Problem Area Localization Method and System Applied to Network Quality Assessment

Abstract

This invention provides a method and system for locating problem areas in network quality assessment. First, it acquires a multi-source network data set containing signal strength, transmission rate, packet loss rate, and user complaint information. Then, it constructs a multi-dimensional feature map of network quality based on the spatiotemporal correlation attributes of each data type. Next, it calls a pre-trained network anomaly clustering model to perform anomaly clustering and source tracing processing on the map, generating network anomaly clustering results. Finally, it combines preset network region division rules to perform spatial correlation and location processing on the results, extracting the features of abnormal data clusters in each network region to generate problem area features. Finally, it generates a problem area location instruction based on the problem area features, including network region identifiers, anomaly feature descriptions, and the source of the abnormal data, and sends it to the network optimization terminal. This invention can accurately assess network quality, precisely locate problem areas, provide clear guidance for network optimization, improve network optimization efficiency, and enhance user experience and network performance.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence, and more specifically, to a method and system for locating problem areas in network quality assessment. Background Technology

[0002] In the field of network communication, network quality assessment is crucial for ensuring stable network operation and providing high-quality services. With the continuous expansion of network scale and the increasing diversification of user needs, accurately assessing network quality and quickly locating problem areas has become an urgent issue. Currently, existing network quality assessment methods often rely on only a single type of data, such as signal strength or transmission rate. This single-source assessment method has significant limitations and cannot comprehensively reflect the actual network condition. For example, relying solely on signal strength may not accurately determine whether packet loss exists during network transmission; focusing only on transmission rate may ignore the actual user experience reflected in user complaints. Moreover, existing methods lack in-depth analysis of the spatiotemporal correlations of multi-source data when locating problem areas, making it difficult to accurately determine the specific location and scope of the problem. This results in a lack of focus and efficiency in network optimization, failing to address network quality issues promptly and effectively, thus impacting user experience and overall network performance. Summary of the Invention

[0003] In view of the aforementioned problems, and in conjunction with the first aspect of the present invention, embodiments of the present invention provide a method for locating problem areas in network quality assessment, the method comprising: Obtain a multi-source network data set required for network quality assessment, wherein the multi-source network data set includes signal strength data, transmission rate data, packet loss rate data, and user complaint information data; Based on the spatiotemporal correlation attributes of each data type in the multi-source network data set, a multi-dimensional feature map of network quality is constructed. The multi-dimensional feature map of network quality reflects the temporal variation and spatial distribution correlation of each data type in different network regions. The pre-trained network anomaly clustering model is invoked to perform anomaly clustering and source tracing processing on the multi-dimensional feature map of network quality, generating network anomaly clustering results. The network anomaly clustering results include network data clusters with similar anomaly features and anomaly feature descriptions corresponding to the network data clusters. Based on the preset network region division rules, the network anomaly clustering results are spatially correlated and located to extract the abnormal data cluster features corresponding to each network region divided within the network coverage area and generate problem region features. Based on the characteristics of the problem area, a problem area location instruction is generated, which includes a network area identifier, an anomaly feature description, and the source of the anomaly data. The problem area location instruction is then sent to the network optimization terminal.

[0004] In another aspect, embodiments of the present invention also provide a problem area localization system applied in network quality assessment, including a processor and a machine-readable storage medium connected to the processor. The machine-readable storage medium is used to store programs, instructions, or code, and the processor is used to execute the programs, instructions, or code in the machine-readable storage medium to implement the above-described method.

[0005] Based on the above, this invention, by acquiring a multi-source network data set including signal strength data, transmission rate data, packet loss rate data, and user complaint information, can collect multi-dimensional information reflecting network quality, overcoming the limitations of a single data source and providing a solid foundation for accurate network quality assessment. A multi-dimensional feature map of network quality is constructed based on the spatiotemporal correlation attributes of various data types, deeply exploring the temporal variation patterns and spatial distribution correlations of data in different network regions, making network quality assessment more scientific and reasonable. Calling a pre-trained network anomaly clustering model for anomaly clustering and source tracing can quickly and accurately identify network data clusters with similar anomaly characteristics and generate detailed anomaly feature descriptions. Combining preset network region division rules for spatial correlation and location processing can accurately extract the features of anomaly data clusters in each network region, generating problem region features. Finally, a problem region location instruction containing network region identifiers, anomaly feature descriptions, and the source of anomaly data is generated and sent to the network optimization terminal, providing clear and specific guidance for network optimization work, effectively improving the targeting and efficiency of network optimization, enabling timely resolution of network quality problems, and enhancing user experience and overall network performance. Attached Figure Description

[0006] Figure 1 This is a schematic diagram of the execution flow of the problem area localization method applied to network quality assessment provided in the embodiments of the present invention.

[0007] Figure 2 This is a schematic diagram of exemplary hardware and software components of a problem area localization system applied in network quality assessment, provided in an embodiment of the present invention. Detailed Implementation

[0008] The present invention will now be described in detail with reference to the accompanying drawings. Figure 1 This is a flowchart illustrating a problem area localization method for network quality assessment provided by an embodiment of the present invention. The following is a detailed description of this problem area localization method for network quality assessment.

[0009] Step S110: Obtain the multi-source network data set required for network quality assessment. The multi-source network data set includes signal strength data, transmission rate data, packet loss rate data, and user complaint information data.

[0010] In this embodiment, this step is used to collect various network-related data required for network quality assessment and problem area location. Specifically, signal strength data can be obtained through signal acquisition devices deployed within the network coverage area, transmission rate data can be collected through traffic monitoring devices at network nodes, packet loss rate data can be collected through packet detection devices along the network transmission path, and user complaint information data can be collected through user feedback platforms or customer service systems. All collected data will be aggregated and stored in a data server, forming a multi-source network data set. In practical applications, the collection of this data must comply with relevant laws and regulations. For information that may involve user privacy, such as personal identifiers in user complaint information, data anonymization technology must be used to remove or replace personally identifiable information to protect user privacy.

[0011] Step S120: Based on the spatiotemporal correlation attributes of each data type in the multi-source network data set, construct a multi-dimensional feature map of network quality.

[0012] In this embodiment, this step aims to integrate the spatiotemporal attributes of multi-source network data to form a feature map that reflects the network quality status. By analyzing and correlating the temporal and spatial attributes of different data types, the scattered data is transformed into a structured map with spatiotemporal dimensions.

[0013] Step S121: Decompose the signal strength data, transmission rate data, packet loss rate data, and user complaint information data in the multi-source network data set according to data dimensions, and extract the time attribute features and spatial attribute features corresponding to the signal strength data, the transmission rate data, the packet loss rate data, and the user complaint information data. The time attribute features include the time series information of data collection, and the spatial attribute features include the network area location information corresponding to the data.

[0014] In this embodiment, a unified urban network coverage area is used as the application scenario. When decomposing signal strength data, the time attribute feature is the specific time point of data collection and the time sequence of continuous collection, and the spatial attribute feature is the network area location of the device collecting the signal strength data; the time attribute feature of transmission rate data is the time interval of data monitoring, and the spatial attribute feature is the network area corresponding to the network node monitoring the transmission rate data; the time attribute feature of packet loss rate data is the time point of packet detection, and the spatial attribute feature is the network area corresponding to the transmission path of the packet loss rate data; the time attribute feature of user complaint information data is the time of submission of the user complaint, and the spatial attribute feature is the network area where the user filed the complaint.

[0015] Step S122: Establish a time attribute association model, perform time-series alignment processing on the time-series information of signal strength data, transmission rate data, packet loss rate data, and user complaint information data, and generate a joint time-series association relationship for multiple data types.

[0016] In this embodiment, in the scenario of urban network coverage area, the model is used to adjust the time information of different data types to the same time dimension.

[0017] Step S1221: Determine the smallest time unit for the four types of time series information: time series information of signal strength data, time series information of transmission rate data, time series information of packet loss rate data, and time series information of user complaint data.

[0018] In this embodiment, within an urban network coverage area scenario, the collection or recording time intervals for four types of time series information are examined, and the smallest time interval is selected as the minimum time unit. For example, if signal strength data is collected every ten minutes, transmission rate data is monitored every five minutes, packet loss rate data is detected every minute, and user complaint information is submitted in real time, then the minimum time unit is one minute.

[0019] Step S1222: Based on the determined minimum time unit, divide the time series information of signal strength data into multiple time slices, divide the time series information of transmission rate data into multiple time slices, divide the time series information of packet loss rate data into multiple time slices, and divide the time series information of user complaint information data into multiple time slices.

[0020] In this embodiment, in the scenario of urban network coverage area, the signal strength data is divided into time intervals of one minute, with one minute as the smallest time unit, and each time interval is a time slice; similarly, the transmission rate data, packet loss rate data, and user complaint information data are also divided into time intervals of one minute to obtain their respective corresponding time slices.

[0021] Step S1223: For each time slice, extract the average value of signal strength data, the average value of transmission rate data, the average value of packet loss rate data, and the number of user complaint information data within that time slice to form a time-related data group for that time slice.

[0022] In this embodiment, in the scenario of urban network coverage area, for each one-minute time slice, the average value of all signal strength data in the time slice is calculated, the average value of all transmission rate data in the time slice is calculated, the average value of all packet loss rate data in the time slice is calculated, and the number of user complaint information data in the time slice is counted. These four data are combined together to form the time-related data group of the time slice.

[0023] Step S1224: Arrange all time-related data groups of time slices in chronological order to construct a time-related sequence.

[0024] In this embodiment, in the scenario of urban network coverage area, the time-related data groups corresponding to each time slice are arranged sequentially according to the order of time slices to form a continuous time-related sequence. This sequence can reflect the comprehensive situation of various network data at different time points.

[0025] Step S1225: Calculate the degree of difference between time-related data groups of adjacent time slices. The degree of difference is determined by calculating the sum of the absolute differences of the corresponding indicators in the two data groups.

[0026] In this embodiment, in the scenario of urban network coverage area, for the time-related data groups corresponding to two adjacent time slices in the time-related sequence, the absolute difference of the average signal strength, the absolute difference of the average transmission rate, the absolute difference of the average packet loss rate, and the absolute difference of the number of user complaint information are calculated respectively. Then, these four absolute differences are added together to obtain the degree of difference between the time-related data groups of the two adjacent time slices.

[0027] Step S1226: Segment the time-related sequence according to the difference value. Continuous time slices with a difference value less than the preset segmentation threshold are divided into the same time period, generating a joint time series association relationship containing multiple time periods and corresponding time-related data.

[0028] In this embodiment, within an urban network coverage area scenario, a segmentation threshold is set, and the difference between adjacent time slices in the time-related sequence is iterated. When a difference is less than the threshold, the corresponding two time slices are divided into the same time period; when the difference is greater than or equal to the threshold, the time period of the previous time slice ends, and the next time slice becomes the start of a new time period. In this way, the time-related sequence is divided into multiple time periods, and the time-related data within each time period has high similarity, thereby generating a joint time series association relationship.

[0029] Step S123: Establish a spatial attribute association model by spatially overlaying the network area location information of signal strength data, transmission rate data, packet loss rate data, and user complaint information data to generate a joint spatial distribution association relationship of multiple data types.

[0030] In this embodiment, within the urban network coverage area scenario, the model integrates spatial location information of different data types into the same spatial coordinate system. Specifically, the locations of signal strength data acquisition devices, transmission rate data monitoring nodes, packet loss rate data detection paths, and the location of user complaints are all mapped onto a spatial map of the urban network coverage area. Through a spatial overlay algorithm, this spatial location information from different sources is fused to form a joint spatial distribution correlation that can display the spatial distribution of various types of data.

[0031] Step S124: Based on the joint time series correlation and the joint spatial distribution correlation, assign corresponding feature weights to the four types of data included in the multi-source network data set: signal strength data, transmission rate data, packet loss rate data, and user complaint information data. The feature weights are determined according to the impact of the data types on network quality assessment.

[0032] In this embodiment, within an urban network coverage area scenario, this step is used to determine the importance of different data types in network quality assessment. By analyzing the impact of various data types on network quality, corresponding weights are assigned to each data type.

[0033] Step S1241: Retrieve a preset list of data type influence factors in the field of network quality assessment. The list of data type influence factors includes the basic influence factors corresponding to signal strength data, the basic influence factors corresponding to transmission rate data, the basic influence factors corresponding to packet loss rate data, and the basic influence factors corresponding to user complaint information data.

[0034] In this embodiment, in the scenario of urban network coverage areas, the list of data type impact factors is formulated based on professional knowledge and historical experience in network quality assessment. For each data type, a corresponding basic impact factor is set according to its fundamental impact on network quality. For example, packet loss rate data has a significant impact on network quality, so its basic impact factor is relatively high; while user complaint information data directly reflects user experience, it is greatly influenced by subjective user factors, so its basic impact factor is relatively low.

[0035] Step S1242: Analyze the time series fluctuation amplitudes of signal strength data, transmission rate data, packet loss rate data, and user complaint information data in the joint time series correlation relationship. Perform numerical correlation calculations between the time series fluctuation amplitude of signal strength data and the basic influence factors corresponding to the signal strength data to obtain the time dimension weight coefficient of signal strength data. Perform numerical correlation calculations between the time series fluctuation amplitude of transmission rate data and the basic influence factors corresponding to the transmission rate data to obtain the time dimension weight coefficient of transmission rate data. Perform numerical correlation calculations between the time series fluctuation amplitude of packet loss rate data and the basic influence factors corresponding to the packet loss rate data to obtain the time dimension weight coefficient of packet loss rate data. Perform numerical correlation calculations between the time series fluctuation amplitude of user complaint information data and the basic influence factors corresponding to the user complaint information data to obtain the time dimension weight coefficient of user complaint information data.

[0036] In this embodiment, within the urban network coverage area scenario, the time series fluctuation amplitude reflects the degree of drastic change in the data over time. For each data type, its fluctuation amplitude in the joint time series correlation is calculated, and then this fluctuation amplitude is calculated with the corresponding basic impact factor. For example, if the time series fluctuation amplitude of a certain type of data is large, it indicates that its changes over time are more significant and have a greater impact on the time dimension of network quality assessment. By calculating with the basic impact factor, the time dimension weight coefficient of this type of data is obtained.

[0037] Step S1243: Analyze the spatial coverage density of signal strength data, transmission rate data, packet loss rate data, and user complaint information data in the joint spatial distribution correlation relationship. Perform numerical correlation calculations between the spatial coverage density of signal strength data and the basic influence factors corresponding to the signal strength data to obtain the spatial dimension weight coefficient of the signal strength data. Perform numerical correlation calculations between the spatial coverage density of transmission rate data and the basic influence factors corresponding to the transmission rate data to obtain the spatial dimension weight coefficient of the transmission rate data. Perform numerical correlation calculations between the spatial coverage density of packet loss rate data and the basic influence factors corresponding to the packet loss rate data to obtain the spatial dimension weight coefficient of the packet loss rate data. Perform numerical correlation calculations between the spatial coverage density of user complaint information data and the basic influence factors corresponding to the user complaint information data to obtain the spatial dimension weight coefficient of the user complaint information data.

[0038] In this embodiment, within the urban network coverage area scenario, spatial coverage density reflects the density of data distribution in the spatial dimension. For each data type, its coverage density in the joint spatial distribution association is calculated, and then this coverage density is calculated with the corresponding basic influence factor. For example, if the spatial coverage density of a certain type of data is high, it indicates that its spatial distribution is relatively wide and has a significant impact on the spatial dimension of network quality assessment. By calculating with the basic influence factor, the spatial dimension weight coefficient of this type of data is obtained.

[0039] Step S1244: The time dimension weight coefficient and the spatial dimension weight coefficient of the signal strength data are weighted and summed to obtain the comprehensive feature weight corresponding to the signal strength data; the time dimension weight coefficient and the spatial dimension weight coefficient of the transmission rate data are weighted and summed to obtain the comprehensive feature weight corresponding to the transmission rate data; the time dimension weight coefficient and the spatial dimension weight coefficient of the packet loss rate data are weighted and summed to obtain the comprehensive feature weight corresponding to the packet loss rate data; and the time dimension weight coefficient and the spatial dimension weight coefficient of the user complaint information data are weighted and summed to obtain the comprehensive feature weight corresponding to the user complaint information data.

[0040] In this embodiment, in the scenario of urban network coverage area, the comprehensive feature weight takes into account the impact of data in both time and space dimensions. By weighted summing of the weight coefficients for the time dimension and the weight coefficients for the space dimension, the comprehensive feature weight for each type of data is obtained. This comprehensive feature weight can more comprehensively reflect the overall impact of data on network quality assessment.

[0041] Step S1245: Normalize the comprehensive feature weights corresponding to signal strength data, transmission rate data, packet loss rate data, and user complaint information data. Adjust the comprehensive feature weight values ​​of each data type by linear scaling so that the sum of the comprehensive feature weights of each data type reaches a fixed value. Finally, determine the feature weights of the four data types (signal strength data, transmission rate data, packet loss rate data, and user complaint information data) in the multi-dimensional feature map of network quality.

[0042] In this embodiment, in the scenario of urban network coverage area, the purpose of normalization is to ensure that the comprehensive feature weights of various data types are on the same order of magnitude. Through a linear scaling algorithm, the comprehensive feature weights of each data type are adjusted to a fixed range, while ensuring that the sum of the comprehensive feature weights of all data types is a fixed value, such as one. In this way, when constructing a multi-dimensional feature map of network quality, the display weight of each data type in the map can be reasonably allocated according to its feature weight.

[0043] Step S125: Integrate the joint time series correlation with feature weights and the joint spatial distribution correlation into a graph to construct a multi-dimensional feature graph of network quality that includes a time dimension layer, a spatial dimension layer and a data feature layer.

[0044] In this embodiment, within the urban network coverage area scenario, this step integrates the processed temporal and spatial relationships, as well as the feature weights of various data types, into a unified graph structure. The temporal dimension layer displays the changes of various data types at different times, the spatial dimension layer displays the spatial distribution of various data types, and the data feature layer displays the feature weights and specific values ​​of various data types. Through graph integration technology, this information is presented in a visual manner, forming a multi-dimensional feature graph of network quality.

[0045] Step S130: Call the pre-trained network anomaly clustering model to perform anomaly clustering and source tracing processing on the multi-dimensional feature map of network quality, and generate network anomaly clustering results. The network anomaly clustering results include network data clusters with similar anomaly features and anomaly feature descriptions corresponding to the network data clusters.

[0046] In this embodiment, within an urban network coverage area scenario, this step utilizes a pre-trained model to analyze the feature map and identify data clusters with anomalous characteristics. A clustering algorithm is then used to group data with similar anomalous characteristics together, forming anomalous data clusters, and the anomalous characteristics of each cluster are described.

[0047] Step S131: Input the network quality multi-dimensional feature map into the map parsing layer of the network anomaly clustering model, and split the network quality multi-dimensional feature map into multiple sub-map units, each sub-map unit corresponding to a preset time interval and a preset spatial sub-region.

[0048] In this embodiment, in the scenario of urban network coverage area, the map parsing layer adopts an image segmentation algorithm to divide the entire network quality multi-dimensional feature map into multiple sub-map units according to the preset time interval and spatial sub-region size. For example, if the time interval is set to a certain hour of a day and the spatial sub-region is set to a certain block in the urban network coverage area, then each sub-map unit corresponds to the network quality characteristics of that block within that hour.

[0049] Step S132: Extract feature vectors for each of the split sub-graph units, and convert the temporal attribute features, spatial attribute features, and data feature weights in the sub-graph units into sub-graph feature vectors of a unified dimension.

[0050] In this embodiment, in the scenario of urban network coverage area, the feature vector extraction process employs a feature extraction algorithm to convert the temporal information, spatial information, and feature weights of various data in the sub-map unit into a set of vectors that can represent the features of the sub-map unit. For example, temporal attribute features can be converted into vector elements representing time intervals, spatial attribute features can be converted into vector elements representing spatial locations, and data feature weights can be converted into vector elements representing the importance of various data types, ultimately forming a sub-map feature vector with a unified dimension.

[0051] Step S133: Input all sub-graph feature vectors into the clustering layer of the network anomaly clustering model, and use density clustering algorithm to cluster all sub-graph feature vectors to initially generate multiple candidate data clusters.

[0052] In this embodiment, in the scenario of urban network coverage area, the clustering layer utilizes a density clustering algorithm to group sub-map feature vectors with similar characteristics together to form candidate data clusters. Specifically, the algorithm calculates the distance between each sub-map feature vector and other vectors. When the distance between multiple vectors is less than a preset threshold, they are classified into the same candidate data cluster. In this way, multiple candidate data clusters are initially generated, and the sub-map feature vectors within each cluster have high similarity.

[0053] Step S134: Perform anomaly feature verification on each initially generated candidate data cluster, extract the anomaly feature index of each sub-map feature vector in the candidate data cluster, and calculate the consistency of the anomaly feature index within the candidate data cluster.

[0054] In this embodiment, in the scenario of urban network coverage area, the anomaly feature verification process is used to determine whether candidate data clusters are truly anomalous data clusters. First, anomaly feature indicators are extracted from the feature vectors of all sub-graphs in each candidate data cluster. These indicators can be the degree of data fluctuation, the degree of deviation from the normal range, etc. Then, the consistency of these anomaly feature indicators within the candidate data cluster is calculated, for example, by calculating the variance or standard deviation of the indicators. The smaller the variance or standard deviation, the higher the consistency.

[0055] Step S1341: Retrieve a preset abnormal feature index library from the network anomaly clustering model. The abnormal feature index library includes abnormal fluctuation indexes corresponding to signal strength data, abnormal attenuation indexes corresponding to transmission rate data, abnormal change indexes corresponding to packet loss rate data, and abnormal frequency indexes corresponding to user complaint information data.

[0056] In this embodiment, in the scenario of urban network coverage areas, the abnormal feature index library is developed based on professional knowledge of network anomaly detection and historical data. For each type of data, the library sets corresponding anomaly judgment indicators. For example, the abnormal fluctuation indicator for signal strength data is that the change in signal strength within a short period exceeds a preset value; the abnormal attenuation indicator for transmission rate data is that the decrease in transmission rate exceeds a preset value; the abnormal change indicator for packet loss rate data is that the increase in packet loss rate exceeds a preset value; and the abnormal frequency indicator for user complaint information data is that the number of user complaints per unit time exceeds a preset value.

[0057] Step S1342: For each candidate data cluster, extract the index values ​​corresponding to the abnormal feature index library from the feature vectors of all sub-maps within the cluster to form the abnormal index set of the candidate data cluster.

[0058] In this embodiment, within an urban network coverage area scenario, for each candidate data cluster, all sub-spectrum feature vectors are traversed. Based on the indicator definitions in the anomaly indicator library, the corresponding indicator values ​​for each sub-spectrum feature vector are extracted. For example, for anomaly fluctuation indicators of signal strength data, the amplitude of signal strength fluctuations in each sub-spectrum feature vector is extracted; for anomaly attenuation indicators of transmission rate data, the amplitude of transmission rate attenuation in each sub-spectrum feature vector is extracted, and so on. These extracted indicator values ​​are then combined to form the anomaly indicator set for that candidate data cluster.

[0059] Step S1343: Calculate the standard deviation of the index values ​​within each set of abnormal indicators. The standard deviation value corresponds to the consistency of the abnormal indicators within the candidate data cluster. When the standard deviation value is in a preset low value range, the consistency of the abnormal indicators within the candidate data cluster meets the requirements.

[0060] In this embodiment, in the scenario of urban network coverage area, standard deviation is used to measure the dispersion of indicator values ​​within an anomaly indicator set. The smaller the standard deviation, the smaller the difference between indicator values ​​and the higher the consistency; the larger the standard deviation, the greater the difference between indicator values ​​and the lower the consistency. By setting a low value range, when the standard deviation of a certain anomaly indicator set is within this range, the consistency of the anomaly indicator within the candidate data cluster is considered to meet the requirements.

[0061] Step S1344: Calculate the average standard deviation of all outlier indicators for each candidate data cluster to obtain the overall standard deviation of the candidate data cluster.

[0062] In this embodiment, in the scenario of urban network coverage area, the overall standard deviation comprehensively considers the consistency of all abnormal indicators within the candidate data cluster. By averaging the standard deviations of each abnormal indicator, an overall standard deviation that reflects the consistency of abnormal indicators across the entire candidate data cluster is obtained. The smaller this overall standard deviation, the higher the consistency of abnormal indicators within the candidate data cluster, and the greater the likelihood that the candidate data cluster is an abnormal data cluster.

[0063] Step S1345: Compare the overall standard deviation with the preset consistency threshold. If the overall standard deviation is less than the consistency threshold, the consistency of the candidate data cluster is determined to meet the preset standard. If the overall standard deviation is greater than or equal to the consistency threshold, the consistency of the candidate data cluster is determined to not meet the preset standard.

[0064] In this embodiment, in the scenario of urban network coverage area, the consistency threshold is set based on experience and requirements for network anomaly detection. When the overall standard deviation of the candidate data cluster is less than the threshold, it indicates that the anomaly indicators within the candidate data cluster have high consistency and meet the characteristics of an anomaly data cluster; when the overall standard deviation is greater than or equal to the threshold, it indicates that the anomaly indicators within the candidate data cluster have low consistency, may contain non-anomaly data, and do not meet the characteristics of an anomaly data cluster.

[0065] Step S135: Retain candidate data clusters that meet the preset consistency standard as abnormal data clusters, add anomaly feature description tags to each abnormal data cluster, the abnormal feature description tags include the data type composition within the cluster, the range of abnormal feature indicators and the spatiotemporal range of data collection, and integrate all abnormal data clusters and corresponding abnormal feature description tags to generate network anomaly clustering results.

[0066] In this embodiment, within an urban network coverage area scenario, candidate data clusters that meet a preset consistency standard are identified as anomalous data clusters. Then, based on the feature vectors of the sub-graphs within each cluster, the data types contained in the anomalous data cluster are analyzed to determine the range of anomalous feature indicators, as well as the temporal and spatial scope of data collection. This information is then compiled into anomalous feature description labels. Finally, all anomalous data clusters and their corresponding anomalous feature description labels are integrated to form a network anomaly clustering result. This result clearly demonstrates the anomalous data present in the network.

[0067] Step S140: Combine the preset network region division rules to perform spatial correlation and localization processing on the network anomaly clustering results, extract the corresponding abnormal data cluster features of each network region divided within the network coverage area, and generate problem region features.

[0068] In this embodiment, within an urban network coverage area scenario, this step associates anomalous data clusters with specific network regions to determine which network regions exhibit anomalies. By analyzing the spatial location information of the anomalous data clusters and combining it with network region segmentation rules, the characteristics of the anomalous data clusters within each network region are extracted.

[0069] Step S141: Retrieve preset network area division rules, which include area division standards based on geographical boundaries and area division standards based on network device coverage. Divide the entire network coverage area into multiple independent network areas according to the network area division rules.

[0070] In this embodiment, in the urban network coverage area scenario, the network area division rules are formulated based on the city's geographical boundaries and the coverage range of network devices. Geographical boundary-based division criteria can divide the city into different administrative districts or blocks, while network device coverage-based division criteria can divide areas according to the coverage range of network devices such as base stations and routers. Based on these rules, the entire urban network coverage area is divided into multiple independent network areas, each with clearly defined boundaries and corresponding network devices.

[0071] Step S142: Establish a spatial mapping relationship between network regions and anomalous data clusters. Match the spatial sub-regions corresponding to each anomalous data cluster in the network anomaly clustering results with the divided network regions to determine the network region to which each anomalous data cluster belongs.

[0072] In this embodiment, in the scenario of urban network coverage area, the spatial mapping relationship is implemented through a spatial matching algorithm. First, the coordinates of the spatial sub-region corresponding to each abnormal data cluster are obtained. Then, these coordinates are compared with the boundary coordinates of the divided network regions. When the spatial sub-region of an abnormal data cluster is completely contained within a certain network region, or when the overlap area between the spatial sub-region of an abnormal data cluster and a certain network region exceeds a preset ratio, the abnormal data cluster is determined to belong to that network region. In this way, the correspondence between network regions and abnormal data clusters is established.

[0073] For example, in step S1421: obtain the spatial boundary coordinates of each network region after division, and generate a spatial boundary range map of each network region based on the obtained spatial boundary coordinates.

[0074] In this embodiment, in the scenario of urban network coverage area, the spatial boundary coordinates can be obtained through a geographic information system. The boundary coordinates of each network area include the coordinates of the four vertices of the area or other coordinate points that can determine the boundary of the area. Based on these coordinate points, a spatial boundary range map of each network area is generated using a graphics generation algorithm. This map can clearly show the boundary and range of the network area.

[0075] Step S1422: Obtain the coordinate range of the spatial sub-region corresponding to each abnormal data cluster in the network abnormal clustering results, and generate a spatial range map of each abnormal data cluster based on the obtained coordinate range.

[0076] In this embodiment, in a scenario with urban network coverage, the spatial sub-region coordinate range corresponding to the anomalous data clusters can be extracted from the network anomaly clustering results. Based on these coordinate ranges, a spatial range map of each anomalous data cluster is generated using a graph generation algorithm. This map can show the spatial distribution range of the anomalous data clusters.

[0077] Step S1423: Overlay and compare the spatial extent map of each anomalous data cluster with the spatial boundary extent map of all network regions, and calculate the overlap area between the spatial extent of the anomalous data cluster and the spatial extent of each network region.

[0078] In this embodiment, in the scenario of urban network coverage area, the overlay comparison process is implemented through a graphical overlay algorithm. The spatial extent map of the abnormal data clusters is overlaid with the spatial boundary extent map of each network region, and then the area of ​​the overlapping part is calculated. In this way, the overlap area between the abnormal data clusters and each network region is obtained.

[0079] Step S1424: Determine the network region with the largest overlap area as the candidate region to which the abnormal data cluster belongs.

[0080] In this embodiment, within an urban network coverage area scenario, the network region with the largest overlap area is identified by comparing the overlap area between the anomalous data cluster and each network region. This region is then designated as a candidate region for the anomalous data cluster. This is because a larger overlap area indicates a higher degree of correlation between the anomalous data cluster and the network region.

[0081] Step S1425: Calculate the ratio of the overlap area between the spatial range of the abnormal data cluster and the spatial range of the candidate region to the total spatial range of the abnormal data cluster. If the ratio of the overlap area to the total spatial range of the abnormal data cluster reaches the preset attribution threshold, the network region to which the abnormal data cluster belongs is determined as the candidate region. If the ratio of the overlap area to the total spatial range of the abnormal data cluster does not reach the preset attribution threshold, the accuracy of the spatial boundary coordinates is rechecked, the spatial boundary coordinates are corrected, and the overlap area and ratio are recalculated until the network region to which the abnormal data cluster belongs is determined.

[0082] In this embodiment, in the urban network coverage area scenario, the attribution threshold is set based on the accuracy of network area division and the distribution of abnormal data clusters. When the proportion of the overlapping area to the total spatial range of the abnormal data clusters reaches this threshold, it indicates that the abnormal data clusters are mainly distributed within the candidate attribution area, and therefore the candidate attribution area is determined as the network area to which the abnormal data clusters belong. If the proportion does not reach the threshold, it indicates that there may be inaccurate spatial boundary coordinates, and the coordinates need to be rechecked and corrected. Then, the overlapping area and proportion are recalculated until the network area to which the abnormal data clusters belong is determined.

[0083] Step S143: Perform feature integration on all abnormal data clusters in each divided network region, extract the abnormal feature description labels of each abnormal data cluster, and count the number of abnormal data clusters and the frequency of occurrence of various abnormal features in the network region.

[0084] In this embodiment, in a city network coverage area scenario, the feature integration process first collects the anomaly feature description tags of all abnormal data clusters within each network area. Then, these tags are analyzed to extract various anomaly features. Next, the total number of abnormal data clusters within the network area and the frequency of each anomaly feature are counted. In this way, the quantity and type distribution of abnormal data within each network area can be understood.

[0085] Step S144: Calculate the anomaly impact of each network region. The anomaly impact is calculated based on the number of abnormal data clusters in the network region, the frequency of occurrence of various anomaly features, and the feature weights corresponding to the abnormal data clusters.

[0086] In this embodiment, within an urban network coverage area scenario, this step is used to assess the impact of anomalies on network quality within each network area. By comprehensively considering the number of anomalous data clusters, the frequency of occurrence of anomalous features, and the feature weights of anomalous data clusters, the impact of anomalies on each network area is calculated to determine which network areas require focused attention.

[0087] Step S1441: Assign an influence coefficient to each anomalous data cluster within the network area. The influence coefficient is determined based on the feature weights corresponding to the anomalous data cluster. The feature weights corresponding to the anomalous data cluster correspond to the influence coefficients. When the feature weights corresponding to the anomalous data cluster change, the influence coefficients change accordingly.

[0088] In this embodiment, within the urban network coverage area scenario, the influence coefficient reflects the contribution of anomalous data clusters to the anomaly impact on the network area. Anomalous data clusters with higher feature weights have larger influence coefficients. Specifically, based on the feature weights of the anomalous data clusters, a mapping algorithm converts the feature weights into corresponding influence coefficients. For example, an anomalous data cluster with a feature weight of 0.5 may have an influence coefficient of 0.5; an anomalous data cluster with a feature weight of 0.8 may have an influence coefficient of 0.8. When the feature weights of the anomalous data clusters change, the influence coefficients will also change accordingly.

[0089] Step S1442: Calculate the sum of the influence coefficients of all anomalous data clusters within the network area to obtain the basic influence value of the network area.

[0090] In this embodiment, within an urban network coverage area scenario, the basic impact value reflects the overall impact of all anomalous data clusters within the network area on network quality. The basic impact value for the network area is obtained by summing the impact coefficients of each anomalous data cluster. A larger basic impact value indicates a greater impact of anomalous data clusters on network quality within that area.

[0091] Step S1443: Count the occurrence frequency of various abnormal features within the network area, and perform weighted processing on the occurrence frequency of each abnormal feature. The occurrence frequency value of the abnormal feature corresponds to the corresponding weighting coefficient value. When the occurrence frequency value of the abnormal feature changes, the weighting coefficient value changes accordingly. Calculate the sum of the weighted frequencies of all abnormal features to obtain the frequency influence value.

[0092] In this embodiment, within an urban network coverage area scenario, the frequency impact value reflects the influence of the frequency of occurrence of various anomalies within the network area on network quality. First, the frequency of occurrence of each anomaly within the network area is statistically analyzed. Then, based on the severity of the anomaly, a corresponding weighting coefficient is assigned to each anomaly; the higher the severity of the anomaly, the larger the weighting coefficient. The weighted frequency of each anomaly is obtained by multiplying its frequency by its corresponding weighting coefficient. Finally, the weighted frequencies of all anomalies are summed to obtain the frequency impact value for the network area.

[0093] Step S1444: The basic impact value and the frequency impact value are weighted and summed according to a preset ratio to obtain the initial abnormal impact of the network area.

[0094] In this embodiment, in the scenario of urban network coverage areas, the initial anomaly impact comprehensively considers the influence of the number of anomalous data clusters and the frequency of occurrence of anomalous features on network quality. The preset proportions are set according to the needs of network quality assessment; for example, the basic impact value accounts for 0.6%, and the frequency impact value accounts for 0.4%. The initial anomaly impact of the network area is obtained by multiplying the basic impact value by its corresponding proportion and the frequency impact value by its corresponding proportion, and then adding the two together.

[0095] Step S1445: Sort the initial anomaly impact of all network regions, classify the initial anomaly impact according to the sorting results, add a level label to the anomaly impact of each network region, and finally determine the anomaly impact of each network region.

[0096] In this embodiment, within an urban network coverage area scenario, the initial anomaly impact of all network areas is first sorted from highest to lowest. Then, based on the sorting results, the initial anomaly impact is categorized into different levels, such as severe, moderate, and serious. A corresponding threshold is set for each level; when the initial anomaly impact of a network area exceeds a certain threshold, it is classified into the corresponding level, and a level label is added. This method clearly displays the degree of anomaly impact for each network area.

[0097] Step S145: Generate a feature description for each network region that includes the composition of abnormal data clusters, the type of abnormal features, the impact of abnormalities, and the spatiotemporal range of data collection. Use the feature description of the composition of abnormal data clusters, the type of abnormal features, the degree of abnormal impact, and the spatiotemporal range of data collection as the feature of the problem region.

[0098] In this embodiment, within an urban network coverage area scenario, the composition of abnormal data clusters is analyzed based on the abnormal data clusters within each network area. The types of abnormal features are determined, the impact of the anomalies is assessed, and the temporal and spatial scope of data collection is defined. This information is then compiled into a feature description, which comprehensively reflects the anomalies within the network area and serves as a feature of the problem area.

[0099] Step S150: Based on the characteristics of the problem area, generate a problem area location instruction that includes a network area identifier, an anomaly feature description, and the source of the anomaly data, and send the problem area location instruction to the network optimization terminal.

[0100] In this embodiment, in a scenario with urban network coverage, this step generates location instructions that can guide network optimization work based on the characteristics of the problem area. By integrating information such as network area identifiers, descriptions of abnormal characteristics, and sources of abnormal data into the instructions, these instructions are sent to the network optimization terminal so that network optimization personnel can promptly understand the problems existing in the network and take corresponding optimization measures.

[0101] Step S151: Analyze the features of the problem area and extract the network area identifier, anomaly feature type, anomaly impact level, anomaly data cluster composition, and spatiotemporal range of anomaly data collection for each network area.

[0102] In this embodiment, in a scenario with urban network coverage, the parsing process first reads the information from the problem area features, then extracts the identifier of each network area, such as area number or name; the type of abnormal feature, such as abnormal signal strength or abnormal transmission rate; the level of abnormal impact, such as severe or relatively severe; the composition of abnormal data clusters, such as abnormal data clusters containing signal strength data and packet loss rate data; and the time and spatial range of abnormal data collection. In this way, the information in the problem area features is decomposed into specific fields.

[0103] Step S152: Construct a positioning instruction framework according to the instruction format recognizable by the network optimization terminal. The positioning instruction framework includes an instruction header, a network area information segment, an abnormal feature information segment, an abnormal data source segment, and an instruction tail.

[0104] In this embodiment, in the scenario of urban network coverage areas, the instruction format is formulated according to the interface requirements and communication protocol of the network optimization terminal. The instruction header contains information such as the instruction type and version; the network area information segment is used to fill in the network area identifier; the anomaly feature information segment is used to fill in the anomaly feature type and the level of anomaly impact; the anomaly data source segment is used to fill in the composition of the anomaly data cluster and the spatiotemporal range of the anomaly data collection; and the instruction tail contains verification information, etc. Based on these requirements, a positioning instruction framework is constructed, which provides the structural foundation for generating specific problem area positioning instructions.

[0105] Step S153: Fill the extracted network area identifier into the network area information segment of the positioning command frame, fill the abnormal feature type and abnormal impact level into the abnormal feature information segment of the positioning command frame, and fill the abnormal data cluster composition and the spatiotemporal range of abnormal data collection into the abnormal data source segment of the positioning command frame.

[0106] In this embodiment, within an urban network coverage area scenario, based on the parsed problem area feature information, the network area identifier is accurately filled into the network area information segment of the positioning command framework; the anomaly feature type and anomaly impact level are filled into the anomaly feature information segment; and the anomaly data cluster composition and the spatiotemporal range of anomaly data collection are filled into the anomaly data source segment. This ensures the accuracy of information in each field so that the network optimization terminal can correctly parse the command content.

[0107] Step S154: Encode the filled positioning instruction frame by converting the text information in the positioning instruction frame into binary code that conforms to the network transmission protocol, and generate the problem area positioning instruction.

[0108] In this embodiment, in a city network coverage area scenario, the instruction encoding process employs an encoding algorithm to convert the text information in the filled location instruction framework into binary data. Specifically, each character is converted into its corresponding binary code, and the binary data is encapsulated and formatted according to network transmission protocol requirements to generate a problem area location instruction that can be transmitted over the network. The encoded instruction has high transmission efficiency and security, ensuring that the instruction is not tampered with or lost during transmission.

[0109] Step S155: Establish a two-way communication link with the network optimization terminal, send a connection request through the two-way communication link, and after the network optimization terminal sends back connection confirmation information, transmit the problem area location instruction to the network optimization terminal through the two-way communication link, and receive the instruction reception confirmation information sent back by the network optimization terminal.

[0110] In this embodiment, within an urban network coverage area scenario, a bidirectional communication link is first established with the network optimization terminal using a network communication protocol. Then, a connection request is sent to the network optimization terminal, and confirmation information is awaited. Once confirmation is received, it indicates that the communication link has been successfully established, and the problem area location command is transmitted to the network optimization terminal through this link. Finally, a confirmation message is received from the network optimization terminal confirming that the command has been successfully delivered. This method ensures that the problem area location command is sent to the network optimization terminal in a timely and accurate manner, enabling network optimization personnel to take prompt measures to resolve network problems.

[0111] Figure 2 The illustration shows exemplary hardware and software components of a problem area localization system 100 for network quality assessment, which can implement the ideas of this application, according to some embodiments of this application. For example, a processor 120 can be used in the problem area localization system 100 for network quality assessment and to perform the functions of this application.

[0112] The problem area localization system 100 applied to network quality assessment can be a general-purpose server or a special-purpose server; both can be used to implement the problem area localization method for network quality assessment described in this application. Although only one server is shown in this application, for convenience, the functions described in this application can be implemented in a distributed manner on multiple similar platforms to balance the load.

[0113] For example, a problem area location system 100 used in network quality assessment may include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and various forms of storage media 140, such as a disk, ROM, or RAM, or any combination thereof. Exemplarily, the problem area location system 100 may also include program instructions stored in ROM, RAM, or other types of non-transitory storage media, or any combination thereof. The methods of this application can be implemented according to these program instructions. The problem area location system 100 also includes an I / O interface 150 between the computer and other input / output devices.

[0114] For ease of explanation, only one processor is described in the problem area localization system 100 used in network quality assessment. However, it should be noted that the problem area localization system 100 used in network quality assessment in this application may also include multiple processors, and therefore the steps performed by one processor described in this application may also be performed jointly by multiple processors or individually. For example, if the processor of the problem area localization system 100 used in network quality assessment performs steps A and B, it should be understood that steps A and B may also be performed jointly by two different processors or individually by one processor. For example, the first processor performs step A, the second processor performs step B, or the first processor and the second processor jointly perform steps A and B.

[0115] Furthermore, embodiments of the present invention also provide a readable storage medium, wherein computer-executable instructions are preset in the readable storage medium, and when the processor executes the computer-executable instructions, the problem area localization method applied to network quality assessment as described above is implemented.

[0116] It should be noted that, in order to simplify the description of the present invention and thus help to understand one or more embodiments of the invention, multiple features may sometimes be grouped into one embodiment, drawing or description thereof in the foregoing description of the embodiments of the present invention.

Claims

1. A method for locating problem areas in network quality assessment, characterized in that, The method includes: Obtain a multi-source network data set required for network quality assessment, wherein the multi-source network data set includes signal strength data, transmission rate data, packet loss rate data, and user complaint information data; Based on the spatiotemporal correlation attributes of each data type in the multi-source network data set, a multi-dimensional feature map of network quality is constructed. The multi-dimensional feature map of network quality reflects the temporal variation and spatial distribution correlation of each data type in different network regions. The pre-trained network anomaly clustering model is invoked to perform anomaly clustering and source tracing processing on the multi-dimensional feature map of network quality, generating network anomaly clustering results. The network anomaly clustering results include network data clusters with similar anomaly features and anomaly feature descriptions corresponding to the network data clusters. Based on the preset network region division rules, the network anomaly clustering results are spatially correlated and located to extract the abnormal data cluster features corresponding to each network region divided within the network coverage area and generate problem region features. Based on the characteristics of the problem area, a problem area location instruction is generated, which includes a network area identifier, an anomaly feature description, and the source of the anomaly data. The problem area location instruction is then sent to the network optimization terminal.

2. The problem area localization method applied to network quality assessment according to claim 1, characterized in that, The construction of a multi-dimensional feature map of network quality based on the spatiotemporal correlation attributes of various data types in the multi-source network data set includes: The signal strength data, transmission rate data, packet loss rate data, and user complaint information data in the multi-source network data set are decomposed into data dimensions, and the time and spatial attribute features corresponding to the signal strength data, transmission rate data, packet loss rate data, and user complaint information data are extracted. The time attribute features include the time series information of data collection, and the spatial attribute features include the network area location information corresponding to the data. Establish a time attribute association model, perform time-series alignment processing on the time-series information of signal strength data, transmission rate data, packet loss rate data, and user complaint information data, and generate joint time-series association relationships for multiple data types; A spatial attribute association model is established, which spatially overlays the network area location information of signal strength data, transmission rate data, packet loss rate data, and user complaint information data to generate a joint spatial distribution association relationship for multiple data types. Based on the joint time series correlation and the joint spatial distribution correlation, corresponding feature weights are assigned to the four types of data included in the multi-source network data set: signal strength data, transmission rate data, packet loss rate data, and user complaint information data. The feature weights are determined according to the impact of the data types on network quality assessment. By integrating the joint time series correlation with feature weights and the joint spatial distribution correlation in a graph-like manner, a multi-dimensional feature graph of network quality is constructed, which includes a time dimension layer, a spatial dimension layer, and a data feature layer.

3. The problem area localization method applied to network quality assessment according to claim 2, characterized in that, Based on the joint time series correlation and the joint spatial distribution correlation, corresponding feature weights are assigned to the four data types included in the multi-source network data set: signal strength data, transmission rate data, packet loss rate data, and user complaint information data. These weights include: Retrieve a pre-defined list of data type influence factors in the field of network quality assessment. The list of data type influence factors includes the basic influence factors corresponding to signal strength data, the basic influence factors corresponding to transmission rate data, the basic influence factors corresponding to packet loss rate data, and the basic influence factors corresponding to user complaint information data. The time series fluctuation amplitudes of signal strength data, transmission rate data, packet loss rate data, and user complaint information data in the joint time series correlation are analyzed. The time series fluctuation amplitudes of signal strength data are numerically correlated with the corresponding basic influencing factors to obtain the time dimension weight coefficients of the signal strength data, the transmission rate data, the packet loss rate data, and the user complaint information data. The spatial coverage density of signal strength data, transmission rate data, packet loss rate data, and user complaint information data in the joint spatial distribution correlation are analyzed. The spatial coverage density of signal strength data is numerically correlated with the basic influence factors corresponding to the signal strength data to obtain the spatial dimension weight coefficient of signal strength data. The spatial coverage density of transmission rate data is numerically correlated with the basic influence factors corresponding to the transmission rate data to obtain the spatial dimension weight coefficient of transmission rate data. The spatial coverage density of packet loss rate data is numerically correlated with the basic influence factors corresponding to the packet loss rate data to obtain the spatial dimension weight coefficient of packet loss rate data. The spatial coverage density of user complaint information data is numerically correlated with the basic influence factors corresponding to the user complaint information data to obtain the spatial dimension weight coefficient of user complaint information data. The comprehensive feature weights corresponding to the signal strength data are obtained by weighting and summing the time dimension weight coefficients and spatial dimension weight coefficients of the signal strength data; the comprehensive feature weights corresponding to the transmission rate data are obtained by weighting and summing the time dimension weight coefficients and spatial dimension weight coefficients of the transmission rate data; the comprehensive feature weights corresponding to the packet loss rate data are obtained by weighting and summing the time dimension weight coefficients and spatial dimension weight coefficients of the packet loss rate data; and the comprehensive feature weights corresponding to the user complaint information data are obtained by weighting and summing the time dimension weight coefficients and spatial dimension weight coefficients of the user complaint information data. The comprehensive feature weights corresponding to signal strength data, transmission rate data, packet loss rate data, and user complaint information data are normalized. The comprehensive feature weights of each data type are adjusted by linear scaling so that the sum of the comprehensive feature weights of each data type reaches a fixed value. Finally, the feature weights of the four data types (signal strength data, transmission rate data, packet loss rate data, and user complaint information data) in the multi-dimensional feature map of network quality are determined.

4. The problem area localization method applied to network quality assessment according to claim 1, characterized in that, The process involves calling a pre-trained network anomaly clustering model to perform anomaly clustering and source tracing on the multi-dimensional feature map of network quality, generating network anomaly clustering results, including: The network quality multidimensional feature map is input into the map parsing layer of the network anomaly clustering model, and the network quality multidimensional feature map is split into multiple sub-map units, each sub-map unit corresponding to a preset time interval and a preset spatial sub-region. Feature vectors are extracted from each of the split sub-graph units, and the temporal attribute features, spatial attribute features, and data feature weights in the sub-graph units are converted into sub-graph feature vectors of a unified dimension. All sub-graph feature vectors are input into the clustering layer of the network anomaly clustering model, and density clustering algorithm is used to cluster all sub-graph feature vectors to initially generate multiple candidate data clusters; For each initially generated candidate data cluster, anomaly feature verification is performed, and anomaly feature indices of the feature vectors of each sub-map in the candidate data cluster are extracted. The consistency of the anomaly feature indices within the candidate data cluster is calculated. Candidate data clusters that meet the preset consistency criteria are retained as abnormal data clusters. An abnormal feature description label is added to each abnormal data cluster. The abnormal feature description label includes the data type composition within the cluster, the range of abnormal feature indicators, and the spatiotemporal range of data collection. All abnormal data clusters and their corresponding abnormal feature description labels are integrated to generate network abnormal clustering results.

5. The problem area localization method applied to network quality assessment according to claim 4, characterized in that, The process of performing anomaly feature verification on each initially generated candidate data cluster, extracting anomaly feature indices from the feature vectors of each sub-map within the candidate data cluster, and calculating the consistency of anomaly feature indices within the candidate data cluster includes: The pre-set abnormal feature index library is retrieved from the network anomaly clustering model. The abnormal feature index library includes abnormal fluctuation indexes corresponding to signal strength data, abnormal attenuation indexes corresponding to transmission rate data, abnormal change indexes corresponding to packet loss rate data, and abnormal frequency indexes corresponding to user complaint information data. For each candidate data cluster, extract the index values ​​corresponding to the abnormal feature index library from the feature vectors of all sub-maps within the cluster to form the abnormal index set of that candidate data cluster. Calculate the standard deviation of the indicator values ​​within each set of outliers. The standard deviation value corresponds to the consistency of the outlier within the candidate data cluster. When the standard deviation value is in a preset low range, the consistency of the outlier within the candidate data cluster meets the requirements. The average standard deviation of all outlier indicators for each candidate data cluster is calculated to obtain the overall standard deviation of that candidate data cluster. The overall standard deviation is compared with a preset consistency threshold. If the overall standard deviation is less than the consistency threshold, the consistency of the candidate data cluster is determined to meet the preset standard. If the overall standard deviation is greater than or equal to the consistency threshold, the consistency of the candidate data cluster is determined to not meet the preset standard.

6. The problem area localization method applied to network quality assessment according to claim 1, characterized in that, The process involves combining preset network region segmentation rules to perform spatial correlation and localization processing on the network anomaly clustering results, extracting the corresponding anomaly data cluster features within each network region segmented within the network coverage area, and generating problem region features, including: Retrieve preset network area division rules, which include area division standards based on geographical boundaries and area division standards based on network device coverage. Divide the entire network coverage area into multiple independent network areas according to the network area division rules. Establish a spatial mapping relationship between network regions and anomalous data clusters, match the spatial sub-regions corresponding to each anomalous data cluster in the network anomaly clustering results with the divided network regions, and determine the network region to which each anomalous data cluster belongs; For each network region, feature integration is performed on all anomalous data clusters, and anomalous feature description labels of each anomalous data cluster are extracted. The number of anomalous data clusters and the frequency of occurrence of various anomalous features in the network region are counted. The impact of anomalies in each network region is calculated, and the impact of anomalies is calculated based on the number of anomalous data clusters in the network region, the frequency of occurrence of various anomalous features, and the feature weights corresponding to the anomalous data clusters. For each network region, a feature description is generated that includes the composition of abnormal data clusters, the type of abnormal features, the impact of the abnormality, and the spatiotemporal range of data collection. The feature descriptions of the composition of abnormal data clusters, the type of abnormal features, the degree of abnormal impact, and the spatiotemporal range of data collection are used as the features of the problem region.

7. The problem area localization method applied to network quality assessment according to claim 6, characterized in that, The calculation of the anomaly impact of each network region is derived by comprehensively calculating the number of anomalous data clusters within the network region, the frequency of occurrence of various anomaly features, and the feature weights corresponding to the anomalous data clusters, including: An influence coefficient is assigned to each anomalous data cluster within the network area. The influence coefficient is determined based on the feature weights corresponding to the anomalous data cluster. The feature weights corresponding to the anomalous data cluster correspond to the influence coefficients. When the feature weights corresponding to the anomalous data cluster change, the influence coefficients change accordingly. The sum of the influence coefficients of all anomalous data clusters within the network region is calculated to obtain the basic influence value of the network region. The frequency of occurrence of various abnormal features within the statistical network area is counted. The frequency of occurrence of each abnormal feature is weighted separately. The frequency value of the abnormal feature corresponds to the corresponding weighting coefficient value. When the frequency value of the abnormal feature changes, the weighting coefficient value changes accordingly. The frequency influence value is obtained by calculating the sum of the weighted frequencies of all abnormal features. The basic impact value and the frequency impact value are weighted and summed according to a preset ratio to obtain the initial anomaly impact of the network area; The initial anomaly impact of all network regions is sorted, and the initial anomaly impact is classified into levels according to the sorting results. A level label is added to the anomaly impact of each network region, and the anomaly impact of each network region is finally determined.

8. The problem area localization method applied to network quality assessment according to claim 1, characterized in that, The step of generating a problem area location instruction based on the characteristics of the problem area, including a network area identifier, anomaly feature description, and anomaly data source, and sending the problem area location instruction to the network optimization terminal, includes: The characteristics of the problem area are analyzed, and the network area identifier, anomaly feature type, anomaly impact level, anomaly data cluster composition, and spatiotemporal range of anomaly data collection corresponding to each network area are extracted. According to the instruction format that the network optimization terminal can recognize, a positioning instruction framework is constructed. The positioning instruction framework includes an instruction header, a network area information segment, an abnormal feature information segment, an abnormal data source segment, and an instruction tail. Fill the extracted network area identifier into the network area information segment of the positioning command frame, fill the abnormal feature type and abnormal impact level into the abnormal feature information segment of the positioning command frame, and fill the abnormal data cluster composition and the spatiotemporal range of abnormal data collection into the abnormal data source segment of the positioning command frame. The filled positioning instruction frame is encoded, and the text information in the positioning instruction frame is converted into binary code that conforms to the network transmission protocol to generate the problem area positioning instruction. A two-way communication link is established with the network optimization terminal. A connection request is sent through the two-way communication link. After the network optimization terminal sends back connection confirmation information, the problem area location instruction is transmitted to the network optimization terminal through the two-way communication link. The instruction reception confirmation information is received from the network optimization terminal.

9. The problem area localization method applied to network quality assessment according to claim 2, characterized in that, The establishment of the time attribute association model involves time-series alignment processing of the time-series information of signal strength data, transmission rate data, packet loss rate data, and user complaint information data to generate a joint time-series association relationship for multiple data types, including: Determine the smallest time unit for four types of time series information: time series information of signal strength data, time series information of transmission rate data, time series information of packet loss rate data, and time series information of user complaint data. Based on a defined minimum time unit, the time series information of signal strength data is divided into multiple time slices, the time series information of transmission rate data is divided into multiple time slices, the time series information of packet loss rate data is divided into multiple time slices, and the time series information of user complaint information data is divided into multiple time slices. For each time slice, extract the average value of signal strength data, the average value of transmission rate data, the average value of packet loss rate data, and the number of user complaint information data within that time slice to form a time-related data group for that time slice; Arrange all time-related data groups of time slices in chronological order to construct a time-related sequence; The degree of difference between time-related data sets of adjacent time slices is calculated. The degree of difference is determined by summing the absolute differences of corresponding indicators in the two sets of data. The time-related sequence is segmented based on the difference value. Continuous time slices with a difference value less than a preset segmentation threshold are divided into the same time period, generating a joint time series relationship that includes multiple time periods and corresponding time-related data.

10. A problem area localization system applied in network quality assessment, characterized in that, The problem area localization system applied to network quality assessment includes a processor and a memory, the memory and the processor being connected, the memory being used to store programs, instructions or code, and the processor being used to execute the programs, instructions or code in the memory to implement the problem area localization method applied to network quality assessment as described in any one of claims 1-9.

Images