A network evaluation method based on MR big data analysis

Abstract

The present invention discloses a network evaluation method based on MR big data analysis, which includes the following steps: collecting MR information data; rasterizing the MR sampling point information; performing value scoring on the raster: obtaining the optimal solution coefficient for the raster by using geographical and clustering algorithms, and determining the total raster value score based on multi-dimensional scoring of the raster in terms of coverage and value; establishing an MR fingerprint library; matching the MR information with the fingerprint library to distinguish between indoor and outdoor. The present invention performs value information scoring on the MR information raster, rates the traffic according to the indoor and outdoor traffic prediction information, and combines the scene importance rating for comprehensive rating and ranking. According to the above rating and ranking, high-level areas can be preferentially addressed, and finally network coverage and value evaluation can be completed. The data of the present invention is adaptable to various scenarios, mines weak coverage and high-value areas, can quickly improve the network perception of high-value users, and effectively improve network quality and user perception.

Description

Technical Field

[0001] The present invention relates to the technical field of network coverage quality assessment. Specifically, it relates to a network assessment method based on MR big data analysis. Background Art

[0002] With the continuous development of communication technologies, mobile networks have penetrated into people's daily lives, and the requirements for network performance and quality are getting higher and higher. As an important means to ensure network performance, the accuracy and effectiveness of network assessment are directly related to user experience and the business development of operators. However, traditional network assessment methods often rely on small sample data and local observations, making it difficult to comprehensively and deeply reflect the true state of the network.

[0003] The growth of network scale and complexity, the increasing requirements of users for network quality, the richness and value of MR data, the development of big data technologies, and cost - benefit considerations. These factors have jointly promoted the research and application of network assessment methods based on MR big data analysis.

[0004] Scene - specific and systematic network assessment and optimization work have become increasingly important. How to quantify user perception and quickly locate and identify scene problems has become an urgent problem to be solved. Summary of the Invention

[0005] In view of the above - mentioned technical problems in the related art, the present invention proposes a network assessment method based on MR big data analysis, which can overcome the above - mentioned deficiencies of the prior art.

[0006] To achieve the above - mentioned technical objectives, the technical solution of the present invention is realized as follows:

[0007] A network assessment method based on MR big data analysis includes the following steps:

[0008] S1. Collect MR information data;

[0009] S2. Gridify the MR sampling point information;

[0010] S3. Score the value of the grid: For the grid, use geographical and clustering algorithms to obtain the optimal solution coefficient, and perform multi - dimensional scoring on the grid based on coverage and value to determine the total value score of the grid;

[0011] The specific steps of step S3 are as follows:

[0012] In S31, select the top 20% of the grids with a comprehensive score above -100 dBm in the selected area as the training sample data, use model training and machine learning with all data, calculate the weight ratio relationship of the number of users, traffic, number of service connections, and service connection duration through a linear regression function, and determine the optimal solution coefficient; the linear regression function is V = f(X1, X2, X3, X4), where V is the total grid value score, and X1, X2, X3, X4 are the number of users, traffic, number of service connections, and service connection duration respectively; F is a linear function;

[0013] S32 Calculate the total grid value score using the optimal solution coefficient. The calculation formula for the total grid value score is: Total grid value score = [Grid user number + Grid traffic * a 1 +(Grid number of service connections + Grid service connection duration * a 2 ) * a 3 , where a 1 is the optimal solution coefficient of grid traffic, a 2 is the optimal solution coefficient of grid service connection duration, and a 3 is the optimal solution coefficient of grid services;

[0014] S4. Establish an MR fingerprint database;

[0015] S5. Match the MR information with the fingerprint database to distinguish indoor and outdoor;

[0016] S6. Conduct value analysis on indoor covered buildings: The predicted traffic of indoor covered buildings is obtained by assigning respective weight coefficients to MR traffic, surrounding macro-station traffic, and competitor's indoor distribution traffic;

[0017] S7. Conduct value analysis on outdoor covered macro-stations: The outdoor macro-station traffic prediction calculates the base station coverage design distance according to the wireless propagation model and its development model, based on the shared cell model and shared cell traffic, and then corrects the increased traffic by comprehensively considering service coverage and service quality capabilities;

[0018] S8. Determine the comprehensive rating ranking according to the MR grid score obtained in S3 and the traffic predicted in S6 - S7, combined with the importance of the scenario;

[0019] S9. According to the annual or quarterly investment plan, prioritize solving high-level areas to complete network coverage and value assessment.

[0020] Furthermore, the MR information data in step S1 includes cell ID, RSCP, RSRQ, CQI, AOA, transmit power, longitude and latitude, and neighbor cell information.

[0021] Further, in step S2, a 5m * 5m map is used for grid geographicalization, and coverage and service information are assigned. The coverage includes the number of sampling points, field strength, and weak coverage ratio, and the service information includes traffic, the number of users, user connection duration, and the number of service link times.

[0022] Further, in step S4, the signal strength of each cell in the grid is calculated and combined with the trained propagation model, Clutter map, grid, primary cell, and neighbor cell identifiers. Based on the in-network test log, MDT data, and map feature information, a fingerprint database is established for the area. The inverse distance interpolation method is used to construct the fingerprint database algorithm. The inverse distance interpolation formula is:

[0023] where the weight W ij The calculation formula is:

[0024] In the above formula, RSRP j is the j-th element in the RSRP vector received at multiple different base stations at the interpolation point. K is the number of neighboring sampling points selected for interpolation. W ij is the weight of the i-th sampling point acting on the j-th RSRP of the point to be interpolated. d ij is the distance from the i-th sampling point to the point to be interpolated;

[0025] The specific construction process is as follows:

[0026] S41 Determine the area range where the fingerprint database needs to be constructed. According to the positioning accuracy requirements, divide an appropriate grid size and determine each grid point as a potential unknown location;

[0027] S42 For each grid point, use the inverse distance weight function to calculate the weight of each known data point for the unknown point. According to the selected weight function, calculate the weight of each known point for the unknown point;

[0028] S43 According to the attribute value of each known data point and its corresponding weight, calculate the attribute value of the unknown point by weighted average;

[0029] S44 Store the estimated attribute values of all grid points together with their location information to construct a complete fingerprint database.

[0030] Further, in step S5, specifically: match the level feature, motion feature, cell feature, handover feature, neighbor cell feature of MR and the feature of the ground objects in the 5-meter precision map with the fingerprint database, determine whether the MR is located in a building to distinguish indoor and outdoor, and finally assign the MR sampling points to indoor buildings or outdoor areas;

[0031] The basis for judging and distinguishing between indoor and outdoor is as follows: Wear and tear estimation is carried out based on MR cell type + TA / PD; Behavioral characteristics are judged based on horizontal positioning results + MR time stamp; Cell handover analysis + level fluctuation analysis are used to analyze level characteristics; Building attribution judgment is made based on horizontal positioning results + electronic map;

[0032] The process of allocating MR sampling points to indoor buildings or outdoor areas is as follows:

[0033] S51 Determine the signal strength of the serving cell and neighboring cells;

[0034] S52 In the fingerprint grid set to which the MR sampling point belongs to the serving cell, find the grid with the feature information closest to the feature information contained in the current MR, select the RP with the smallest Euclidean distance from the RSRP in the sampling point as the nearest RP, and calculate the coordinate weight of the nearest RP; The Euclidean distance formula of RSRP in the sampling point is:

[0035]

[0036] where ED i，* represents the Euclidean distance between the RSRP fingerprint information of the grid in the i-th fingerprint database and the RSRP in the MR. and are the RSRP values of the grid in the i-th fingerprint database and the RSRP in the MR, where the superscript u represents that the RSRP comes from the u-th base station BS. M is the number of base stations;

[0037] S53 Use the weighted coordinates of the nearest RP as the location of this MR.

[0038] Furthermore, the indoor coverage building traffic prediction formula in step S6 is: where α i is the traffic of MR sampling points at the coverage cell level of the building; β i is the number of MR sampling points at the coverage cell level of the building. γ i is the number of MR sampling points predicted for a single cell of the building using big data positioning output; θ is 1 - the weak coverage ratio of the predicted building, δ = the average traffic of the nearest macro station around the building * the ratio of the current network indoor to outdoor traffic of the scene to which the building belongs, ε = the indoor distribution traffic of competitors * the market share of this enterprise / the market share of competitors * the DOU of this enterprise / the DOU of competitors, K 1 、K 2 、K 3 are weight coefficients, K 1 、K 2 、K 3 are 70%, 15%, and 15% respectively.

[0039] Furthermore, the specific steps of step S7 are as follows:

[0040] S71 Determine the site location for the new base station according to the output information of weak MR coverage;

[0041] S72 Design the coverage distance a of the new base station according to the formulas of the Okumura-Hata model and the COST231-Hata model;

[0042] When the base station transmission frequency is 150 - 1500 MHz, L = 46.3 + 33.9lg f - 13.82lgh b -α(h m )+(44.9 - 6.551lgh b )lgd + C m ;

[0043] When the base station transmission frequency is 1500 - 2000 MHz, L = 69.55 + 26.16lg f - 13.82lgh b -a(h m )+(44.9 - 6.551lgh b )lgd;

[0044] Where: L is the median value of the basic propagation loss; f is the base station frequency; h b 、h m are the effective heights of the base station and mobile station antennas; a(h m ) is the mobile station antenna height correction factor; C m is the urban correction factor;

[0045] The designed coverage distance a = MIN((transmitting antenna height / SIN(RADIANS(down tilt angle))) / 1000, propagation radius * COS(RADIANS(down tilt angle))

[0046] propagation radius (km)) = 10^((maximum path loss - L) / (44.9 - 6.55 * LOG(transmitting antenna height)));

[0047] S73 Determine the cells shared by the existing network, extract the traffic q of the shared cells i , and calculate the station distance b through the longitude and latitude of the site of the shared cell,

[0048] b=IFERROR(6370*ACOS(COS(3.141592654 / 2-D1*3.141592654 / 180)*COS(3.141592654 / 2-B1*3.141592654 / 180)+SIN(3.141592654 / 2-D1*3.141592654 / 180)*SIN(3.141592654 / 2-B1*3.141592654 / 180)*COS(C1*3.141592654 / 180-A1*3.141592654 / 180))*1000,0)

[0049] A1 and B1 are the longitude and latitude of the newly built base station; C1 and D1 are the longitude and latitude of the existing network sharing cell;

[0050] S74 calculates the cell sharing coefficient γ according to the sharing coefficient model.

[0051]

[0052] S75 Among them, Q is the traffic volume that is calculated based on the shared cell traffic volume and the coefficient for demand capacity growth. Improve traffic for business quality capabilities,

[0053] ω is the traffic expansion for business coverage capability.

[0054] Furthermore, step S8 specifically: rates the total value score of the grid, predicted traffic, and importance of the scene, with the TOP30% being determined as level A and the Bottom70% being determined as level B. The comprehensive rating is determined to be 1-4 levels based on the combination of A and B levels of the three dimensions of total value score of the grid, predicted traffic, and importance of the scene.

[0055] Beneficial effects of the present invention: The present invention scores the value information of the MR information grid, rates the traffic according to the indoor and outdoor traffic prediction information, and combines the scene importance rating to perform comprehensive rating sorting. According to the above rating sorting, high-level areas can be prioritized to finally complete network coverage and value assessment. The data of the present invention is comprehensive, multi-dimensional, and adaptable to a variety of scenarios. It mines areas with weak coverage and high value, which can quickly improve the network perception of high-value users, concentrate advantageous resources to quickly and accurately create competitive advantages in key areas, and effectively improve network quality and user perception. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0057] Figure 1 is a flow chart of a network evaluation method based on MR big data analysis according to an embodiment of the present invention;

[0058] Figure 2 is a flow chart of indoor coverage value analysis of a network evaluation method based on MR big data analysis according to an embodiment of the present invention;

[0059] Figure 3 is a flow chart of outdoor macro base station value analysis according to a network evaluation method based on MR big data analysis according to an embodiment of the present invention;

[0060] Figure 4 It is a flowchart of comprehensive rating and ranking of network coverage and value of a network evaluation method based on MR big data analysis according to an embodiment of the present invention.

[0061] Figure 5 It is a diagram for distinguishing indoor and outdoor structures according to the network evaluation method based on MR big data analysis described in an embodiment of the present invention. DETAILED DESCRIPTION

[0062] The following will be combined with the accompanying drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field belong to the scope of protection of the present invention.

[0063] like Figure 1 As shown, a network evaluation method based on MR big data analysis according to an embodiment of the present invention includes the following steps:

[0064] S1. Collect MR information data.

[0065] Collect MR information data for one month, including cell ID, RSCP, RSRQ, CQI, AOA, transmit power, longitude and latitude, neighboring cell information, etc.

[0066] S2. Rasterize the MR sampling point information.

[0067] The step S2 uses a 5m*5m map to geo-map the grid, and assigns coverage (number of sampling points, field strength, weak coverage ratio) and service information (traffic, number of users, user connection duration, number of service links).

[0068] S3. Assess the value of the grid.

[0069] The step S3 uses geographicization and clustering algorithms to obtain the optimal solution coefficient, and determines the total grid value score based on the grid multi-dimensional score of coverage and value. The total grid value score of the proposed building or base station (new station) is predicted using the existing network data (built stations).

[0070] Use the existing network data of a certain area to identify the value of the entire network. There are 50,000 grids in the University Town area within the Third Ring Road of XX City. According to the 20 / 80 principle, the top 10,000 grids are considered to be high-value grids.

[0071] The top 10,000 grids are judged based on four attributes: ① existing revenue ② valuable brand ③ geographical factors ④ user activity.

[0072] The grid score is related to the number of users, traffic, business connection times, and business connection duration. The goal is to find a linear regression relationship between the four factors so that the top 10,000 grids analyzed above have the highest score.

[0073] The specific implementation process is as follows:

[0074] S31. Use the top 10,000 rasters of relevant attributes as training data, so that the rasters with the highest comprehensive scores have as many top 10,000 feature attributes as possible.

[0075] (1) The total value score of the grid is evaluated based on its relevant attributes, including four main factors: traffic, service connection duration, number of users, and number of service connections;

[0076] (2) Let V = f(X1, X2, X3, X4), where V is the total value score of the grid, X1, X2, X3, and X4 are the number of users, traffic, number of service connections, and service connection duration, respectively. F is a linear function, that is, a linear combination of X1, X2, X3, and X4;

[0077] (3) Taking the grids above -100 dBm in a certain area as samples, model training and full data machine learning are used to calculate the proportional relationship of each attribute through linear regression to determine the optimal solution.

[0078] S32. The optimal solution coefficients are 0.3486, 0.068, and 0.0007, which are the weight coefficients of the four factors. The goal here is to sort the top 10,000 communities that contain as many characteristic attributes as possible according to the comprehensive score.

[0079] S33. Grid total value score = [number of grid users + grid traffic*0.3486+(number of grid service connections + grid service connection duration*0.068)*0.007].

[0080] The purpose of this step is to use the existing network data to study a linear relationship of the four factors so that the grids with confirmed high value can obtain the highest score, and then the results are applied to the evaluation of new buildings or base stations.

[0081] S4. Establish MR fingerprint database.

[0082] The step S4 calculates the signal strength of each cell in the grid and considers the trained propagation model, Clutter map, grid, main cell, neighbor cell identification and other features, and establishes a fingerprint library for the area based on the existing network test log, MDT data, and map feature information (indoor and outdoor borders, road surfaces, etc.).

[0083] The fingerprint library algorithm is constructed using the inverse distance interpolation method.

[0084] Inverse distance interpolation formula:

[0085] The weight W ij The calculation formula is:

[0086] In the above formula, RSRP j is the jth element in the RSRP vector received at multiple different base stations at the interpolation point, K is the number of adjacent sampling points selected during interpolation, and W ij is the weight of the i-th sampling point acting on the j-th RSRP of the interpolation point. ij is the distance from the i-th sampling point to the point to be interpolated.

[0087] Specific steps:

[0088] (1) Clearly define the area where the fingerprint library needs to be built. According to the positioning accuracy requirements, divide the grid into appropriate sizes and identify each grid point as a potential unknown location.

[0089] (2) For each grid point, use the inverse distance weight function to calculate the weight of each known data point to the unknown point. The weight is inversely proportional to the distance, that is, the closer the point is, the greater the weight is. Then, based on the selected weight function, calculate the weight of each known point to the unknown point.

[0090] (3) According to the attribute value (signal strength value) of each known data point and its corresponding weight, the attribute value of the unknown point is calculated by weighted average.

[0091] (4) The estimated attribute values ​​of all grid points are stored together with their location information to construct a complete fingerprint library.

[0092] S5. Match the MR information with the fingerprint database to distinguish between indoor and outdoor.

[0093] The step S5 matches the MR's level features, motion features, cell features, switching features, neighboring area features and the features of the 5-meter precision map with the fingerprint library, locates the MR to the building, distinguishes between indoor and outdoor areas through multi-dimensional judgment, and finally allocates the MR sampling points to indoor buildings or outdoor areas.

[0094] The basic principle of matching MR information with fingerprint database to distinguish indoor and outdoor is: fingerprint represents the data distribution of various features including signal and field strength in time and space at each geographical location. In theory, this data distribution is unique. With the unique matching of geographical location and location fingerprint, the above matching algorithm can be used to solve the location estimation under the known distribution of each feature data of the current location, such as Figure 5 shown.

[0095] 1. MR cell type (indoor / macro base station) + TA / PD (delay characteristics) for wear estimation.

[0096] The TA value can indirectly reflect the distance between the UE and the base station. By comparing the TA values ​​of indoor and outdoor UEs, the additional delay of the signal when penetrating a building can be estimated, and then the penetration loss can be calculated. The RSRP of outdoor users is 10 to 20 dB higher than that of indoor users, with a median value of about 15 dB.

[0097] The specific data of delay characteristics can be directly used to evaluate the loss during signal propagation. By comparing the delay differences between different paths (such as indoor to outdoor and outdoor to indoor), the loss estimation can be further verified and corrected.

[0098] 2. Horizontal positioning results + MR time cut to determine behavioral characteristics (moving / stationary).

[0099] MR timestamps can be used to determine behavioral characteristics, especially when distinguishing between indoor and outdoor user behaviors. By setting specific thresholds and judgment logic, user behavior characteristics can be effectively analyzed to make indoor and outdoor judgments.

[0100] The characteristics of indoor and outdoor users are defined by adjusting parameters such as the RSRP threshold, the threshold setting for the number of switching times of the user's adjacent cell reporting data, and the threshold setting for the number of switching times of the user's primary cell.

[0101] 3. Cell switching analysis + level fluctuation analysis Analyze level characteristics (multiple peaks indoors and outdoors).

[0102] The indoor-outdoor primary serving cell handover has obvious mobility characteristics: By analyzing historical MR data and measured data, it is statistically determined whether handover occurs and the number of handovers within a certain period of time, and the differences in the mobility characteristics of the indoor-outdoor primary serving cells are obtained. Outdoor users generally experience handovers within a short period of time and have a relatively large number of handovers.

[0103] 4. Use the horizontal positioning result + electronic map to make a building attribution judgment.

[0104] The MR horizontal positioning result provides information on the strength and distribution of wireless signals in a specific area. By combining it with an electronic map, the location of the signal source can be accurately determined, and the wireless signal coverage within the building can also be obtained, including key parameters such as signal strength and coverage range. Combining this information with the electronic map can visually display the relationship between the signal source and the building, thereby making a building attribution judgment to distinguish between indoor and outdoor.

[0105] Calculate the Euclidean distance between the RSRP in the reported MR information and the RSRP in the fingerprint database. The formula for calculating the Euclidean distance is:

[0106]

[0107] where ED i，* represents the Euclidean distance between the RSRP fingerprint information of the grid in the i-th fingerprint database and the RSRP in the MR. and are the RSRP values of the grid in the i-th fingerprint database and the RSRP value in the MR, where the superscript u indicates that the RSRP comes from the u-th base station BS. M is the number of base stations.

[0108] The basic principle of the Euclidean formula: The wireless positioning method based on the network refers to collecting signals related to positioning transmitted by the mobile station at the base station side, such as reference signal received strength, time advance, etc. The base station side uses this information to obtain the coordinates of the mobile station through a positioning algorithm.

[0109] First, determine the signal strengths of the serving cell and neighboring cells. Then, in the set of fingerprint grids to which the MR belongs to the primary serving cell, find the grid with the feature information closest to the feature information contained in the current MR. Select the RP with the smallest Euclidean distance from the RSRP at the sampling point as the nearest RP, and calculate the coordinate weight of the nearest RP. Finally, use the weighted coordinates of the nearest RP as the location of this MR. S6. Conduct a value analysis on the buildings covered indoors.

[0110] As Figure 2 shown, the traffic prediction of the buildings covered indoors in step S6 is obtained by assigning respective weight coefficients to the MR traffic, the traffic of the surrounding macro base stations, and the traffic of the indoor distribution of friendly operators.

[0111] This step mainly relies on the data processing and analysis capabilities of MR big data to form a weak coverage building library-level weak coverage area, which is combined with the historical and existing traffic of other operators to calculate the predicted traffic.

[0112]

[0113] Among them, α i is the traffic of MR sampling points at the coverage cell level of the building; β i is the number of MR sampling points at the coverage cell level of the building. γ i is the number of MR sampling points predicted for a single cell of the building using big data positioning; θ is 1 - the weak coverage ratio of the predicted building.

[0114] δ = the average traffic of the three nearest macro stations around the building * the ratio of in-network indoor to outdoor traffic of the building's affiliated scenario.

[0115] ε = the indoor distribution traffic of the competitor * the market share of this enterprise / the market share of the competitor * the DOU (Dataflow of usage) of this enterprise / the DOU (Dataflow of usage) of the competitor.

[0116] This enterprise can be China Unicom, and the competitors refer to China Telecom, China Mobile, China Radio and Television, etc. The purpose is to predict and correct the traffic results using the relationship between the market traffic data and market share in a certain area.

[0117] K 1 、K 2 、K 3 are weight coefficients. According to the importance level of traffic prediction, and since MR predicted traffic is the main means of this invention, the weight coefficients of K1, K2, and K3 are given as 70%, 15%, and 15% respectively.

[0118] S7. Conduct a value analysis on the outdoor coverage macro station.

[0119] As Figure 3 shown, the step S7 includes the following sub-steps:

[0120] S7.1 Based on the MR weak coverage output information, that is, determine the site location for building a new station.

[0121] Extract the MR data for one week, map the MR data onto a 5m * 5M map, screen out the located weak coverage grids according to a certain number of MR sampling points and weak coverage ratio, cluster the weak coverage grids to form weak coverage clusters, judge whether to build a macro station or a small station indoor distribution according to the attributes of the weak coverage clusters and the distribution of surrounding existing sites, and then select a station according to the center point of the weak coverage cluster.

[0122] S7.2 Determine the designed coverage distance a of the new base station according to the wireless propagation model Okumura - Hata model and its developed model COST231 - Hata model formula. The COST231 - Hata model is a developed model of the wireless propagation Okumura - Hata model.

[0123] When the base station transmission frequency is 150 - 1500 MHz;

[0124] L = 46.3 + 33.9lg f - 13.82lgh b -a(h m ) + (44.9 - 6.551lgh b )lgd + C m .

[0125] When the base station transmission frequency is 1500 - 2000 MHz;

[0126] L = 69.55 + 26.16lg f - 13.82Lgh b - α (h m ) + (44.9 - 6.551lgh b )lgd.

[0127] Where: L is the median value of the basic propagation loss; f is the base station frequency; h b , h m are the effective heights of the base station and mobile station antennas; a(h m ) is the mobile station antenna height correction factor; C m is the urban correction factor.

[0128] S7.3 Determine the existing network sharing cells and extract the sharing cell traffic q i . Determine the station distance b through the longitude and latitude of the sharing cell site.

[0129] b = IFERROR(6370 * ACOS(COS(3.141592654 / 2 - D1 * 3.141592654 / 180) * COS(3.141592654 / 2 - B1 * 3.141592654 / 180) + SIN(3.141592654 / 2 - D1 * 3.141592654 / 180) * SIN(3.141592654 / 2 - B1 * 3.141592654 / 180) * COS(C1 * 3.141592654 / 180 - A1 * 3.141592654 / 180)) * 1000, 0)

[0130] Where A1 and B1 are the longitude and latitude of the new base station; C1 and D1 are the longitude and latitude of the existing network sharing cell.

[0131] S7.4 Calculate the sharing cell coefficient γ according to the sharing coefficient model.

[0132]

[0133] S7.5

[0134] Among them, Q is the traffic for calculating the growth of demand capacity based on the sharing cell traffic and coefficient.

[0135]

[0136] Among them, ψ is the traffic for improving the service quality ability.

[0137]

[0138] Among them, ω is the traffic for expanding the service coverage ability, and generally takes values according to the predicted growth data of the market and network optimization departments.

[0139] S8. As Figure 4 described above, combine the value analysis results of steps S6 to S7 with step S3, and conduct a comprehensive rating and ranking according to the traffic rating, grid value rating, and importance rating.

[0140] Determine the comprehensive rating and ranking based on the MR grid score obtained in step S3 and the traffic predicted in steps S6 to S7, and then combine the importance of scenarios and complaints.

[0141] The results analyzed in the above steps (phases) are determined as grade A according to the TOP30%, and grade B according to the Bottom70%. The final comprehensive rating is divided into levels 1-4 based on the A and B grade combinations of the three dimensions. Refer to Figure 4 .

[0142] S9. According to the annual or quarterly investment plan, give priority to solving high-level areas, and finally complete network coverage and value assessment.

[0143] To sum up, with the above technical solutions of the present invention, by scoring the value information for the MR information grid, rating the traffic according to the indoor and outdoor traffic prediction information, and combining the scenario importance rating to conduct a comprehensive rating and ranking, according to the above rating and ranking, high-level areas can be given priority to solve, and finally network coverage and value assessment can be completed. The data of the present invention is comprehensive, multi-dimensional, adaptable to various scenarios, can excavate weak coverage and high-value areas, can quickly improve the network perception of high-value users, intensively utilize superior resources to quickly and accurately create competitive advantages in key areas, and effectively improve network quality and user perception.

[0144] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A network evaluation method based on MR big data analysis, characterized in that: The steps include: S1. Collect MR information data; S2. rasterizing the MR sampling point information; S3. Value scoring of the grid: using geographicization and clustering algorithms to obtain the optimal solution coefficient for the grid, and performing multi-dimensional scoring based on the coverage and value of the grid to determine the total value score of the grid; The specific steps of step S3 are as follows: S31 selects the top 20% of the grids with a comprehensive score of -100dBm or above in the region as training sample data, uses model training and full data machine learning, calculates the weighted proportional relationship between the number of users, traffic, number of business connections, and business connection duration through a linear regression function, and determines the optimal solution coefficient; the linear regression function is V=f(X1, X2, X3, X4), where V is the total value score of the grid, X1, X2, X3, and X4 are the number of users, traffic, number of business connections, and business connection duration respectively; and F is a linear function; S32 uses the optimal solution coefficient to calculate the grid total value score. The calculation formula of the grid total value score is: grid total value score = [number of grid users + grid traffic*a1+(number of grid service connections + grid service connection duration*a2)*a3], where a1 is the optimal solution coefficient of grid traffic, a2 is the optimal solution coefficient of grid service connection duration, and a3 is the optimal solution coefficient of grid service; S4. Establish MR fingerprint database; S5. Match the MR information with the fingerprint database to distinguish between indoor and outdoor; S6. Perform value analysis on indoor coverage buildings: The predicted traffic of indoor coverage buildings is obtained by assigning weight coefficients to MR traffic, surrounding macro base station traffic, and indoor distribution traffic of friendly companies; S7. Value analysis of outdoor coverage macro base stations: The outdoor macro base station traffic forecast is based on the wireless propagation model and its development model to calculate the base station coverage design distance, based on the shared cell model and shared cell traffic, and then the service coverage and service quality capabilities are combined to correct the growth traffic; S8. Determine the comprehensive rating ranking based on the MR grid score obtained in S3 and the traffic predicted by S6-S7, combined with the importance of the scene; S9. In accordance with this year’s or quarterly investment plan, priority will be given to completing network coverage and value assessment in high-level areas.

2. The network evaluation method based on MR big data analysis according to claim 1 is characterized in that: The MR information data in step S1 includes cell ID, RSCP, RSRQ, CQI, AOA, transmit power, longitude and latitude, and neighboring cell information.

3. The network evaluation method based on MR big data analysis according to claim 1 is characterized in that: The step S2 uses a 5m*5m map to geo-grid and assign coverage and service information. The coverage includes the number of sampling points, field strength, and weak coverage ratio. The service information includes traffic, number of users, user connection duration, and number of service links.

4. The network evaluation method based on MR big data analysis according to claim 1 is characterized in that: In step S4, the signal strength of each cell in the grid is calculated and combined with the trained propagation model, Clutter map, grid, primary cell, and neighbor cell identifiers. A fingerprint library is established for the area based on the existing network test log, MDT data, and map feature information. The fingerprint library algorithm is constructed using the inverse distance interpolation method. The inverse distance interpolation formula is: The weight W ij The calculation formula is: In the above formula, RSRP j is the jth element in the RSRP vector received at multiple different base stations at the interpolation point, K is the number of adjacent sampling points selected during interpolation, and W ij is the weight of the i-th sampling point acting on the j-th RSRP of the interpolation point, d ij is the distance from the i-th sampling point to the point to be interpolated; The specific construction process is as follows: S41 determines the area range where the fingerprint library needs to be constructed, divides the area into appropriate grid sizes according to the positioning accuracy requirements, and determines each grid point as a potential unknown location; S42, for each grid point, using an inverse distance weight function to calculate the weight of each known data point to the unknown point, and calculating the weight of each known point to the unknown point according to the selected weight function; S43 calculates the attribute value of the unknown point by weighted average according to the attribute value of each known data point and its corresponding weight; S44 stores the estimated attribute values ​​of all grid points together with their location information to construct a complete fingerprint library.

5. The network evaluation method based on MR big data analysis according to claim 1 is characterized in that: Specifically, in step S5: by matching the MR level features, motion features, cell features, switching features, neighboring area features and the features of the 5-meter precision map with the fingerprint library, the MR is positioned to the building to distinguish between indoor and outdoor, and finally the MR sampling points are allocated to indoor buildings or outdoor areas; The basis for distinguishing indoor and outdoor is: MR cell type + TA / PD for damage estimation; horizontal positioning result + MR time cutoff for behavior feature judgment; Cell switching analysis + level fluctuation analysis Analyze level characteristics; Horizontal positioning results + electronic map to determine the ownership of buildings; The process of assigning MR sampling points to indoor buildings or outdoor areas is: S51 determines the signal strength of the current cell and the neighboring cells; S52 searches for the grid whose characteristic information is closest to the characteristic information contained in the current MR in the fingerprint grid set to which the MR sampling point belongs to the primary serving cell, selects the RP with the smallest Euclidean distance to the RSRP in the sampling point as the nearest RP, and calculates the coordinate weight of the nearest RP; the RSRP Euclidean distance formula in the sampling point is: Among them ED i，* represents the Euclidean distance between the RSRP fingerprint information of the grid in the i-th fingerprint library and the RSRP in the MR, and is the RSRP value of the grid in the i-th fingerprint library and the RSRP value in MR, where the upper corner u indicates that the RSRP comes from the u-th base station BS, and M is the number of base stations; S53 uses the coordinates of the nearest weighted RP as the position of the MR.

6. The network evaluation method based on MR big data analysis according to claim 1 is characterized in that: The indoor coverage building flow prediction formula in step S6 is: where α i is the coverage area-level MR sampling point traffic of the building; β i is the number of MR sampling points at the building coverage area level; γ i To use big data positioning output to predict the number of MR sampling points in a single cell of a building; θ is 1-the predicted weak coverage ratio of the building, δ = the average traffic of the nearest macro station around the building * the ratio of the existing indoor and outdoor traffic of the scene to which the building belongs, ε = the indoor distribution traffic of friendly companies * the market share of this company / the market share of friendly companies * the DOU of this company / the DOU of friendly companies, K1, K2, K3 are weight coefficients, K1, K2, K3 are 70%, 15%, 15% respectively.

7. The network evaluation method based on MR big data analysis according to claim 1 is characterized in that: The specific steps of step S7 are as follows: S71 determines the location of a new site according to the MR weak coverage output information; S72 Design coverage distance a of new base station based on Okumura-Hata model and COST231-Hata model formula of wireless propagation model; When the base station transmission frequency is 150-1500MHZ, L=46.3+33.9lgf-13.82lgh b -a(h m )+(44.9-6.551lgh b )lgd+C m ; When the base station transmission frequency is 1500-2000MHZ, L=69.55+26.16lg f-13.82lgh b -a(h m )+(44.9-6.551lgh b )lgd; Where: L is the median of the basic propagation loss; f is the base station frequency; h b 、h m is the effective height of the base station and mobile station antenna; a(h m ) is the mobile station antenna height correction factor; C m is the city correction factor; Designed coverage distance a = MIN ((transmitting antenna height / SIN (RADIANS (downtilt angle))) / 1000, propagation radius * COS (RADIANS (downtilt angle)) Propagation radius (km) = 10 ((maximum path loss - L) / (44.9 - 6.55 * LOG (transmitting antenna height))); S73 determines the existing network sharing cell and extracts the shared cell traffic q i , calculate the station distance b by the longitude and latitude of the site of the shared cell, b=IFERROR(6370*ACOS(COS(3.141592654 / 2-D1*3.141592654 / 180)*COS(3.141592654 / 2-B1*3.141592654 / 180)+SIN(3.141592654 / 2-D1*3.141592654 / 180)*SIN(3.141592654 / 2-B1*3.141592654 / 180)*COS(C1*3.141592654 / 180-A1*3.141592654 / 180))*1000,0) A1 and B1 are the longitude and latitude of the newly built base station; C1 and D1 are the longitude and latitude of the existing network sharing cell; S74 calculates the cell sharing coefficient γ according to the sharing coefficient model. S75 Among them, Q is the traffic volume that is calculated based on the shared cell traffic volume and the coefficient for demand capacity growth. Improve traffic for business quality capabilities, ω is the traffic expansion for business coverage capability.

8. The network evaluation method based on MR big data analysis according to claim 1 is characterized in that: Step S8 specifically: rate the total value score of the grid, predicted traffic and the importance of the scene, with the TOP30% determined as level A and the Bottom70% determined as level B. The comprehensive rating is determined by the combination of A and B levels of the three dimensions of total value score of the grid, predicted traffic and the importance of the scene, and is divided into levels 1-4.

Images