Method and apparatus for evaluating and optimizing user experience rate, device and storage medium

Abstract

The application provides a user experience rate evaluation method and optimization method, device, electronic equipment and computer readable storage medium, and relates to the technical field of communication quality evaluation. The evaluation method comprises: collecting measurement report MR data of a target area in a preset time period, wherein the MR data comprises field information of a plurality of users; performing N-dimensional gridding on the MR data of the target area to obtain field information of each N-dimensional grid, wherein N comprises one of 2, 3 and 4; and evaluating the user experience rate of each N-dimensional grid based on the field information of each N-dimensional grid. At least the problems of the related art, such as inaccurate division of network performance data, poor geographical presentation effect of network performance data and low evaluation accuracy of user experience rate, are solved. The application is suitable for user experience evaluation and optimization scenarios.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of communication quality evaluation, and in particular to a user experience rate evaluation method and optimization method, device, electronic equipment and computer readable storage medium. BACKGROUND

[0002] User experience guarantee of key scenarios is an important part of communication services, and an effective evaluation method of user experience is usually measured by using user experience rate as a standard. The accuracy of user rate capability evaluation is a key factor affecting user experience guarantee.

[0003] However, the existing user experience rate evaluation method can realize geographical presentation of network performance data to a certain extent, but the evaluation accuracy of user experience rate still needs to be improved. Specifically, (1) most of the existing user experience rates are cell-level and regional-level statistics, which cannot efficiently support user experience and customer satisfaction under different services. (2) For a small part of the existing technology that introduces geographical grids, the size and resolution of the grid are often rough, which cannot fully reflect the subtle geographical differences, and may lead to inaccurate division of network performance data, poor geographical presentation effect of network performance data, and low evaluation accuracy of user experience rate. SUMMARY

[0004] The technical problem to be solved by the present application is to solve the above-mentioned deficiencies of the prior art, and to provide a user experience rate evaluation method, optimization method, device, electronic equipment and computer readable storage medium. The method can reduce the grid size of network performance data, improve the geographical presentation effect of network performance data, expand the application of user experience and customer satisfaction evaluation under different service scenarios, and improve the evaluation accuracy, efficiency and applicability of user experience rate.

[0005] In a first aspect, the present application provides a user experience rate evaluation method, comprising: collecting measurement report (MR) data of a target area in a preset time period, wherein the MR data includes field information of a plurality of users; performing N-dimensional gridding on the MR data of the target area to obtain field information of each N-dimensional grid, wherein N includes one of 2, 3 and 4; and evaluating the user experience rate of each N-dimensional grid based on the field information of each N-dimensional grid.

[0006] Preferably, the MR data of the target area is N-dimensional rasterized to obtain field information of each N-dimensional grid, specifically including: obtaining geographical position information of the target area, and performing rasterization and grid layering processing on the target area to obtain geographical position information of each N-dimensional grid, wherein the N-dimensional grid is used to represent grids at different levels; performing user position positioning on the MR data to obtain N-dimensional position information of a plurality of users; based on the geographical position information of each N-dimensional grid, the N-dimensional position information of the plurality of users, and a clustering algorithm, backfilling field information of the plurality of users to obtain field information of each N-dimensional grid, wherein the clustering algorithm includes any one of the following: K-means algorithm, density-based spatial clustering with noise application DBSCAN algorithm, and Mean Shift algorithm.

[0007] Preferably, the user position positioning on the MR data to obtain N-dimensional position information of a plurality of users specifically includes: performing MR aggregation on the MR data to obtain a plurality of MR groups; performing angle of arrival (AOA) estimation on the MR data to obtain arrival angle information of a plurality of users; performing ray intersection based on topological relationship on the arrival angle information of the same MR group to obtain initial N-dimensional position information of a plurality of users; performing MR calibration on the initial N-dimensional position information of a plurality of users based on a ray tracing model to obtain N-dimensional position information of a plurality of users.

[0008] Preferably, the field information includes modulation and coding (MCS) field information, rank (RANK) field information, and resource block (RB) field information, and the user experience rate of each N-dimensional grid is evaluated based on the field information of each N-dimensional grid, specifically including: performing mean regression calculation on the field information of each N-dimensional grid to obtain regression MCS field information, regression RANK field information, and regression RB field information of each N-dimensional grid; calculating the user experience rate of each N-dimensional grid based on the regression MCS field information, the regression RANK field information, and the regression RB field information of each N-dimensional grid.

[0009] Preferably, the mean regression calculation on the field information of each N-dimensional grid to obtain mean regression field information of each N-dimensional grid specifically includes: performing mean regression calculation on the field information of each N-dimensional grid according to formula (1):

[0010] Yt = a + bXt + et (1), wherein Yt represents regression field information at time t, Xt represents field information at time t-1, a and b represent constants, and et represents a mean regression error term at time t.

[0011] Preferably, the regression MCS field information, the regression RANK field information and the regression RB field information of each N-dimensional grid are used to calculate the user experience rate of each N-dimensional grid, and specifically, the user experience rate of each N-dimensional grid is calculated according to formula (2):

[0012] Grid Rate=AVERAGE(MCS)×AVERAGE(Rank)×AVERAGE(RNNum) (2),

[0013] wherein Grid Rate represents the user experience rate of each N-dimensional grid, MCS represents the regression MCS field information of each N-dimensional grid, Rank represents the regression RANK field information of each N-dimensional grid, and RB Num represents the regression RB field information of each N-dimensional grid.

[0014] In a second aspect, the present application further provides a user experience rate optimization method, comprising: obtaining the grid-level user experience rate evaluated by the user experience rate evaluation method provided in the first aspect, wherein the grid-level user experience rate refers to the user experience rate of each N-dimensional grid; clustering all grid-level user experience rates in the same scene and calculating the scene-level user experience rate, wherein the scene-level user experience rate refers to the average of all grid-level user experience rates in each scene; visually presenting the problem area, wherein the problem area refers to an area where the grid-level user experience rate or the scene-level user experience rate is lower than a preset threshold; performing root cause positioning on the problem area based on an anomaly detection algorithm and / or a path loss and interference analysis model, so as to optimize the user experience rate of the problem area based on the root cause positioning result.

[0015] In a third aspect, the present application further provides a user experience rate evaluation device, comprising: a collection module, a gridding module and an evaluation module, the collection module is used to collect measurement report (MR) data of a target area in a preset time period, wherein the MR data comprises position information and field information of a plurality of users, the gridding module is connected with the collection module and is used to perform N-dimensional gridding on the MR data of the target N-dimensional space area to obtain field information of each N-dimensional grid, wherein N includes one of 2, 3 and 4, and the evaluation module is connected with the gridding module and is used to evaluate the user experience rate of each N-dimensional grid based on the field information of each N-dimensional grid.

[0016] In a fourth aspect, the present application further provides a user experience rate optimization device, comprising: an acquisition module, a clustering module, a visualization module and a root cause positioning module. The acquisition module is configured to acquire the grid-level user experience rate evaluated by the user experience rate evaluation device provided in the third aspect. The grid-level user experience rate refers to the user experience rate of each N-dimensional grid. The clustering module is connected to the acquisition module and configured to cluster all grid-level user experience rates in the same scene and calculate the scene-level user experience rate. The visualization module is connected to the clustering module and configured to visualize the problem area where the grid-level user experience rate or the scene-level user experience rate is lower than the preset threshold. The root cause positioning module is connected to the visualization module and configured to perform root cause positioning on the problem area based on an anomaly detection algorithm and / or a path loss and interference analysis model, so as to optimize the user experience rate of the problem area based on the root cause positioning result.

[0017] In a fifth aspect, the present application further provides an electronic device comprising a memory and a processor. The memory stores a computer program. The processor is configured to run the computer program to implement the user experience rate evaluation method provided in the first aspect or the user experience rate optimization method provided in the second aspect.

[0018] In a sixth aspect, the present application further provides a computer readable storage medium storing a computer program. When the computer program is executed by a processor, the user experience rate evaluation method provided in the first aspect or the user experience rate optimization method provided in the second aspect is implemented.

[0019] The user experience rate evaluation method and optimization method, device, electronic device and computer readable storage medium provided by the present application can effectively improve the geographic presentation effect of network performance data by performing multi-dimensional gridding on the MR data in the target area and then presenting the field information in the MR data in a fine-grained grid. The field information in the fine-grained grid can be used to better evaluate the grid-level user experience rate, which can then be applied to the user experience and customer satisfaction evaluation in different business scenarios, thereby improving the evaluation accuracy and applicability of the user experience rate. Therefore, the present application can reduce the gridding granularity of network performance data, improve the geographic presentation effect of network performance data, expand the application of user experience and customer satisfaction evaluation in different business scenarios, and improve the evaluation accuracy, efficiency and applicability of the user experience rate. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 A flowchart of a user experience rate evaluation method according to Embodiment 1 of the present application;

[0021] Figure 2 An example diagram of user location positioning on MR data according to Embodiment 1 of the present application;

[0022] Figure 3 An example diagram for AOA estimation on MR data in embodiment 1 of the present application;

[0023] Figure 4 An example diagram for ray intersection based on topological relationship on AOA information of the same MR group in embodiment 1 of the present application;

[0024] Figure 5 An example diagram for backfilling field information of several users in embodiment 1 of the present application;

[0025] Figure 6 An example diagram for field information obtained by occupying different service cells in the same grid in embodiment 1 of the present application;

[0026] Figure 7 An example diagram for one kind of user experience rate visualization presentation in embodiment 1 of the present application;

[0027] Figure 8 An example diagram for another kind of user experience rate visualization presentation in embodiment 1 of the present application;

[0028] Figure 9 An example diagram for a scene root cause positioning in embodiment 1 of the present application;

[0029] Figure 10 An example diagram for coverage and rate before optimization in embodiment 1 of the present application;

[0030] Figure 11 An example diagram for coverage and rate after optimization in embodiment 1 of the present application;

[0031] Figure 12 A flow chart of one kind of user experience rate optimization method in embodiment 2 of the present application;

[0032] Figure 13 A structural schematic diagram of one kind of user experience rate evaluation device in embodiment 3 of the present application;

[0033] Figure 14 A structural schematic diagram of one kind of user experience rate optimization device in embodiment 4 of the present application. DETAILED DESCRIPTION

[0034] In order to make the skilled in the art better understand the technical solutions of the present application, the embodiments of the present application will be further described in detail below with reference to the drawings.

[0035] It can be understood that the specific embodiments and drawings described herein are only used to explain the present application, but not to limit the present application.

[0036] It can be understood that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict.

[0037] It can be understood that, for the convenience of description, only parts related to the present application are shown in the drawings of the present application, and parts irrelevant to the present application are not shown in the drawings.

[0038] It can be understood that each unit and module involved in the embodiments of the present application can correspond to only one entity structure, or can be composed of multiple entity structures, or multiple units and modules can be integrated into one entity structure.

[0039] It can be understood that the functions and steps marked in the flowcharts and block diagrams of the present application can occur in an order different from that marked in the drawings without conflict.

[0040] It can be understood that in the flowcharts and block diagrams of the present application, the architecture, functions and operations of the possible implementations of the system, device, equipment and method according to the embodiments of the present application are shown. Each block in the flowchart or block diagram can represent a unit, module, program segment, code, which contains executable instructions for realizing the specified functions. Moreover, each block or combination of blocks in the block diagram and flowchart can be realized by a hardware-based system for realizing the specified functions, or by a combination of hardware and computer instructions.

[0041] It can be understood that the units and modules involved in the embodiments of the present application can be realized in the form of software or in the form of hardware, for example, the units and modules can be located in a processor.

[0042] Embodiment 1:

[0043] As shown in Figure 1 , the present embodiment provides a user experience rate evaluation method.

[0044] The user experience rate evaluation method comprises the following steps.

[0045] S101, collecting measurement report (MR) data of a target area in a preset time period, wherein the MR data comprises field information of a plurality of users.

[0046] In this embodiment, network performance data is usually presented in two forms: MR (Measurement Report) data from the network management system and MDT (Multi-Domain Test) data collected through road testing. MR data is performance data generated by user equipment (UE) and reported to the network management system, and typically includes user signal quality, signal strength, and connection state-related field information. MDT data is performance data collected through road testing or specialized network testing equipment, and typically involves detailed testing of network performance in a specific area or route. Therefore, MR data itself usually does not include specific geographic location information of the user, while MDT data collected during testing at a specific location can contain location data to provide information about the user's specific location in the network. This embodiment can directly collect MR data of a target area from the network management system within a preset time period to evaluate user perceived speed, which is more real-time than collecting MDT data of the area through road testing, and can reflect the current network performance, saving time and effort, improving the evaluation efficiency and accuracy of user experience speed, and more timely identifying network coverage and performance problems.

[0047] S102, performing N-dimensional gridding on the MR data of the target area to obtain field information of each N-dimensional grid, wherein N includes one of 2, 3, and 4.

[0048] Specifically, S102: performing N-dimensional gridding on the MR data of the target area to obtain field information of each N-dimensional grid, including steps S1021-S1023:

[0049] S1021, obtaining geographic location information of the target area and performing gridding and gridding layering on the target area to obtain geographic location information of each N-dimensional grid, wherein the N-dimensional grid is used to represent grids at different levels.

[0050] In this embodiment, N=3 is taken as an example, geographical position information of the target area is acquired, and rasterization and raster layering processing are performed on the target area to divide the target area into a plurality of 3-dimensional raster units (i.e., 3-dimensional raster), wherein 3-dimensional refers to longitude, latitude and height. Each 3-dimensional raster is assigned a unique raster ID for identifying its geographical position. The size of the raster can be flexibly set according to the requirements of the scene (such as 10m x 10m or 100m x 100m) to achieve a suitable spatial resolution. Taking an urban scene as an example, since the buildings in the urban area are dense and the user distribution density is high, in order to facilitate subsequent obtaining of higher positioning accuracy, the target area is divided into 20m x 20m raster units. Taking a rural and suburban scene as an example, since the user density is low and the environment is open, the rural and suburban areas can be divided into 50m x 50m raster units. The larger raster unit can reduce the amount of calculation while maintaining a certain positioning accuracy.

[0051] It should be noted that the raster layering processing includes a plurality of different division modes, which depends on the application scenario and the data type, for example, the raster can be divided according to indoor and outdoor or the height of multi-layer buildings. In this embodiment, N=3 is taken as an example, and the 3-dimensional raster refers to the rasterization of the target area and the raster after the preset height raster layering processing.

[0052] If the target area includes urban and rural and suburban scenes, the target area can be divided into different rasters, such as 20m x 20m and 50m x 50m.

[0053] S1022, user position positioning is performed on the MR data to obtain N-dimensional position information of a plurality of users.

[0054] Specifically, S1022: user position positioning is performed on the MR data to obtain N-dimensional position information of a plurality of users, including: MR aggregation is performed on the MR data to obtain a plurality of MR groups; angle of arrival (AOA) estimation is performed on the MR data to obtain angle of arrival information of a plurality of users; ray intersection based on topological relationship is performed on the angle of arrival information of the same MR group to obtain initial N-dimensional position information of a plurality of users; MR calibration is performed on the initial N-dimensional position information of a plurality of users based on a ray tracing model to obtain N-dimensional position information of a plurality of users.

[0055] In this embodiment, user position positioning is performed on the MR data to obtain 3-dimensional position information of a plurality of users, such as Figure 2As shown, specifically includes: ① using MR aggregation, aggregate data features in MR data, complete MR coarse positioning, get several MR groups, MR group can represent user group, coarse positioning result is used to represent the user belongs to the group, wherein the data features include but not limited to: user features, call features, cell features and level features. ② using user orientation AOA(Angle-of-Arrival, angle of arrival ranging) estimation, first accurate correction of user positioning for several MR groups, get several user arrival angle information, first accurate correction result is used to represent the arrival angle information of each user in MR group. ③ based on the topological relationship ray intersection, secondary correction of user positioning for several MR groups, get several user's initial 3D position information, secondary correction result is used to represent the initial 3D position information of each user in MR group. ④ after using MR aggregation, aggregate data features in MR data, get several MR groups, also use several MR groups to build MR indoor and outdoor model; based on the positioning result (i.e. Figure 2 Map information) and MR indoor and outdoor model, the initial 3D position information of several users (i.e. the initial 3D position information of all users in all MR groups) is corrected three times, and the 3D position information of several users is obtained, and the three times correction result is the 3D position information of several users, wherein the indoor and outdoor model includes but is not limited to ray tracing model; in addition, in addition to correcting the initial 3D position information of several users based on the positioning result and the indoor and outdoor model of MR, this embodiment can also use Markov chain, Kalman filter or time correlation method to correct the initial 3D position information of several users. This embodiment realizes indoor and outdoor separation and 3D positioning by gridding and gridding layering processing of target area and user position positioning of MR data, improves the positioning accuracy of MR data, improves the matching accuracy of field information under the grid, and improves the distribution accuracy of network performance data.

[0056] Using MR aggregation, aggregate data features in MR data, complete MR coarse positioning, specifically includes: modeling the path of each user in MR data and data features, the purpose is to simulate the level performance and abnormal event performance of users in indoor and outdoor, which can roughly judge the position (i.e. indoor or outdoor) of users in the same level features and call features, get indoor MR group and outdoor MR group. For the superposition of level features, it can also be based on the existing model, such as level coverage model function for coverage modeling, while superimposing call features and many data features of users, wherein the formula of level coverage model function is:

[0057] The first accurate correction of user positioning for several MR groups is performed by using user azimuth AOA (Angle-of-Arrival) estimation, as shown in Figure 3 The angle position estimation of the user can be performed by using a private field AOA in the MR of the actual user. When the user has multiple serving cells (which can also be regarded as base stations) as the adjacent cells, the rough and fast positioning of the talking user in the network is performed by using network measurement, network structure and user behavior, a multi-dimensional equation set is constructed according to the connection state of the user and the base station in the network, the initial positioning result (i.e. the angle of arrival information of several users) is calculated by simultaneous solution, and the initial positioning result (i.e. the MR groups) after the MR aggregation is corrected.

[0058] The second correction of user positioning for several MR groups is performed by using the ray intersection based on the topological relationship, as shown in Figure 4 The relevant information of the serving cell recognized by the user is spliced from the MR data, and the relevant information of the serving cell includes the serving cell ID (i.e. CELL1, CELL2 and CELL3 in Figure 4 ), the serving cell position (i.e. (x1, y1), (x2, y2) and (x3, y3) in Figure 4 ), the antenna hanging height, the azimuth angle, the downtilt angle, the lobe width, the pilot power and the serving cell marker. The time delay (i.e. Time Dlay in Figure 4 ) and the level information of the angle of arrival information of the user are calculated from the MR, the 3D algorithm of line-circle intersection is used to further calibrate the angle of arrival information of several users (i.e. F(RSCP1, Araval1,...) in Figure 4 , F(RSCP2, Araval2,...) and F(RSCP2, Araval2,...)), and the initial 3D position information of several users (i.e. UEPoisition in Figure 4 ) is obtained. This embodiment takes the 3D algorithm of line-circle intersection as the WCCL (Weighted Centroid Correction Localization) fast positioning algorithm as an example.

[0059] Taking the indoor and outdoor model of the MR as a ray tracing model as an example, based on the positioning result and the indoor and outdoor model of the MR, the initial 3D position information of a plurality of users is corrected three times, specifically including: ① according to the 3D map work parameter (i.e. the positioning result), the ray tracing model (Volcano) forms a basic 3D library, so as to strictly correspond the initial N-dimensional position information of a plurality of users and the map trajectory. ② Then, the initial 3D position information of a plurality of users is corrected (including indoor and outdoor, height, latitude and longitude) by using the historical positioning data feature library and the initial 3D position information of a plurality of users for feature matching. ③ The corrected result (i.e. the N-dimensional position information of a plurality of users) is imported into the map for final calibration, and the calibrated result is imported into the historical positioning data feature library, and the historical positioning data feature library is iteratively updated for the next time of user position positioning of the MR data.

[0060] In S1023, based on the geographic position information of each N-dimensional grid, the N-dimensional position information of a plurality of users and a clustering algorithm, the field information of a plurality of users is backfilled to obtain the field information of each N-dimensional grid, wherein the clustering algorithm includes any one of the following: K-means algorithm, density-based spatial clustering with noise application DBSCAN algorithm and Mean Shift algorithm.

[0061] In this embodiment, as shown in Figure 5 The field information of a plurality of users is backfilled, specifically including: based on the field information and the N-dimensional position information of a plurality of users of the MR data, a data set is created. The massive data set and the grid layer position (i.e. the geographic position information of each N-dimensional grid) are matched by grid clustering, all data information of the corresponding position grid of the attribution grid ID is clustered based on the grid ID, and the field information based on the grid ID is output, as shown in Table 1. Through the clustering algorithm, the field information of the massive MR data and the grid can be matched in this embodiment, accurate, grid-level network performance data distribution is provided, and then the evaluation accuracy of the subsequent user experience rate is improved.

[0062] Table 1 Field information based on grid ID

[0063]

[0064] It should be noted that the K-means algorithm is a common unsupervised clustering algorithm, which iteratively assigns data to K cluster centers so that the distance between each data point and the nearest cluster center is minimized. The DBSCAN algorithm (Density-Based Spatial Clustering of Applications with Noise) is a density-based clustering algorithm suitable for spatial data clustering and can effectively identify noise data. The DBSCAN algorithm is particularly suitable for processing MR data with uneven distribution and can identify areas with high density and isolated points when clustering in a grid. The Mean Shift clustering algorithm is a kernel density estimation-based clustering method that gradually moves data points to the location with the highest density, eventually forming clusters, and is suitable for situations where data points have uneven density distributions.

[0065] Based on the geographic location information of each 3D grid, the 3D location information of a plurality of users, and the K-means algorithm, the field information of a plurality of users is backfilled to obtain the field information of each 3D grid, specifically including: ① taking the 3D location information (longitude, latitude, and height) of a plurality of users and the field information of a plurality of users as input feature vectors. ② Set the number of clusters K according to the density of grid division and the number of grids in the target area, and determine the geographic location information of the K cluster centers according to the geographic location information of each 3D grid. ③ Through multiple iterations, calculate the distance between the 3D location information of a plurality of users and the geographic location information of the K cluster centers, assign a plurality of users to the nearest cluster center, and continuously update the geographic location information of the cluster center. ④ Finally, assign the field information of a plurality of users to a grid ID, and calculate the average or other statistical characteristics of each field information in the grid as the clustering result of the grid.

[0066] Based on the geographic location information of each 3D grid, the 3D location information of a plurality of users, and the DBSCAN algorithm, the field information of a plurality of users is backfilled to obtain the field information of each 3D grid, specifically including: ① The input data is also the 3D location information (longitude, latitude, and height) of a plurality of users and the field information of a plurality of users. ② According to the geographic location information of each 3D grid, set the neighborhood radius (eps) and the minimum point number (minPts), which are used to determine whether the field information of a plurality of users belongs to an area with high enough density. ③ The DBSCAN algorithm traverses each point (i.e., user) in the data set to determine whether the number of points in its neighborhood range is greater than minPts. If so, it is determined whether all points in the neighborhood range greater than minPts belong to the same cluster, otherwise they are marked as noise points. ④ Output result: a series of grid IDs and clustering results are obtained, and the field information in the grid is statistically summarized.

[0067] Based on the geographical position information of each 3D grid, the 3D position information of the users, and the Mean Shift clustering algorithm, the field information of the users is backfilled to obtain the field information of each 3D grid, specifically including: ① The input data is also the 3D position information (longitude, latitude, and height) of the users and the field information of the users. ② An appropriate kernel function (such as a Gaussian kernel) and a bandwidth parameter are selected to determine the direction and distance of the movement of the data points (i.e., the field information and 3D position information of any user in the data set). ③ The 3D position information of the data points is constantly moved until all the data points are clustered in a high-density area to form a cluster. ④ The output result: according to the final 3D position information of the data points, the data points are assigned to the corresponding grid, and the statistical result after clustering is output.

[0068] In addition, as shown in Figure 6 , in the same grid, the service cells occupied by the users (i.e., cell 1, cell 2, and cell 3 in Figure 6 ) can be more than one. Due to the different distances of the service cells to the current grid, the field information (i.e., MCS1, RANK1, RB1, MCS2, RANK2, RB2, MCS3, RANK3, and RB3 in Figure 6 ) obtained by the users occupying different service cells in the same grid can have a large difference. Therefore, regardless of which clustering algorithm is used, the statistical result after clustering needs to be processed by grid data, which includes but is not limited to field statistical calculation, data filtering, and outlier processing. Field statistical calculation: the field information of each grid is statistically calculated, and at least one of the following is output: mean, standard deviation, maximum, and minimum. Data filtering and outlier processing: according to the distribution characteristics of the field information of all grids, abnormal grids are filtered out, or noise field information is further processed.

[0069] Specifically, the field information includes modulation and coding MCS field information, rank RANK field information, and resource block RB field information.

[0070] It should be noted that the MCS (Modulation and Coding Scheme, used to represent the air interface quality) defines a modulation scheme (Modulation) and a redundancy coding scheme (Code Rate) when a time-frequency resource carries binary data. The higher the MCS, the higher the modulation mode and the corresponding transmission code rate. Under the condition that other conditions (including but not limited to RANK and RB) remain unchanged, to maintain the same bit error rate, the requirement for the air interface is also higher. In low-order MCS, the quantization error of the 256QAM table is larger than that of the 64QAM table, so the middle / remote point in the scenario where low-order MCS is required is preferred to select the 64QAM table. For users in the same cell coverage, near points mean that the probability of obtaining network resource capability and user experience rate is relatively high, and remote points are affected by signal coverage and interference, attenuation, and user experience rate is relatively low. In order to better reflect the user experience rate capability, the user experience rate of the edge user is usually evaluated as the user experience rate capability, so in order to restore the cell network capability rate, the MCS of the 64QAM table is used to evaluate the user experience rate, and the spectral efficiency is determined according to the MCS selection step table. In summary, when the base station selects the modulation, the 256QAM table is selected for the middle / near point user, and the 64QAM table is selected for the middle / remote point user. The base station periodically judges the switching of the two tables, and selects the MCS / CQI (Channel Quality Indication, Channel Quality Indication) / spectral efficiency threshold.

[0071] RANK (multi-path environment or rank) refers to the rank of the MIMO (Multipe Input Multiple Output, Multi-Input Multi-Output) channel matrix, which represents the number of layers that can be simultaneously transmitted. The maximum RANK does not exceed the short side size of the MIMO channel matrix, that is, in the case of Massive MIMO, the maximum value of RANK depends on the number of receiving antennas at the UE side, which is usually 4, that is, there are 1-4 channels. Simply put, the same time-frequency resource is divided into several parts for simultaneous transmission. In the case of constant time-frequency resources, the higher the RANK, the higher the actual throughput rate and the greater the speed.

[0072] RB(Resource Block, resource block) number: 5G(5th Generation Mobile Communication Technology, 5th generation mobile communication technology) adopts TDD(Time Division Duplexing, time division duplexing) networking mode, uses 100M bandwidth (i.e. 273 RBs) as resources to provide user network services, according to time-frequency distribution, in the frequency domain, 1 RB is composed of 12 subcarriers (Resource Element, RE), and in the time domain, each time slot has 14 symbols. Taking the time slot ratio of 7:3 as an example, according to the uplink and downlink distribution, the downlink accounts for 70%, and the uplink accounts for 30%, and the network capacity of each cell remains consistent, that is, the total RB number of the network cell and the number of RE per RB remain basically consistent.

[0073] S103, based on the field information of each N-dimensional grid, evaluating the user experience rate of each N-dimensional grid.

[0074] In this embodiment, there is a detailed introduction and explanation of CQI Table in 3GPP(3rd Generation Partnership Project, 3rd Generation Partnership Project). Each equipment manufacturer uses the terminal to report CQI, and finally gets the user MCS after mapping the wireless environment. The base station determines the transmission efficiency and transmission quality of the user service through the MCS, that is, it also determines the rate capability of the user. Therefore, the air interface rate is mainly affected by MCS, RANK and RB number. Generally, the air interface rate capability of the user at the location is calculated according to the MCS of the main service cell where the user is located, the RANK of the user at the location and the RB number of the user, which specifically includes: the air interface rate capability of the user at the location = (X = MCS) x (Z = RB number) x (Y = Rank), wherein X is the MCS value of the sampling point ≥90%, X is evaluated under full filling condition, so RB corresponds to bandwidth, such as TDD 100MHz carrier bandwidth supports 273 RBs, Y is the average rank value of the user.

[0075] The air interface rate is used to evaluate the maximum carrying capacity of the network, helping the operator to understand and optimize the network architecture, while the user experience rate is used to measure the real experience of the user, analyze the quality of service, and guide the improvement of user satisfaction. The air interface rate and the user experience rate are dependent on each other to a certain extent. Higher air interface rate can provide better user experience rate under ideal conditions. However, the user experience rate is limited by the end-to-end, and is not purely a reflection of the air interface rate capacity of the wireless network. For example, when a user sends an instant message, the user experience rate may be lower than 10 Mbps, but the actual network air interface has no interference and good coverage, and the actual air interface can reach an air interface rate of 1 Gbps. Therefore, in order to isolate the user experience rate and the air interface rate capacity, the air interface rate capacity of the user's location needs to be converted to obtain the user experience rate, wherein the conversion means includes but is not limited to protocol tail packets.

[0076] In summary, the embodiment can calculate the user experience rate by using the index field information for calculating the air interface rate, that is, by using the MCS field information, the RANK field information and the RB field information to calculate the user experience rate, considering the efficiency of MCS, the effect of MIMO technology and the use of resource blocks, so that the evaluation of user experience rate is more accurate and comprehensive.

[0077] Specifically, S103: based on the field information of each N-dimensional grid, evaluating the user experience rate of each N-dimensional grid, including steps S1031-S1032:

[0078] S1031, the field information of each N-dimensional grid is calculated by mean regression to obtain the regression MCS field information, the regression RANK field information and the regression RB field information of each N-dimensional grid.

[0079] Specifically, S1031: the field information of each N-dimensional grid is calculated by mean regression to obtain the regression MCS field information, the regression RANK field information and the regression RB field information of each N-dimensional grid, including: according to formula (1), the field information of each N-dimensional grid is calculated by mean regression:

[0080] Yt=α+βXt+εt(1), wherein Yt represents the regression field information at time t, Xt represents the field information at time t-1, α and β represent constants, and εt represents the mean regression error term at time t.

[0081] In this embodiment, since the field information of the user is changing all the time, the user experience rate also changes many times from the time T when the field information is collected to the time T+1 when the user experience rate is evaluated, so the user experience rate evaluated at the time T+1 can only represent the user experience rate at the time T and cannot realize real-time evaluation of the user experience rate. The mean reversion is a data method, which firstly determines a statistical range (grid ID), and calculates the average value in the statistical range, no matter whether any field information in the statistical range is higher or lower than the value center (or average value), any field information in the statistical range will return to the value center with a high probability. Therefore, this embodiment calculates the mean reversion of the field information of each N-dimensional grid at the time T, such as CellDLInitMcs, CellDLRank and CellDLCPUsedRBNum in Table 1, predicts the field information of each N-dimensional grid at the time T+1 in advance, so as to evaluate the user experience rate at the time T+1, which can effectively improve the accuracy and stability of the evaluation of the user experience rate, provide real-time and reliable user experience insight in a dynamic environment, help the network operator to adjust the strategy in time and optimize the resource allocation, so as to maximize the satisfaction and quality of service of the user.

[0082] S1032, based on the regression MCS field information, the regression RANK field information and the regression RB field information of each N-dimensional grid, calculate the user experience rate of each N-dimensional grid.

[0083] Specifically, S1032, based on the regression MCS field information, the regression RANK field information and the regression RB field information of each N-dimensional grid, calculate the user experience rate of each N-dimensional grid, including: calculating the user experience rate of each N-dimensional grid according to formula (2):

[0084] Grid Rate=AVERAGE(MCS)×AVERAGE(Rank)×AVERAGE(RB Num) (2),

[0085] Wherein, Grid Rate represents the user experience rate of each N-dimensional grid, MCS represents the regression MCS field information of each N-dimensional grid, Rank represents the regression RANK field information of each N-dimensional grid, and RB Num represents the regression RB field information of each N-dimensional grid.

[0086] In this embodiment, according to formula (2), the user experience rate of each 3-dimensional grid in Table 1 is AVERAGE(CellDLInitMcs)×AVERAGE(CellDLRank)×AVERAGE(CellDLCPUsedRBNum).

[0087] It should be noted that after evaluating the user experience rate, it is usually followed by the application and analysis of the user experience rate to address different scenario requirements. The application and analysis of the user experience rate typically includes the geographical presentation of the user experience rate, the aggregation of scenario-level user experience rates, and the optimization of the user experience rate. Through the application and analysis of the user experience rate, after identifying problems, measures are taken to optimize them, such as adjusting bandwidth allocation, improving network equipment, or adjusting service strategies, to improve the overall user experience. That is to say, after the network performance data analysis stage is implemented through steps S101 and S102, and the user experience rate is calculated through step S103, this embodiment can also perform geographical presentation, scenario-level user experience rate aggregation, and optimization of the evaluated user experience rate to identify areas or scenarios with low user experience rates and improve low user experience rates.

[0088] The geographical representation of user experience rate refers to displaying the calculated user experience rate on a map in an intuitive way. Different ranges of user experience rate can be used to represent the impact of network speed on user experience; therefore, different ranges of user experience rate are displayed using different identifiers. For example, with N=2, the geographical representation of user experience rate would look like this: Figure 7 As shown, the user experience rate of each 2D grid is... Figure 7 In terms of 2D grid capabilities, a user experience rate below 10 Mbps indicates a very poor network experience, potentially leading to slow webpage loading, frequent video buffering, and high latency in online games. A user experience rate in the range of [10, 20) Mbps indicates that the user experience is beginning to improve, but buffering and slow loading speeds may still occur, especially when using multiple devices simultaneously. A user experience rate in the range of [20, 50) Mbps indicates sufficient support for streaming media services (such as 1080p video) and basic multi-device operation, but problems may occur during peak hours. A user experience rate in the range of [50, 100) Mbps indicates a relatively good user experience, capable of supporting 4K video streaming, online games, and other bandwidth-intensive activities. A user experience rate above 100 Mbps indicates an excellent network connection, capable of supporting simultaneous use of multiple devices and smooth playback of high-resolution videos, large file downloads, and other bandwidth-intensive applications.

[0089] like Figure 8As shown, the embodiment can also use color coding (such as red for low rate, green for high rate) to visualize the user experience rate distribution of each 3D grid, enabling network optimization personnel to quickly identify low-rate areas. By superimposing the user experience rate of each 3D grid on an electronic map, the spatial distribution of network performance and problem areas are immediately apparent, greatly improving optimization efficiency; it can also be combined with actual terrain and building distribution to assist in assessing network performance problems in coverage blind spots and complex scenarios.

[0090] The aggregation of scene-level user experience rate is the aggregation and analysis of the user experience rate of each 3D grid in different geographic scenes (such as commercial areas, residential areas, transportation hubs) in order to more targetedly optimize network performance. The implementation of the aggregation of scene-level user experience rate specifically includes: ① According to geographic information system (GIS) data, the user experience rate of each 3D grid is distributed to different scene categories (such as outdoor urban areas, indoor commercial centers, subway stations, tunnels). ② Each grid is attached with a scene label so that it can be classified and aggregated according to the scene label in the aggregation and analysis. ③ K-means or DBSCAN algorithm is used to group the grid-level user experience rates of the same scene label, where the grid-level experience rate refers to the user experience rate of each 3D grid. Through the aggregation of scene-level user experience rate, the common problems of each scene can be identified, and the performance bottlenecks of the network in specific scenes can be revealed; through the aggregation and comparison of scenes, network optimization personnel can develop more refined optimization strategies to meet the needs of users in different scenes.

[0091] The grouping of the grid-level user experience rates of the same scene label using K-means or DBSCAN algorithm specifically includes: ① Selecting the feature vectors of each 3D grid: grid-level user experience rate, interference level, and number of users, where the interference level includes but is not limited to RSRP (Reference Signal Receiving Power), SINR (Signal to Interference plus Noise Ratio). ② Setting the initial parameters of clustering, such as the number of clusters K or the minimum cluster density (for DBSCAN). ③ Executing the clustering algorithm to output the center point and range of each cluster group. ④ Scene aggregation formula: aggregating the clustering results within the same scene and calculating the scene-level user experience rate according to the formula: i is the user experience rate of the i-th 3D grid.

[0092] ​The optimization of user experience rate includes root cause positioning and formulating targeted optimization scheme. The root cause positioning aims to find the root cause behind the low user experience rate of the grid through in-depth analysis. The optimization of user experience rate combines multiple algorithms such as multiple linear regression, decision tree, and anomaly detection to investigate the key factors affecting network performance. Common root causes include severe signal interference, uneven resource allocation, and poor coverage. Root cause positioning can identify the main cause of rate decline through multi-dimensional analysis of MR data such as signal strength, signal-to-noise ratio, and resource block utilization. Optimization schemes include adjusting antenna direction, optimizing resource scheduling, or increasing network coverage.

[0093] The multi-dimensional analysis of MR data specifically includes combining multiple indicators of MR data such as RSRP, SINR, CQI, and resource block utilization (RB Utilization) to analyze the impact of each indicator on user experience rate through a multi-dimensional regression model. For example, using a multiple linear regression or decision tree model, i.e., user experience rate = β0 + β1 × RSRP + β2 × SINR + β3 × CQI, the indicator with the greatest impact on user experience rate can be found, where β i is the regression coefficient for each indicator, determined by training the model.

[0094] In addition to combining multiple indicators of MR data and analyzing the impact of each indicator on user experience rate through a multi-dimensional regression model, this embodiment can also perform root cause positioning on problem areas based on an anomaly detection algorithm and / or a path loss and interference analysis model to optimize the user experience rate in problem areas based on the root cause positioning results.

[0095] Based on the anomaly detection algorithm, the root cause positioning of the problem area specifically includes using statistical methods such as Z-score or Tukey's fences or machine learning anomaly detection algorithms such as Isolation Forest and LOF (Local Outliers Factor) to identify outliers in MR data and analyze the impact of outliers on user experience rate.

[0096] Based on the path loss and interference analysis model, the root cause positioning of the problem area specifically includes calculating the path loss of each 3D grid according to the wireless propagation model and locating possible interference sources based on interference analysis, where the wireless propagation model includes but is not limited to the COST231-Hata model and the Okumura-Hata model. Locating possible interference sources based on interference analysis specifically includes finding possible interfering cells or user equipment by analyzing the RSRP and SINR distribution of cells within and around the grid.

[0097] Taking N = 2, the target area as A lake and the surrounding city of B scenic spot as an example, after evaluating the grid-level user experience rate and calculating the scene-level user experience rate, there are 321 scenes in which the user experience rate of A lake and the surrounding city of B scenic spot is at risk, and the proportion of problem scenes is 4.4%. Through TOP scene screening, that is, sorting the user experience rate insufficient scenes, the TOP k scenes in the sorting result are screened to determine the target scene, and the a scene is the target scene in the user experience rate insufficient scene. The scene-level user experience rate of the a scene is geographically presented and the root cause is located to determine that the a scene is mainly a weak coverage problem, and the problem is characterized by 5G downlink rate less than 100Mbps accounting for 21%, as shown in Figure 9 .

[0098] The scene-level user experience rate of the a scene is geographically presented and the root cause is located, specifically including: the problem area under the a scene can be directly presented by geographically presenting all the grid-level user experience rates under the a scene, and it is determined that the problem area has 5G off-network problem, and the root cause of the problem area is located. The average RSRP, SINR, downlink rate and comprehensive coverage rate of the problem area 4G network are-66dBm, 39db, 35Mbps and 100%, respectively, and the average RSRP, SINR, downlink rate and comprehensive coverage rate of the 5G network are-98dBm, 1.6db, 40Mbps and 35.71%, respectively. The 4G coverage is good, the 5G is off-network inside, and it is determined that the problem area has no 5G room distribution in the 4G room distribution. The optimization scheme is developed as follows: shared room distribution distribution common mode, open NR2.1 room distribution to perform 5G coverage blind. After the NR2.1G is opened, the network coverage and user experience are greatly improved. As shown in Figure 10 before optimization and Figure 11 after optimization, the 5G network coverage is from partial off-network to continuous, the overall RSRP value (i.e. NR SS-RSRP in Figure 10 and Figure 11 ) is improved from-98dBm to about-70dBm, the downlink rate is improved from 32Mbps to 113Mbps, the uplink rate is improved from 5Mbps to 66Mbps, and the user perception experience is fully guaranteed.

[0099] The five links of analysis of network performance data, calculation of user experience rate, geographical presentation of user experience rate, aggregation of scene-level user experience rate and optimization of user experience rate are closely connected, and constitute a complete closed-loop optimization process. Through fine evaluation of network quality and rate capacity, and geographical presentation and scene-level aggregation, the root problem of low-rate area can be accurately located, the optimization process is gradually promoted, the network performance is continuously improved, the network coverage quality is improved, and the actual experience of users in different scenes is significantly improved, which provides strong technical support for network optimization and user experience guarantee.

[0100] The user experience rate evaluation method provided in the embodiment can effectively improve the geographical presentation effect of network performance data by performing multi-dimensional gridding on MR data in a target area and then presenting field information in the MR data in a fine-grained grid. The field information in the fine-grained grid can be used to better evaluate the grid-level user experience rate, which can be applied to user experience and customer satisfaction evaluation in different business scenarios, improve the evaluation accuracy and applicability of user experience rate, realize accurate evaluation of user experience rate, reduce the gridding granularity of network performance data, improve the geographical presentation effect of network performance data, expand the application of user experience and customer satisfaction evaluation in different business scenarios, and improve the evaluation accuracy, efficiency and applicability of user experience rate.

[0101] Embodiment 2

[0102] As shown in FIG. 8, the embodiment provides a user experience rate optimization method. The user experience rate optimization method includes the following steps. Figure 12

[0103] S202, obtaining a grid-level user experience rate, wherein the grid-level user experience rate refers to a user experience rate of each N-dimensional grid.

[0104] ​In this embodiment, after the user experience rate is evaluated, the application and analysis of the user experience rate are usually accompanied to solve different scene requirements. The application and analysis of the user experience rate usually include the geographic presentation of the user experience rate, the aggregation of the scene-level user experience rate and the optimization of the user experience rate. Through the application and analysis of the user experience rate, after the problem is identified, measures are taken for optimization, for example, adjusting the bandwidth allocation, improving the network equipment or adjusting the service strategy, to improve the overall experience of the user. That is to say, after the analysis of the network performance data is realized through steps S101 and S102, and the calculation of the user experience rate is realized through step S103, the evaluated user experience rate can be presented geographically, aggregated at the scene level and optimized in this embodiment, so as to identify the area or scene with low user experience rate and improve the low user experience rate.

[0105] In step S202, all grid-level user experience rates in the same scene are clustered, and a scene-level user experience rate is calculated, wherein the scene-level user experience rate refers to the average of all grid-level user experience rates in each scene.

[0106] In this embodiment, the aggregation of the scene-level user experience rate refers to aggregating and analyzing the user experience rate of each 3D grid in different geographic scenes (such as commercial areas, residential areas and transportation hubs), so as to more specifically optimize the network performance. The aggregation of the scene-level user experience rate specifically includes: ①According to the geographic information system (GIS) data, the user experience rate of each 3D grid is distributed to different scene categories (such as outdoor urban area, indoor commercial center, subway station and tunnel). ②Each grid is attached with a scene label, so that it can be classified and aggregated according to the scene label in the aggregation and analysis. ③The K-means or DBSCAN algorithm is used to group the grid-level user experience rates with the same scene label, wherein the grid-level experience rate refers to the user experience rate of each 3D grid. Through the aggregation of the scene-level user experience rate, the common problems of each scene can be identified, and the performance bottleneck of the network in the specific scene can be revealed; through the aggregation and comparison of the scenes, the network optimization personnel can make more refined optimization strategies to meet the needs of users in different scenes.

[0107] The K-means or DBSCAN algorithm is used to group the grid-level user experience rates of the same scene label, including: ① selecting the feature vectors of each 3D grid: grid-level user experience rate, interference level, and user number, and the interference level includes but is not limited to RSRP (Reference Signal Receiving Power), SINR (Signal to Interference plus Noise Ratio). ② Set the initial parameters of clustering, such as the number of clusters K or the minimum cluster density (for DBSCAN). ③ Execute the clustering algorithm and output the center point and range of each cluster group. ④ Scene aggregation formula: aggregate the clustering results within the same scene, and according to the formula: Calculate the scene-level user experience rate and the distribution characteristics of the scene-level user experience rate, where N is the total number of 3D grids in the scene, Grid Rate i is the user experience rate of the i-th 3D grid.

[0108] In S203, the problem area is visualized and presented, where the problem area refers to the area where the grid-level user experience rate or the scene-level user experience rate is lower than the preset threshold.

[0109] In this embodiment, the geographic presentation of user experience rate refers to displaying the calculation results of user experience rate on the map in an intuitive way, where different ranges of user experience rate can be used to represent the impact of network speed on user experience, so different ranges of user experience rate are displayed with different identifiers. Taking N=2 as an example, the geographic presentation of user experience rate is as shown in Figure 7 , where the user experience rate of each 2D grid is the grid capability in Figure 7 , the user experience rate of the 2D grid lower than 10 Mbps means that the network experience is very poor, which may cause slow webpage loading, frequent video buffering and high online game delay; the user experience rate of the 2D grid in the range of [10, 20) Mbps means that the user experience starts to improve, but there may still be problems such as buffering and slow loading speed, especially when multiple devices are used simultaneously; the user experience rate of the 2D grid in the range of [20, 50) Mbps means that it is sufficient to support streaming services (such as 1080p video) and can perform basic multi-device operations, but problems may occur during peak hours; the user experience rate of the 2D grid in the range of [50, 100) Mbps means that the user's experience is relatively good, which can support 4K video streaming, online games and other bandwidth-intensive activities; the user experience rate of the 2D grid higher than 100 Mbps means that the network connection is very excellent, which can support multiple devices to be used simultaneously, smooth high-resolution video playback, large file download and other highly bandwidth-dependent applications.

[0110] like Figure 8 As shown, this embodiment can also use color coding (e.g., red for low speed and green for high speed) to visualize the user experience rate distribution of each 3D grid, enabling network optimization personnel to quickly identify areas with low speeds. By overlaying the user experience rates of each 3D grid on the electronic map, the spatial distribution of network performance and problem areas are clearly visible, significantly improving optimization efficiency; it can also be combined with actual terrain and building distribution to assist in assessing network performance issues in coverage blind spots and complex scenarios.

[0111] S204, based on anomaly detection algorithms and / or path loss and interference analysis models, performs root cause localization of problem areas, and optimizes the user experience rate of problem areas based on the root cause localization results.

[0112] In this embodiment, the optimization of user experience rate includes root cause analysis and the development of targeted optimization solutions. Root cause analysis aims to find the underlying reasons for low user experience rates in grids through in-depth analysis. The optimization process combines various algorithms, such as multiple linear regression, decision trees, and anomaly detection, to investigate key factors affecting network performance. Common root causes include severe signal interference, uneven resource allocation, and poor coverage. Root cause analysis, through multidimensional analysis of MR data (such as signal strength, signal-to-noise ratio, and resource block occupancy), can identify the main reasons for rate degradation. Optimization solutions include adjusting antenna orientation, optimizing resource scheduling, or increasing network coverage.

[0113] Multidimensional analysis of MR data specifically includes combining multiple indicators of MR data, such as RSRP, SINR, CQI, and Resource Block Utilization (RB Utilization), and analyzing the impact of each indicator on user experience rate using a multidimensional regression model. For example, using a multiple linear regression or decision tree model, i.e., the formula User Experience Rate = β0 + β1 × RSRP + β2 × SINR + β3 × CQI, the indicator with the greatest impact on user experience rate can be identified, where β... i The regression coefficients for each indicator are determined through model training.

[0114] In addition to combining multiple indicators from MR data and analyzing the impact of each indicator on user experience rate through a multidimensional regression model, this embodiment can also perform root cause localization of problem areas based on anomaly detection algorithms and / or path loss and interference analysis models, so as to optimize the user experience rate of problem areas based on the root cause localization results.

[0115] Based on the abnormality detection algorithm, the root cause of the problem area is located, specifically including: using statistical methods (such as Z-score or Tukey's fences) or machine learning abnormality detection algorithms (such as Isolation Forest, LOF (Local Outliers Factor)) to identify outliers in the MR data, and analyze the impact of outliers on user experience rate.

[0116] Based on the path loss and interference analysis model, the root cause of the problem area is located, specifically including: according to the wireless propagation model, the path loss of each 3D grid is calculated, and combined with the interference analysis to locate the possible interference source, wherein the wireless propagation model includes but is not limited to: COST231-Hata model and Okumura-Hata model. Combined with the interference analysis to locate the possible interference source, specifically including: by analyzing the RSRP, SINR distribution of the cells in and around the grid, the possible interference cells or user equipment are found out.

[0117] Taking N=2 and the target areas A lake and B scenic area surrounding city as an example, after evaluating the grid-level user experience rate and calculating the scene-level user experience rate, there are 321 scenes with risk of user experience rate in A lake and B scenic area surrounding city, and the problem scene accounts for 4.4%. Through TOP scene screening, that is, sorting the user experience rate insufficient scenes, the TOP k scenes in the sorting result are selected as the target scenes, and a scene is the target scene in the user experience rate insufficient scene. The scene-level user experience rate of a scene is geographically presented and the root cause is located, and it is determined that a scene is mainly a weak coverage problem, and the problem is characterized by 5G downlink rate capability less than 100Mbps accounting for 21%, as shown in Figure 9 .

[0118] The scene-level user experience rate of a scene is geographically presented and the root cause is located, specifically including: the problem area under a scene can be directly presented by geographically presenting all the grid-level user experience rates under a scene, and the problem area is determined to have 5G off-network problem, and the root cause of the problem area is located. The average RSRP, SINR, downlink rate and comprehensive coverage rate of the problem area 4G network are-66dBm, 39db, 35Mbps and 100%, respectively, and the average RSRP, SINR, downlink rate and comprehensive coverage rate of the 5G network are-98dBm, 1.6db, 40Mbps and 35.71%, respectively. The 4G coverage is good, and the 5G is off-network inside, and it is determined that the problem area has 4G room without 5G room. The optimization scheme is to: share the room distribution common mode, open NR2.1 room to perform 5G coverage blind. After opening NR2.1G, the network coverage and user experience are greatly improved, as shown in Figure 10 ​Figure 11 As shown in the optimized coverage and rate, the 5G network coverage is from partial outage to continuous, the overall RSRP value (i.e. Figure 10 and Figure 11 NR SS-RSRP in the -98dBm is improved to about -70dBm, the downlink rate is improved from 32Mbps to 113Mbps, and the uplink rate is improved from 5Mbps to 66Mbps, and the user experience is fully guaranteed.

[0119] The analysis of network performance data, the calculation of user experience rate, the geographical presentation of user experience rate, the aggregation of scene-level user experience rate and the optimization of user experience rate are closely linked, and constitute a complete closed-loop optimization process. Through the fine evaluation of network quality and rate capability, as well as the geographical presentation and scene-level aggregation, the root problem of low-rate area can be accurately located, the optimization process is gradually promoted, the network performance is continuously improved, not only the coverage quality of the network can be improved, but also the actual experience of users in different scenes can be significantly improved, which provides strong technical support for network optimization and user experience guarantee.

[0120] The optimization method of user experience rate provided by the embodiment can accurately locate the root problem of low-rate area through the fine evaluation of network quality and rate capability, as well as the geographical presentation and scene-level aggregation, the optimization process is gradually promoted, the network performance is continuously improved, not only the coverage quality of the network can be improved, but also the actual experience of users in different scenes can be significantly improved, which provides strong technical support for network optimization and user experience guarantee, realizes accurate evaluation of user experience rate, reduces the grid granularity of network performance data, improves the geographical presentation effect of network performance data, expands the application of user experience and customer satisfaction evaluation in different business scenarios, and improves the evaluation accuracy, efficiency and applicability of user experience rate.

[0121] Embodiment 3:

[0122] As shown in Figure 13 The embodiment provides an evaluation device of user experience rate, which comprises a collection module 31, a grid module 32 and an evaluation module 33. The collection module 31 is used for collecting measurement report MR data of a target area in a preset time period, wherein the MR data comprises position information and field information of a plurality of users. The grid module 32 is connected with the collection module 31 and is used for performing N-dimensional gridding on the MR data of the target area to obtain field information of each N-dimensional grid, wherein N comprises one of 2, 3 and 4. The evaluation module 33 is connected with the grid module 32 and is used for evaluating user experience rate of each N-dimensional grid based on the field information of each N-dimensional grid.

[0123] Specifically, the gridding module 32 comprises a gridding unit 321, a user positioning unit 322 and a backfill unit 323, the gridding unit 321 is configured to obtain geographical position information of a target area, and perform gridding and grid layering processing on the target area to obtain geographical position information of each N-dimensional grid, wherein the N-dimensional grid is used to represent grids at different levels, the user positioning unit 322 is configured to perform user position positioning on MR data to obtain N-dimensional position information of a plurality of users, and the backfill unit 323 is configured to backfill field information of the plurality of users based on the geographical position information of each N-dimensional grid, the N-dimensional position information of the plurality of users and a clustering algorithm to obtain field information of each N-dimensional grid, wherein the clustering algorithm comprises any one of the following: a K-means algorithm, a density-based spatial clustering with noise application DBSCAN algorithm and a Mean Shift algorithm.

[0124] Specifically, the user positioning unit 322 comprises an aggregation subunit, an estimation subunit, a ray intersection subunit and a calibration subunit, the aggregation subunit is configured to perform MR aggregation on MR data to obtain a plurality of MR groups, the estimation subunit is configured to perform angle of arrival estimation AOA on the MR data to obtain angle of arrival information of a plurality of users, the ray intersection subunit is configured to perform ray intersection based on a topological relationship on the angle of arrival information of the same MR group to obtain initial N-dimensional position information of the plurality of users, and the calibration subunit is configured to perform MR calibration on the initial N-dimensional position information of the plurality of users based on a ray tracing model to obtain N-dimensional position information of the plurality of users.

[0125] Specifically, the evaluation module 33 comprises a first calculation unit 331 and a second calculation unit 332, the first calculation unit 331 is configured to perform mean regression calculation on the field information of each N-dimensional grid to obtain regression MCS field information, regression RANK field information and regression RB field information of each N-dimensional grid, and the second calculation unit 332 is configured to calculate user experience rate of each N-dimensional grid based on the regression MCS field information, the regression RANK field information and the regression RB field information of each N-dimensional grid.

[0126] Specifically, the first calculation unit 331 comprises a first calculation subunit configured to perform mean regression calculation on the field information of each N-dimensional grid according to formula (1):

[0127] Yt = a + bXt + et (1), wherein Yt represents regression field information at time t, Xt represents field information at time t-1, a and b represent constants, and et represents a mean regression error term at time t.

[0128] Specifically, the second calculation unit 332 comprises a second calculation subunit configured to calculate user experience rate of each N-dimensional grid according to formula (2):

[0129] Grid Rate = AVERAGE (MCS) x AVERAGE (Rank) x AVERAGE (RB Num) (2),

[0130] wherein, Grid Rate represents the user experience rate of each N-dimensional grid, MCS represents the regression MCS field information of each N-dimensional grid, Rank represents the regression RANK field information of each N-dimensional grid, and RB Num represents the regression RB field information of each N-dimensional grid.

[0131] It can be understood that the above-mentioned user experience rate evaluation device executes the user experience rate evaluation method corresponding to the above-mentioned embodiment 1, and thus the beneficial effects that can be achieved can refer to the beneficial effects of the schemes corresponding to the user experience rate evaluation method of the above-mentioned embodiment 1, which will not be repeated here.

[0132] Embodiment 4:

[0133] As shown in Figure 14 The present embodiment further provides a user experience rate optimization device, which comprises an acquisition module 41, a clustering module 42, a visualization module 43, and a root cause positioning module 44. The acquisition module 41 is configured to acquire grid-level user experience rates, wherein the grid-level user experience rate refers to the user experience rate of each N-dimensional grid. The clustering module 42 is connected to the acquisition module 41 and is configured to cluster all grid-level user experience rates under the same scene and calculate a scene-level user experience rate. The visualization module 43 is connected to the clustering module 42 and is configured to visualize a problem area in which the grid-level user experience rate or the scene-level user experience rate is lower than a preset threshold. The root cause positioning module 44 is connected to the visualization module 43 and is configured to perform root cause positioning on the problem area based on an anomaly detection algorithm and / or a path loss and interference analysis model, so as to optimize the user experience rate of the problem area based on the root cause positioning result.

[0134] It can be understood that the above-mentioned user experience rate optimization device executes the user experience rate optimization method corresponding to the above-mentioned embodiment 2, and thus the beneficial effects that can be achieved can refer to the beneficial effects of the schemes corresponding to the user experience rate optimization method of the above-mentioned embodiment 2, which will not be repeated here.

[0135] The present embodiment further provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to implement the user experience rate evaluation method in the above-mentioned embodiment 1 or the user experience rate optimization method in the above-mentioned embodiment 2.

[0136] Embodiment 6:

[0137] The embodiment also provides a computer readable storage medium, which stores a computer program. The computer program is executed by a processor to implement the method for evaluating user experience rate in the above embodiment 1 or the method for optimizing user experience rate in the above embodiment 2.

[0138] It can be understood that the above implementation is only an exemplary implementation for illustrating the principles of the present application, and the present application is not limited thereto. Various modifications and improvements can be made by those of ordinary skill in the art without departing from the spirit and principle of the present application, and these modifications and improvements are also considered to be within the scope of protection of the present application.

Claims

1. A method of evaluating user experience rate, characterized by, The method comprises the following steps: Collecting measurement report (MR) data of a target area in a preset time period, wherein the MR data comprises field information of a plurality of users; N-dimensional gridding the MR data of the target area to obtain field information of each N-dimensional grid, wherein N comprises one of 2, 3, and 4, and the field information comprises modulation and coding (MCS) field information, rank (RANK) field information, and resource block (RB) field information; Performing mean regression calculation on the field information of each N-dimensional grid to obtain regression MCS field information, regression RANK field information, and regression RB field information of each N-dimensional grid; Based on the regression MCS field information, the regression RANK field information, and the regression RB field information of each N-dimensional grid, calculating a user experience rate of each N-dimensional grid.

2. The method of evaluating user experience rates according to claim 1, characterized in that, The N-dimensional gridding of the MR data of the target area to obtain the field information of each N-dimensional grid comprises the following steps: Obtaining geographical position information of the target area, and performing gridding and grid layering processing on the target area to obtain geographical position information of each N-dimensional grid, wherein the N-dimensional grid is used to represent grids at different levels; Positioning the users in the MR data to obtain N-dimensional position information of the plurality of users; Based on the geographical position information of each N-dimensional grid, the N-dimensional position information of the plurality of users, and a clustering algorithm, backfilling the field information of the plurality of users to obtain the field information of each N-dimensional grid, wherein the clustering algorithm comprises any one of the following: a K-means algorithm, a density-based spatial clustering with noise (DBSCAN) algorithm, and a Mean Shift algorithm.

3. The method of evaluating user experience rates according to claim 2, wherein, The positioning of the users in the MR data to obtain the N-dimensional position information of the plurality of users comprises the following steps: Performing MR aggregation on the MR data to obtain a plurality of MR groups; Performing angle of arrival (AOA) estimation on the MR data to obtain arrival angle information of the plurality of users; Based on a topological relationship, performing ray intersection on the arrival angle information of the same MR group to obtain initial N-dimensional position information of the plurality of users; Based on a ray tracing model, performing MR calibration on the initial N-dimensional position information of the plurality of users to obtain the N-dimensional position information of the plurality of users.

4. The method of claim 1, wherein, The mean regression calculation on the field information of each N-dimensional grid to obtain mean regression field information of each N-dimensional grid comprises the following steps: According to formula (1), performing mean regression calculation on the field information of each N-dimensional grid: Yt = α + βXt + εt (1), wherein Yt represents regression field information at time t, Xt represents field information at time t-1, α and β represent constants, and εt represents a mean regression error term at time t.

5. The method of claim 1, wherein, The calculation of the user experience rate of each N-dimensional grid based on the regression MCS field information, the regression RANK field information, and the regression RB field information of each N-dimensional grid comprises the following steps: According to formula (2), calculating the user experience rate of each N-dimensional grid: Grid Rate=AVERAGE(MCS)×AVERAGE(Rank)×AVERAGE(RBNum)(2), wherein, Grid Rate represents the user experience rate of each N-dimensional grid, MCS represents the regression MCS field information of each N-dimensional grid, Rank represents the regression RANK field information of each N-dimensional grid, and RB Num represents the regression RB field information of each N-dimensional grid.

6. A method for optimizing user experience rate, characterized in that, The method comprises the following steps: obtaining the grid-level user experience rate evaluated by the evaluation method of the user experience rate in any one of claims 1-5, wherein the grid-level user experience rate refers to the user experience rate of each N-dimensional grid; clustering all grid-level user experience rates in the same scene and calculating the scene-level user experience rate, wherein the scene-level user experience rate refers to the average of all grid-level user experience rates in each scene; visually presenting the problem area, wherein the problem area refers to an area where the grid-level user experience rate or the scene-level user experience rate is lower than a preset threshold; based on an anomaly detection algorithm and / or a path loss and interference analysis model, performing root cause positioning on the problem area to optimize the user experience rate of the problem area based on the root cause positioning result.

7. An apparatus for evaluating a user experienced rate, characterized by The method comprises the following steps: a collection module, a gridding module and an evaluation module, the collection module is configured to collect measurement report (MR) data of a target area in a preset time period, wherein the MR data comprises position information and field information of a plurality of users, the gridding module is connected to the collection module and is configured to perform N-dimensional gridding on the MR data of the target area to obtain field information of each N-dimensional grid, wherein N comprises one of 2, 3 and 4, and the field information comprises modulation and coding (MCS) field information, rank (RANK) field information and resource block (RB) field information, the evaluation module is connected to the gridding module and is configured to perform mean regression calculation on the field information of each N-dimensional grid to obtain regression MCS field information, regression RANK field information and regression RB field information of each N-dimensional grid, and is further configured to calculate the user experience rate of each N-dimensional grid based on the regression MCS field information, the regression RANK field information and the regression RB field information of each N-dimensional grid.

8. A device for optimizing user experience rate, characterized in that, The method comprises the following steps: an acquisition module, a clustering module, a visualization module and a root cause positioning module, the acquisition module is configured to obtain the grid-level user experience rate evaluated by the user experience rate evaluation device in claim 7, wherein the grid-level user experience rate refers to the user experience rate of each N-dimensional grid, the clustering module is connected to the acquisition module and is configured to cluster all grid-level user experience rates in the same scene and calculate the scene-level user experience rate, the visualization module is connected to the clustering module and is configured to visually present a problem area where the grid-level user experience rate or the scene-level user experience rate is lower than a preset threshold, the root cause positioning module is connected to the visualization module and is configured to perform root cause positioning on the problem area based on an anomaly detection algorithm and / or a path loss and interference analysis model to optimize the user experience rate of the problem area based on the root cause positioning result.

9. An electronic device, comprising: A computer program product comprising a memory having stored therein a computer program and a processor arranged to execute the computer program to implement a method of evaluating user experience rate as claimed in any one of claims 1 to 5, or a method of optimizing user experience rate as claimed in claim 6.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program product, when executed by a processor, implements a method of evaluating user experience rate as claimed in any one of claims 1 to 5, or a method of optimizing user experience rate as claimed in claim 6.

Images