Base station position intelligent prediction method and device based on improved DBSCAN algorithm

Abstract

The invention relates to the technical field of mobile communication network optimization, in particular to a base station position intelligent prediction method and device based on an improved DBSCAN algorithm, and the method comprises the steps: carrying out the rasterization of collected original data and stock base station data, obtaining an effective Enodeb list, carrying out the classification according to an operator and a network type, and obtaining a plurality of effective Enodeb sub-lists; positioning a raster data index of each Enodeb by using a binary search algorithm, extracting latitude and longitude coordinates of all grids under the same Enodeb, and constructing a clustering data list; performing clustering analysis on the raster data in each Enodeb to generate a predicted base station position data table; wherein during clustering analysis, clustering parameters are dynamically adjusted based on the number of cell identifiers under each Enodeb; and quantitatively predicting the matching degree of the base station and the actual base station, screening out the optimal parameter of the improved DBSCAN algorithm, and predicting the position of the base station based on the optimal parameter. Therefore, different base station density scenes are adapted, and the base station position prediction precision and efficiency are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of mobile communication network optimization, and in particular to a base station position intelligent prediction method and device based on an improved DBSCAN algorithm. BACKGROUND

[0002] The base station latitude and longitude position information is the core basic data of the mobile communication network operation, and its integrity, accuracy and correlation directly determine the rationality of network planning, the effectiveness of optimization adjustment and the efficiency of maintenance work. At the same time, it is crucial for the reliability of external system applications such as network quality problem positioning (such as signal blind area identification) and user complaint analysis (such as call lag reason troubleshooting).

[0003] Currently, the base station position data mainly relies on account records and traditional correction methods (such as GPS latitude and longitude field measurement, field survey), but the above methods have significant defects: on the one hand, the account data is prone to inaccuracy due to manual input errors, base station migration information not updated in time and other problems, which seriously interferes with daily network optimization and planning; on the other hand, traditional GPS correction and field survey require a lot of manpower and resources, and there are problems of data loss and low update efficiency, which cannot meet the demand for rapid correction of base station positions after large-scale deployment of 5G networks.

[0004] In the prior art, although attempts have been made to build a base station positioning model based on Internet OTT (Over-The-Top) user perception big data, and to carry out pilot applications in the existing network to assist in correcting incorrect base station parameter information, but such models still have room for improvement in data processing efficiency, clustering algorithm adaptability and prediction accuracy. For example, the traditional DBSCAN algorithm uses fixed clustering parameters, which cannot adapt to different base station density scenarios (such as the difference in base station distribution between dense urban areas and suburban areas), and is prone to clustering results deviating from reality; at the same time, the query and matching efficiency of large-scale user data is low, which affects the overall running speed of the model. SUMMARY

[0005] In order to overcome the deficiencies in the prior art, the present application provides a base station position intelligent prediction method and device based on an improved DBSCAN algorithm, which can adapt to different base station density scenarios and improve the prediction accuracy and efficiency of base station positions.

[0006] In a first aspect, the present application provides a base station position intelligent prediction method based on an improved DBSCAN algorithm, which comprises the following steps: The collected original data and inventory base station data are subjected to gridding processing, and the data corresponding to Enodebs with a grid number less than a set threshold value are deleted to obtain an effective Enodeb list, and the effective Enodeb list is classified according to the operator and network standard to obtain a plurality of effective Enodeb sub-lists; For each of the effective Enodeb sub-list, a binary search algorithm is used to locate the grid data index of each Enodeb, and based on the grid data index, the latitude and longitude coordinates of all grids under the same Enodeb are extracted to construct a clustering data list; The grid data in each Enodeb in the clustering data list is subjected to clustering analysis to generate a predicted base station position data table; wherein, in the clustering analysis, the clustering parameters are dynamically adjusted based on the number of cell identifiers under each Enodeb; Based on the predicted base station position data table, the matching degree of the predicted base station and the actual base station is quantified, and the optimal parameters of the improved DBSCAN algorithm are screened out to predict the base station position based on the optimal parameters.

[0007] In a possible implementation, the collected raw data and inventory base station data are subjected to rasterization processing, and the data corresponding to the Enodeb with a grid number less than a set threshold is deleted to obtain an effective Enodeb list, including the following steps: The continuous geographic space is divided into discrete grids; The collected raw data is allocated to the corresponding grid according to its latitude and longitude coordinates, and the attribute values of the raw data in each grid are aggregated to generate a rasterization data table; wherein, the raw data includes user identification name, network, Enodeb, cell identifier, latitude and longitude, and RSRP average value; The number of grids corresponding to each Enodeb is counted, and the data corresponding to the Enodeb with a grid number less than a set threshold is deleted from the rasterization data table to obtain an effective Enodeb list.

[0008] In a possible implementation, the binary search algorithm is used to locate the grid data index of each Enodeb for each of the effective Enodeb sub-list, including the following steps: Each of the effective Enodeb sub-list is sorted in ascending order according to the Enodeb identifier; For the sorted effective Enodeb sub-list, a binary search algorithm is used to recursively match according to the set initial boundary and intermediate index calculation formula to obtain all grid data indexes corresponding to each Enodeb; Wherein, the intermediate index calculation formula is mid = left + (right - left) / 2; left and right are the left boundary and right boundary of the search range of the binary search, and initially, left = 0, right = enodeb_list-1, and enodeb_list is the length of the effective Enodeb sub-list.

[0009] In a possible implementation, the clustering analysis on the grid data in each Enodeb in the cluster data list generates a predicted base station position data table, including the following steps: The number of cell identifiers under each Enodeb is counted, and a corresponding clustering parameter is determined based on a preset cell identifier quantity threshold; the clustering parameter includes a clustering radius and a clustering number; Based on the clustering parameter, clustering boundary extraction and clustering gravity calculation are performed to generate a predicted base station position data table including Enodeb identifier, cell identifier quantity, and predicted base station longitude and latitude.

[0010] In a possible implementation, when the number of cell identifiers under an Enodeb is less than or equal to a set cell identifier quantity threshold, a first set of clustering parameters is used; when the number of cell identifiers under an Enodeb is greater than the set cell identifier quantity threshold, a second set of clustering parameters is used to perform a first clustering, and a third set of clustering parameters is used to perform a second clustering based on the result of the first clustering.

[0011] In a possible implementation, the matching degree of the predicted base station and the actual base station is quantified based on the predicted base station position data table to filter out optimal parameters for improving the DBSCAN algorithm, including the following steps: The distance between the predicted base station longitude and latitude in the predicted base station position data table and the actual base station longitude and latitude in the inventory base station data is calculated; The calculated distance is matched with a set distance interval to quantify the matching degree of the predicted base station and the actual base station, and the optimal parameters for improving the DBSCAN algorithm are output when a set matching result threshold is met.

[0012] In a possible implementation, the distance interval is divided into four distance intervals: inventory station, suspected inventory station, suspected self-keeping station, and self-keeping station; and a matching result threshold is set based on the proportion of inventory stations and suspected inventory stations.

[0013] In a second aspect, the application provides a base station position intelligent prediction device based on an improved DBSCAN algorithm, including: A preprocessing module is configured to perform grid processing on collected raw data and inventory base station data, delete data corresponding to Enodebs with a grid quantity less than a set threshold to obtain an effective Enodeb list, and classify the effective Enodeb list according to operators and network standards to obtain multiple effective Enodeb sublists. The module is used to locate the raster data index of each Enodeb for each valid Enodeb sublist using a binary search algorithm, and to extract the latitude and longitude coordinates of all rasters under the same Enodeb based on the raster data index to construct a clustered data list. The clustering module is used to perform clustering analysis on the raster data in each Enodeb of the clustering data list and generate a predicted base station location data table; wherein, during the clustering analysis, the clustering parameters are dynamically adjusted based on the number of cell identifiers under each Enodeb; The filtering module is used to quantify the matching degree between the predicted base stations and the actual base stations based on the predicted base station location data table, filter out the optimal parameters of the improved DBSCAN algorithm, and predict the base station location based on the optimal parameters.

[0014] Thirdly, this application provides an electronic device, including: a processor, a memory, and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor communicates with the memory via the bus, and when the machine-readable instructions are executed by the processor, they perform the steps of the base station location intelligent prediction method based on the improved DBSCAN algorithm as described in any of the first aspects.

[0015] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, performs the steps of the base station location intelligent prediction method based on the improved DBSCAN algorithm as described in any of the first aspects.

[0016] This embodiment provides a method and apparatus for intelligent prediction of base station locations based on an improved DBSCAN algorithm. The method involves rasterizing the collected raw data and existing base station data, deleting data corresponding to Enodebs with fewer than a set threshold of raster counts to obtain a list of valid Enodebs. This list is then categorized according to operator and network standard to obtain multiple sub-lists of valid Enodebs. For each sub-list, a binary search algorithm is used to locate the raster data index of each Enodeb, and the latitude and longitude coordinates of all rasters under the same Enodeb are extracted based on the raster data index to construct a clustered data list. Cluster analysis is performed on the raster data in each Enodeb of the clustered data list to generate a predicted base station location data table. During cluster analysis, clustering parameters are dynamically adjusted based on the number of cell identifiers under each Enodeb. The predicted base station location data table quantifies the matching degree between predicted and actual base stations, and the optimal parameters of the improved DBSCAN algorithm are selected to predict base station locations based on these optimal parameters. This adapts to different base station density scenarios and improves the accuracy and efficiency of base station location prediction. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A flowchart of a base station location intelligent prediction method based on an improved DBSCAN algorithm according to an embodiment of this application is shown; Figure 2 The flowchart illustrates an embodiment of this application that performs rasterization processing on collected raw data and existing base station data, and deletes data corresponding to Enodeb with a raster count less than a set threshold to obtain a list of valid Enodeb. Figure 3 This document illustrates a flowchart illustrating how a binary search algorithm is used to locate the raster data index of each Enodeb for each of the valid Enodeb sublists, as described in an embodiment of this application. Figure 4 This document illustrates a flowchart of an embodiment of the present application describing the clustering analysis of raster data in each Enodeb of the clustered data list to generate a predicted base station location data table. Figure 5 The flowchart illustrates an embodiment of this application, which quantifies the matching degree between predicted base stations and actual base stations based on the predicted base station location data table and filters out the optimal parameters of the improved DBSCAN algorithm. Figure 6 A schematic diagram of the structure of a base station location intelligent prediction device based on an improved DBSCAN algorithm according to an embodiment of this application is shown; Figure 7 A structural block diagram of an electronic device according to an embodiment of this application is shown. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the accompanying drawings in this application are for illustrative and descriptive purposes only and are not intended to limit the scope of protection of this application. Furthermore, it should be understood that the schematic drawings are not drawn to scale. The flowcharts used in this application illustrate operations implemented according to some embodiments of this application. It should be understood that the operations in the flowcharts may not be implemented in sequence, and steps without logical contextual relationships may be reversed or implemented simultaneously. In addition, those skilled in the art, guided by the content of this application, may add one or more other operations to the flowcharts, or remove one or more operations from the flowcharts.

[0020] Furthermore, the described embodiments are merely some, not all, of the embodiments of this application. The components of the embodiments of this application described and illustrated herein can typically be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0021] It should be noted that the term "comprising" will be used in the embodiments of this application to indicate the presence of the features declared thereafter, but does not exclude the addition of other features.

[0022] In view of the technical problems raised in the background art, this application provides a method and device for intelligent prediction of base station location based on the improved DBSCAN algorithm, which can not only adapt to different base station density scenarios, but also improve the accuracy and efficiency of base station location prediction.

[0023] In one embodiment, see the appendix to the specification. Figure 1 This application provides a method for intelligent prediction of base station location based on an improved DBSCAN algorithm, comprising the following steps: S1. The collected raw data and existing base station data are rasterized, and the data corresponding to Enodeb with less than a set threshold of raster count is deleted to obtain a list of valid Enodeb. The list of valid Enodeb is then classified according to the operator and network standard to obtain multiple sub-lists of valid Enodeb. S2. For each of the valid Enodeb sublists, a binary search algorithm is used to locate the raster data index of each Enodeb, and the latitude and longitude coordinates of all rasters under the same Enodeb are extracted based on the raster data index to construct a clustered data list. S3. Perform cluster analysis on the raster data in each Enodeb of the clustered data list to generate a predicted base station location data table; wherein, during the cluster analysis, the clustering parameters are dynamically adjusted based on the number of cell identifiers under each Enodeb. S4. Based on the predicted base station location data table, quantify the matching degree between the predicted base station and the actual base station, select the optimal parameters of the improved DBSCAN algorithm, and predict the base station location based on the optimal parameters.

[0024] Specifically, in step S1, please refer to the appendix of the instruction manual. Figure 2 The process of rasterizing the collected raw data and existing base station data, and deleting data corresponding to Enodebs with fewer than a set threshold of raster counts to obtain a list of valid Enodebs, includes the following steps: S101. Divide the continuous geographic space into discrete grids; S102. The collected raw data is allocated to the corresponding grid according to its latitude and longitude coordinates, and the attribute values ​​of the raw data in each grid are aggregated to generate a rasterized data table; wherein, the raw data includes user identification name, network, Enodeb, cell identification, latitude and longitude, and RSRP mean; S103. Count the number of raster cells corresponding to each Enodeb, and delete the data corresponding to Enodeb whose number of raster cells is less than a set threshold from the rasterized data table to obtain a list of valid Enodeb cells.

[0025] The collected raw data mainly includes user identifier name (opt_name), network, enodeb (E-UTRANNode B, Evolved Universal Terrestrial Radio Access Network Node B), cell identifier (ci), longitude, latitude, and mean RSRP (Reference Signal Received Power; a higher RSRP value indicates a closer distance to the base station and better signal quality; conversely, a lower RSRP value indicates a weaker signal or greater distance). Existing base station data mainly includes the base station number, base station name, and the latitude and longitude of the existing base stations.

[0026] To ensure data accuracy and consistency, and to provide a reliable foundation for subsequent analysis and clustering operations, this application also cleans the collected raw data and existing base station data to identify and remove missing, duplicate, and invalid values. For example, the `df.dropna()` function is used to delete rows or columns containing missing values ​​(NaN or None), and it can flexibly remove missing values ​​from the data according to different parameter settings.

[0027] In step S101, the continuous geographic space is divided into discrete rasters, each representing a small geographic area. This allows user data and existing base station data to be mapped onto the rasters, facilitating subsequent processing. First, the raw data to be rasterized is acquired, and the size, shape, and extent of the rasters are defined according to the analysis requirements. Raster shapes can be regular rectangles, squares, etc., and their size and resolution should be determined based on the characteristics of the data and the analysis objectives. The raw data points are assigned to the corresponding rasters based on their geographic coordinates. This typically involves calculating the relationship between the latitude and longitude of each data point and the raster boundary to determine its assigned raster. For each raster, attribute values ​​(such as average, sum, maximum, etc.) of all data points within it are calculated to generate the raster values ​​for each raster dataset. The rasterized data is then stored as a data table structure.

[0028] In one embodiment, GeoHash encoding is used for rasterization: each geographic coordinate (latitude and longitude) in the original data is GeoHash encoded. GeoHash is a method for converting latitude and longitude coordinates into short strings by dividing the Earth's surface into a series of adjacent grid cells and encoding each grid cell. The longitude of the GeoHash code is determined based on the analysis requirements. The higher the longitude, the longer the encoded string and the more precise the identified geographic location. Typically, the length of the GeoHash code can be adjusted as needed. GeoHash rasters are generated based on the GeoHash codes and the defined raster size. This typically involves mapping the GeoHash codes to a raster coordinate system and dividing the raster into different regions based on the similarity of the GeoHash codes. The original data points are assigned to the corresponding GeoHash rasters according to their GeoHash codes. Then, the data points within each GeoHash raster are aggregated to calculate the attribute values ​​within the raster. The generated GeoHash raster data is stored in a table structure.

[0029] In step S103, because the number of rasters corresponding to Enodeb varies greatly in the original data, some Enodeb data has too few rasters, causing deviations in the optimization settings of the base station location intelligent prediction and analysis model based on the improved DBSCAN algorithm, affecting the overall analysis effect of the model. Therefore, it is necessary to filter out Enodeb data with too few rasters. For example, Counter is a class in the collections module of the Python standard library, used to count hashable objects. Counter is a dictionary subclass where elements are keys and their counts are values. This class is particularly suitable for counting tasks, such as counting the number of times each Enodeb appears in the original data list, that is, the number of rasters under each Enodeb. In one embodiment, Enodeb identifiers with a raster count of less than 30 are counted, and the corresponding identifier rows are deleted.

[0030] Furthermore, in this application, the list of valid Enodeb entries is categorized according to major operators and network standards. This categorization adapts to the data characteristics of different scenarios, avoids mutual interference in model prediction and analysis, provides a homogeneous data foundation for improving the parameter optimization of the DBSCAN algorithm, and ultimately enhances the accuracy and scenario-based application value of base station location prediction. In one embodiment, the entries are divided into six types based on existing operators and network standards: China Mobile 4G, China Mobile 5G, China Unicom 4G, China Unicom 5G, China Telecom 4G, and China Telecom 5G.

[0031] In addition, it should be noted that the cluster radius and number of clusters in the DBSCAN algorithm need to be specified. The selection of these two parameter values ​​is a challenge because different parameter values ​​may lead to completely different results. Grid Search is a systematic hyperparameter optimization method that is widely used in the parameter tuning process of machine learning models. The basic idea of ​​Grid Search is to try different combinations of hyperparameters one by one in the parameter space specified by the user through exhaustive search, and select the parameter combination that makes the model perform optimal based on a certain evaluation criterion. Here, "grid" refers to a multi-dimensional space formed by all candidate hyperparameter value combinations, and the algorithm searches for each point (i.e., each parameter combination) in this space one by one. When using Grid Search, the following key parameters need to be clarified: (1) Hyperparameters to be optimized: These are parameters that need to be set before training the base station prediction analysis model. They directly affect the performance and generalization ability of the model. This paper focuses on the DBSCAN algorithm, so the hyperparameters to be optimized include the cluster radius eps and the number of clusters minpts; (2) Range of hyperparameter values: For each hyperparameter to be optimized, a range of values ​​or a list of candidate values ​​needs to be specified. These value ranges will form a parameter grid, and the algorithm will search the grid one by one. For example, the cluster radius range is [0.0005, 0.003], and the number of clusters ranges is [10, 30]. (3) Evaluation criteria: The standard by which users evaluate the model performance under each parameter combination. This evaluation criterion is set to achieve a predetermined prediction accuracy.

[0032] Therefore, it is also necessary to perform grid search parameter settings that are suitable for model tuning, specifying a range of values ​​or a list of candidate values ​​for each hyperparameter to be optimized. These ranges of values ​​will form a parameter grid. By setting appropriate parameter ranges according to the specific situation, the algorithm will try each combination of parameters in the parameter grid one by one. For each combination, a model will be trained, and its performance will be evaluated using the specified evaluation criteria.

[0033] In step S2, see the appendix to the instruction manual. Figure 3 The step of locating the raster data index of each Enodeb for each valid Enodeb sublist using a binary search algorithm includes the following steps: S201. Sort each of the valid Enodeb sublists in ascending order according to the Enodeb identifier; S202. For the sorted valid Enodeb sublist, a binary search algorithm is used to recursively match according to the set initial boundary and intermediate index calculation formula to obtain all raster data indices corresponding to each Enodeb.

[0034] Due to the massive volume of the original data, traditional iterative analysis methods are inefficient. Binary search algorithms are commonly used to find specific elements in large datasets, accelerating querying or filtering operations on the original data. In this application, the binary search algorithm combined with recursion can repeatedly halve the search range to find multiple index positions within the Enodeb list under a given Enodeb, i.e., the indexes of all raster data in a given Enodeb. This results in high query efficiency and is beneficial for subsequent raster data clustering analysis within the Enodeb.

[0035] The parameters are set as follows: (1) target value: the value of the Enodeb to be searched; (2) ordered array (enodeb_list): the length of the effective Enodeb sublist; (3) left and right boundaries (left, right): initially, the left boundary left is 0 and the right boundary right is the array length minus 1; (4) middle index (mid): each time a search is performed, the middle index mid is calculated to access the middle element of the array. Its calculation formula is mid = left + (right - left) / 2.

[0036] The process is as follows: (1) Initialization: Set the target value of Enodeb to be searched; determine the left and right boundaries of the array to be searched. Initially, left = 0; right = array length - 1; (2) Calculate the middle index: take the middle position of the left and right boundaries as the current search point, i.e., mid = (left + night) / 2; (3) Compare the target value with the middle element: check whether the element in the middle position of the array is equal to the target value target; if they are equal, an index position of the target value has been found, but at this time it is necessary to continue to search for other index positions; if the target value is less than the middle element, adjust the right boundary to mid-1, and then repeat steps (2) and (3); if the target value is greater than the middle element, adjust the left boundary to mid + 1. Then repeat steps (2) and (3); (4) Process multiple index positions: After finding the first matching enodeb index, do not return immediately, but continue to search to the left and right until no more matching elements are found; two pointers can be used to record the position of the currently found matching element and expand to the left and right until the condition is no longer met; (5) Termination condition: When the left boundary is greater than the right boundary, it means that there are no more target values ​​in the array and the search process ends; (6) Return result: Return a list or set containing all the found index positions.

[0037] Then, based on the obtained raster data index, the latitude and longitude coordinates of all rasters under the same Enodeb are extracted, and a clustered data list is constructed. The constructed clustered data list is as follows: ,in This represents the number of distinct Enodeb entries in the enodeb_list list. ( ),in This represents the number of raster cells within the same Enodeb.

[0038] In step S3, the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) clustering algorithm is used to perform cluster analysis on the user network data. DBSCAN is a density-based clustering algorithm that can divide points with high-density connectivity into clusters and identify noise points (i.e., points that do not belong to any cluster). The cluster radius eps and the number of clusters minpts are two key parameters of the DBSCAN algorithm. eps defines the neighborhood radius of a point, and minpts defines the minimum number of points required to form a dense region. The relevant definitions of the DBSCAN algorithm are as follows: - Neighborhood: that is, for a sample point In the sample set In, with Distance not greater than The set of points, i.e. Core object: If of - The neighborhood contains more than minpts of sample points, i.e. Then it is called It is a core object. The DBSCAN algorithm divides sample points into three categories: core points: if sample points... of The neighborhood contains at least If there are 100 sample points, then the sample points are 100. Core point; boundary point: if sample point of Sample points within the neighborhood are less than And if it belongs to the neighborhood of other core points, then the sample point Core points; noise points: sample points that are neither core points nor boundary points. Density reach: if... exist Within the neighborhood, If it is a core object, then it is called... Depend on Density is directly attainable. The converse is not necessarily true; density is achievable: for and , such as sample sequences ,satisfy Depend on Density reaches directly, then it is called Depend on Density is achievable. Density achievable satisfies transitivity; density is interconnected: if a core point exists. ,make and All by If the density can reach, then it is called and Density connected.

[0039] See the instruction manual appendix Figure 4 The step of performing cluster analysis on the raster data in each Enodeb of the clustered data list to generate a predicted base station location data table includes the following steps: S301. Count the number of cell identifiers under each Enodeb, and determine the corresponding clustering parameters based on the preset threshold for the number of cell identifiers; the clustering parameters include the cluster radius and the number of clusters; S302. Based on the clustering parameters, perform cluster boundary extraction and cluster centroid calculation to generate a predicted base station location data table including Enodeb identifier, number of cell identifiers, and predicted base station latitude and longitude.

[0040] In step S301, since there are several ci in the same Enodeb, and the number of ci is positively correlated with the number of base stations under the Enodeb, this application sets the DBSCAN parameter differently based on the number of ci under the Enodeb. For example, the number of different ci under the same Enodeb is counted, divided into two cases: the number of ci is less than or equal to 4 and the number of ci is greater than 4, and the base station location is predicted and analyzed for each case.

[0041] In one embodiment, the parameters of the DBSCAN algorithm are set and optimized according to the data distribution characteristics and actual needs. The data distribution can be observed using visualization tools (such as scatter plots) to select appropriate eps and minpts values. In this application, when the number of ci is less than or equal to 4, eps1 is set to 150 meters and minpts1 is set to 30. When the number of ci is greater than 4, eps2 is set to 50 meters and minpts2 is set to 10 to remove noise points and prevent the number of base stations predicted by the model from being too high and deviating from the actual situation. Then, eps3 is set to 80 meters and minpts3 is set to 30.

[0042] When the number of cluster points (CIs) is less than or equal to 4, the cluster points obtained by the DBSCAN algorithm can be directly used as the set of Enodeb cluster point coordinates. When the number of CIs is greater than 4, the DBSCAN algorithm obtains cluster points, which are then subjected to a second clustering process combined with the settings of parameters eps3 and minpts3. Only the cluster points obtained after this second clustering process can be used as the set of Enodeb cluster point coordinates. This method yields results that are closer to real-world scenarios. The set of cluster points can be described as follows: ,in This represents the number of distinct clusters in the `enodeb_list` list. The set of point coordinates within each cluster can be represented as: ,in This represents the number of raster cells under the same enodeb.

[0043] In step S302, since the obtained clustering result is a set of points, this application uses its centroid as the base station location for prediction analysis. Therefore, it is necessary to first extract the boundary of the cluster point set and then calculate the centroid. Through analysis The distance and angle relationship between the midpoint and its nearest neighbor can be used to detect potential edges. This typically involves setting distance and angle thresholds to determine which points should be considered vertices of edges. Once a polygon is identified or fitted, vertex coordinates can be extracted from the polygon's edges or contours, thus forming the cluster boundaries of these clustered point sets.

[0044] Calculating the centroid of a polygon is a geometric problem, which involves manipulating the coordinates of the polygon's vertices. The following is a detailed process for calculating the centroid of a polygon: (1) Prepare data: A data list consisting of the vertices of the cluster boundary, whose coordinates are respectively (2) Accumulator for calculating area and centroid coordinates: Area accumulator: Initialize a variable The value is 0, used to store the area of ​​the polygon; the barycentric coordinate accumulator initializes two variables. and =0, used to store the accumulated value of the centroid coordinates. (3) Traverse the vertices to calculate the area and centroid coordinates: For each pair of adjacent vertices of the polygon and (in Iterate from 1 to ,and Considered To form a closed loop, perform the following operations: Calculate area increment: using vertices and Part of the area calculation is done using the following formula: .

[0045] Update total area: Add to total area superior: .

[0046] Calculate the centroid coordinate increment: using the vertex and The formula for calculating the increment of the centroid coordinates is:

[0047] Update the barycentric coordinate accumulator: and Add to and superior:

[0048] (4) Calculate the final centroid coordinates: using the total area To calculate the final centroid coordinates ;

[0049] Here, divided by Because of the calculation and In this case, the area of ​​the triangle formed by each pair of vertices is calculated three times (each vertex forms a triangle with each of its two adjacent vertices in turn), and the formula for the area of ​​a triangle is... Therefore, the total coefficient is 6.

[0050] (5) Output results: Output the calculated centroid coordinates That is, predicting the location of the base station.

[0051] This leads to a predicted base station location data table that corresponds to the effective Enodeb sublist and includes Enodeb identifiers, cell identifiers, and predicted base station latitude and longitude.

[0052] In step S4, see the appendix to the instruction manual. Figure 5 The process of quantifying the matching degree between predicted base stations and actual base stations based on the predicted base station location data table, and selecting the optimal parameters for the improved DBSCAN algorithm, includes the following steps: S401. Calculate the distance based on the predicted base station latitude and longitude in the predicted base station location data table and the actual base station latitude and longitude in the existing base station data. S402. Match the calculated distance with the set distance interval to quantify the matching degree between the predicted base station and the actual base station, and output the optimal parameters of the improved DBSCAN algorithm when the set matching result threshold is met.

[0053] In step S401, the distance is calculated using the following formula:

[0054] in: It is the predicted minimum distance between the base station and the actual base station. It is the first in the predicted base station location data table Predicted base station latitude and longitude coordinates, Indicates the first The latitude and longitude coordinates in the existing base station data.

[0055] In step S402, distance interval partitioning is a necessary condition for setting the evaluation criteria for the grid search algorithm; the more accurate the distance interval matching, the stronger the applicability of the model. Distance interval The classification criteria are as follows: Existing stations have a distance range of [0m-50m]; suspected existing stations have a distance range of [50m-150m]; suspected self-retained stations have a distance range of [150m-250m]; and self-retained stations have a distance range of [250m-1000m]. Each predicted base station has one... ,according to The value is matched with the divided distance interval to obtain the corresponding ratio in each interval. For example, the ratio of existing stations and suspected existing stations can be added together and set to a fixed value as the evaluation standard of the grid search algorithm.

[0056] When the distance interval matching result setting value is met, the optimal parameters of the improved DBSCAN algorithm are output. That is, among all parameter combinations tried by the grid search algorithm, the parameter combination that best meets the evaluation criteria is selected as the optimal parameters, and then the selected optimal parameters are used to predict the base station location. Preferably, the performance of the final model can be re-evaluated using an independent test set to verify its generalization ability before predicting the base station location.

[0057] As can be seen, the intelligent base station location prediction method provided in this application, based on an improved DBSCAN algorithm, adapts to different base station density scenarios by improving the DBSCAN algorithm (dynamically adjusting parameters based on the number of ci), effectively solving the clustering bias problem caused by the fixed parameters of the traditional DBSCAN. Simultaneously, GeoHash rasterization and data cleaning reduce noise interference, further improving prediction accuracy. Furthermore, a binary search algorithm is used to accelerate the matching of Enodeb and raster data, improving efficiency by more than 50% compared to traditional loop traversal. Automated parameter optimization through grid search avoids the subjectivity and time-consuming nature of manual parameter tuning, adapting to the needs of large-scale base station data processing. In addition, a clearly defined evaluation standard is established through distance interval matching, ensuring that the prediction results can be directly used for base station ledger correction, providing reliable support for improving base station coverage quality.

[0058] Based on the same inventive concept, this application also provides a base station location intelligent prediction device based on an improved DBSCAN algorithm. Since the principle of the device in this application is similar to the method of the base station location intelligent prediction method based on an improved DBSCAN algorithm described above, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be described again.

[0059] As per the instruction manual Figure 6 As shown, this application also provides a base station location intelligent prediction device based on an improved DBSCAN algorithm, the device comprising: The preprocessing module 601 is used to perform rasterization processing on the collected raw data and existing base station data, and delete the data corresponding to Enodeb with less than a set threshold number of raster cells to obtain a list of valid Enodeb cells. The list of valid Enodeb cells is then classified according to the operator and network standard to obtain multiple sub-lists of valid Enodeb cells. The construction module 602 is used to locate the raster data index of each Enodeb for each valid Enodeb sublist using a binary search algorithm, and extract the latitude and longitude coordinates of all rasters under the same Enodeb based on the raster data index to construct a clustered data list. Clustering module 603 is used to perform clustering analysis on the raster data in each Enodeb in the clustering data list to generate a predicted base station location data table; wherein, during the clustering analysis, the clustering parameters are dynamically adjusted based on the number of cell identifiers under each Enodeb; The filtering module 604 is used to quantify the matching degree between the predicted base station and the actual base station based on the predicted base station location data table, filter out the optimal parameters of the improved DBSCAN algorithm, and predict the base station location based on the optimal parameters.

[0060] In some embodiments, the preprocessing module 601 performs rasterization processing on the collected raw data and existing base station data, and deletes data corresponding to Enodebs with fewer than a set threshold of raster counts to obtain a list of valid Enodebs. This includes: dividing the continuous geographic space into discrete rasteres using GeoHash encoding; allocating the collected raw data to the corresponding rasteres according to their latitude and longitude coordinates, and aggregating the attribute values ​​of the raw data within each raster to generate a rasterized data table; wherein the raw data includes user identifier name, network, Enodeb, cell identifier, latitude and longitude, and mean RSRP; counting the number of rasteres corresponding to each Enodeb, and deleting data corresponding to Enodebs with fewer than a set threshold of raster counts from the rasterized data table to obtain a list of valid Enodebs.

[0061] In some embodiments, the construction module 602 uses a binary search algorithm to locate the raster data index of each Enodeb for each valid Enodeb sublist, including: sorting each valid Enodeb sublist in ascending order according to the Enodeb identifier; and using a binary search algorithm to recursively match the sorted valid Enodeb sublist according to a set initial boundary and intermediate index calculation formula to obtain all raster data indices corresponding to each Enodeb; wherein, the intermediate index calculation formula is mid = left + (right - left) / 2; left and right are the left and right boundaries of the search range of the binary search, respectively. Initially, left = 0, right = enodeb_list - 1, and enodeb_list is the length of the valid Enodeb sublist.

[0062] In some embodiments, the clustering module 603 performs cluster analysis on the raster data in each Enodeb of the clustering data list to generate a predicted base station location data table, including: counting the number of cell identifiers under each Enodeb, and determining the corresponding clustering parameters based on a preset cell identifier number threshold; the clustering parameters include the cluster radius and the number of clusters; and extracting cluster boundaries and calculating cluster centroids based on the clustering parameters to generate a predicted base station location data table including Enodeb identifiers, the number of cell identifiers, and the predicted base station latitude and longitude. Specifically, when the number of cell identifiers under an Enodeb is less than or equal to the set area identifier number threshold, the first set of clustering parameters is used; when the number of cell identifiers under an Enodeb is greater than the set area identifier number threshold, the second set of clustering parameters is used to perform a first clustering, and then the third set of clustering parameters is used to perform a second clustering based on the result of the first clustering.

[0063] In some embodiments, the filtering module 604 quantifies the matching degree between predicted base stations and actual base stations based on the predicted base station location data table, and filters out the optimal parameters of the improved DBSCAN algorithm, including: calculating the distance based on the latitude and longitude of the predicted base stations in the predicted base station location data table and the latitude and longitude of the actual base stations in the existing base station data; matching the calculated distance with a set distance interval to quantify the matching degree between the predicted base stations and actual base stations, and outputting the optimal parameters of the improved DBSCAN algorithm when the set matching result threshold is met. Specifically, the distance interval is divided into four distance intervals: existing stations, suspected existing stations, suspected self-retained stations, and self-retained stations; and a matching result threshold is set based on the proportion of existing stations and suspected existing stations.

[0064] This application provides a base station location intelligent prediction device based on an improved DBSCAN algorithm, which can adapt to different base station density scenarios and improve the accuracy and efficiency of base station location prediction.

[0065] Based on the same concept of the present invention, the specification is attached. Figure 7 As shown in the figure, an embodiment of this application provides the structure of an electronic device 700, which includes: at least one processor 701, at least one network interface 704 or other user interface 703, a memory 705, and at least one communication bus 702. The communication bus 702 is used to realize the connection and communication between these components. The electronic device 700 may optionally include a user interface 703, including a display (e.g., touch screen, LCD, CRT, holographic imaging, or projector, etc.), a keyboard, or a clicking device (e.g., mouse, trackball, touchpad, or touch screen, etc.).

[0066] Memory 705 may include read-only memory and random access memory, and provides instructions and data to processor 701. A portion of memory 705 may also include non-volatile random access memory (NVRAM).

[0067] In some implementations, memory 705 stores executable modules or data structures, or subsets thereof, or extended sets thereof: The 7051 operating system contains various system programs used to implement various basic business functions and handle hardware-based tasks. Application module 7052 contains various applications, such as desktop (launcher), media player (MediaPlayer), browser (Browser), etc., to implement various application services.

[0068] In this embodiment of the application, by calling the program or instructions stored in the memory 705, the processor 701 is used to execute steps such as in a base station location intelligent prediction method based on an improved DBSCAN algorithm, which can adapt to different base station density scenarios and improve the accuracy and efficiency of base station location prediction.

[0069] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, performs steps such as those in the base station location intelligent prediction method based on the improved DBSCAN algorithm.

[0070] Specifically, the storage medium can be a general-purpose storage medium, such as a portable disk or hard disk. When the computer program on the storage medium is run, it can execute the aforementioned intelligent prediction method for base station location based on the improved DBSCAN algorithm.

[0071] In the embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and there may be other division methods in actual implementation. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the coupling or direct coupling or communication connection shown or discussed may be through some communication interface; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0072] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0073] In addition, the functional units in the embodiments provided in this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0074] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0075] Finally, it should be noted that the above embodiments are merely specific implementations of this application, used to illustrate the technical solutions of this application, and not to limit them. The protection scope of this application is not limited thereto. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features, within the scope of the technology disclosed in this application; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application. All should be covered within the protection scope of this application. Therefore, the protection scope of this application should be determined by the protection scope of the claims.

Claims

1. A base station location intelligent prediction method based on an improved DBSCAN algorithm, characterized in that, The method comprises the following steps: The collected raw data and inventory base station data are rasterized, and the data corresponding to Enodebs with a number of grids less than a set threshold value are deleted to obtain an effective Enodeb list, and the effective Enodeb list is classified according to operators and network modes to obtain a plurality of effective Enodeb sublists; For each effective Enodeb sublist, a binary search algorithm is used to locate the grid data index of each Enodeb, and the latitude and longitude coordinates of all grids under the same Enodeb are extracted based on the grid data index to construct a clustering data list; The grid data in each Enodeb in the clustering data list is subjected to clustering analysis to generate a predicted base station position data table; wherein, during the clustering analysis, the clustering parameters are dynamically adjusted based on the number of cell identifiers under each Enodeb; The matching degree of the predicted base station and the actual base station is quantified based on the predicted base station position data table, and the optimal parameters of the improved DBSCAN algorithm are screened out to predict the base station position based on the optimal parameters. 2.The base station location intelligent prediction method based on the improved DBSCAN algorithm of claim 1, wherein, The rasterization of the collected raw data and inventory base station data, and the deletion of the data corresponding to Enodebs with a number of grids less than a set threshold value to obtain an effective Enodeb list, comprises the following steps: Divide the continuous geographic space into discrete grids; The collected raw data are allocated to the corresponding grids according to their latitude and longitude coordinates, and the attribute values of the raw data in each grid are aggregated to generate a rasterized data table; wherein, the raw data include user identifier name, network, Enodeb, cell identifier, latitude and longitude, and RSRP mean value; The number of grids corresponding to each Enodeb is counted, and the data corresponding to Enodebs with a number of grids less than a set threshold value are deleted from the rasterized data table to obtain an effective Enodeb list. 3.The base station location intelligent prediction method based on the improved DBSCAN algorithm of claim 1, wherein, The binary search algorithm is used to locate the grid data index of each Enodeb for each effective Enodeb sublist, which comprises the following steps: Each effective Enodeb sublist is sorted in ascending order according to the Enodeb identifier; For the sorted effective Enodeb sublist, a binary search algorithm is used to recursively match according to a set initial boundary and a middle index calculation formula to obtain all grid data indexes corresponding to each Enodeb; Wherein, the middle index calculation formula is mid = left + (right - left) / 2; left and right are the left boundary and the right boundary of the search range of the binary search, and initially, left = 0, right = enodeb_list-1, and enodeb_list is the length of the effective Enodeb sublist. 4.The base station location intelligent prediction method based on improved DBSCAN algorithm of claim 1, wherein, The clustering analysis of the grid data in each Enodeb in the clustering data list to generate a predicted base station position data table comprises the following steps: Count the number of cell identifiers under each Enodeb, and determine the corresponding clustering parameters based on a preset threshold of the number of cell identifiers; the clustering parameters include clustering radius and clustering number; Based on the clustering parameters, perform clustering boundary extraction and clustering center calculation to generate a predicted base station position data table including Enodeb identifier, number of cell identifiers, and predicted base station longitude and latitude.

5. The intelligent base station location prediction method based on the improved DBSCAN algorithm according to claim 4, characterized in that, Wherein, When the number of cell identifiers under the Enodeb is less than or equal to the set threshold of the number of cell identifiers, the first set of clustering parameters is used; when the number of cell identifiers under the Enodeb is greater than the set threshold of the number of cell identifiers, the second set of clustering parameters is used to perform the first clustering, and the third set of clustering parameters is used to perform the second clustering based on the results of the first clustering.

6. The base station location intelligent prediction method based on the improved DBSCAN algorithm according to claim 5, characterized in that, The matching degree of the predicted base station and the actual base station is quantified based on the predicted base station position data table, and the optimal parameters of the improved DBSCAN algorithm are screened out, including the following steps: Distance calculation is performed based on the predicted base station longitude and latitude in the predicted base station position data table and the actual base station longitude and latitude in the inventory base station data; The calculated distance is matched with the set distance interval to quantify the matching degree of the predicted base station and the actual base station, and the optimal parameters of the improved DBSCAN algorithm are output when the set matching result threshold is met.

7. The intelligent base station location prediction method based on the improved DBSCAN algorithm according to claim 6, characterized in that, Wherein, The distance interval is divided into four distance intervals: inventory station, suspected inventory station, suspected self-retained station, and self-retained station; and the matching result threshold is set based on the proportion of inventory stations and suspected inventory stations.

8. A base station location intelligent prediction device based on an improved DBSCAN algorithm, characterized in that, The device comprises: A preprocessing module for rasterizing the collected raw data and inventory base station data, and deleting the data corresponding to the Enodeb whose grid number is less than the set threshold to obtain an effective Enodeb list, and classifying the effective Enodeb list according to the operator and network standard to obtain multiple effective Enodeb sublists; A construction module for locating the grid data index of each Enodeb by using the binary search algorithm for each effective Enodeb sublist, and extracting the longitude and latitude coordinates of all grids under the same Enodeb based on the grid data index to construct a clustering data list; A clustering module for performing clustering analysis on the grid data in each Enodeb in the clustering data list to generate a predicted base station position data table; wherein, during the clustering analysis, the clustering parameters are dynamically adjusted based on the number of cell identifiers under each Enodeb; A screening module for quantifying the matching degree of the predicted base station and the actual base station based on the predicted base station position data table, and screening out the optimal parameters of the improved DBSCAN algorithm to predict the base station position based on the optimal parameters.

9. An electronic device, comprising: It comprises: A processor, a memory, and a bus, the memory stores machine readable instructions executable by the processor, when the electronic device is running, the processor and the memory communicate through the bus, and the machine readable instructions are executed by the processor to perform the steps of the base station position intelligent prediction method based on the improved DBSCAN algorithm as claimed in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, and the computer program is run by the processor to execute the steps of the base station position intelligent prediction method based on the improved DBSCAN algorithm according to any one of claims 1 to 7.

Images