claim 1, characterized in that, The process of obtaining the estimated value of the network quality amplitude includes taking the bi-state energy parameter, the ternary anchoring trend parameter and the spatiotemporal label feature in the network quality feature vector as input, using the spatiotemporal label to retrieve samples of the same period and the same scene in the historical database, calculating the contribution ratio of the two types of parameters in the event, and allocating dynamic weights according to the ratio; and obtaining the calibrated estimated value of the network quality amplitude through weighted summation and residual correction.

8. The system according to claim 1, characterized in that, The process of obtaining abnormal information about network quality is as follows: establish a two-dimensional dynamic mapping system between abnormal amplitude quantization value and scene adaptation threshold: retrieve historical data of the same period and scene using spatiotemporal label features, and generate dynamic thresholds for the corresponding scene through quantile dynamic calibration algorithm. A three-dimensional association rule is constructed for abnormal amplitude, impact range and indicator synergy. Based on the quantified abnormal amplitude value, combined with the number of affected network nodes and coverage scale associated with spatiotemporal labels, as well as the multi-indicator synergistic anomaly reflected by bi-state energy parameters and ternary anchoring trend parameters, a mapping score is obtained through weighted matrix operation to obtain abnormal information of network quality.

9. The system according to claim 1, characterized in that, The process of acquiring the early warning information package includes: based on the spatiotemporal label features of the abnormal early warning information, retrieving historical abnormal data of the same period and scenario from the network quality prediction association dataset, extracting the correspondence between historical abnormalities, amplitude change curves and duration, coordinating abnormal indicators and occurrence frequency statistics, associating the bi-state energy and ternary anchoring trend parameters of the steady-state data stream; calculating the similarity between current and historical abnormal features, combining the real-time trend slope change rate, and using a scenario-adaptive piecewise linear interpolation algorithm to obtain the minute-level abnormal duration period; Based on the historical probability of collaborative anomalies and the current parameter collaborative fluctuations, a structured early warning information package is obtained by using the early warning information and multi-dimensional data bound to management.

10. The system according to claim 1, characterized in that, The classified information push process is as follows: based on the linked business control, and combined with the network quality area data bound to the early warning information, a multi-dimensional identity graph is dynamically generated for each user; the network quality monitoring system will atomically decompose the early warning content according to the bound network quality area information, filter out the information modules that match the permissions of different areas, encapsulate them at the role level according to the user's multi-dimensional identity graph, and push the network quality early warning information.

11. The system according to claim 1, characterized in that, The process of obtaining network quality monitoring data for the corresponding test location includes: the front-end interface of the management system pre-configures test location information through the location management function, presents the speed test interface with a scenario-based layout design, displays the location information of the current test location, provides a drop-down option for location selection to associate with the test location pre-configured by the management system, and supports the location photo upload function to supplement on-site information, thereby obtaining the binding between the test location and the network quality monitoring data.

12. The system according to claim 1, characterized in that, The construction of the network quality visualization system includes: using a real-time situation monitoring dashboard for regional positioning; selecting the corresponding administrative region; filtering and displaying network quality data within the target region; and updating the data statistics information synchronously to the target region data; supporting preset filtering conditions for data interaction; and constructing a network quality visualization system from multiple dimensions.

Images