Base station position prediction method, device, equipment and medium

Abstract

The invention relates to the technical field of communication, and provides a base station position prediction method and device, equipment and a medium. The method comprises the steps that a measurement data file and a base station table are obtained, the measurement data file comprises user identifications and a series of base station identification switching records of users, and the base station table comprises base station identifications with known positions and corresponding longitude and latitude information; according to the measurement data file, determining a comprehensive switching graph corresponding to all users; based on the comprehensive switching diagram and the base station table, determining an embedded vector of a base station with an unknown position; the embedded vector of the base station with the unknown position is input into the trained base station position prediction model to obtain a latitude and longitude prediction value of the base station with the unknown position, and the trained base station position prediction model comprises a latitude prediction module used for predicting the latitude value of the base station with the unknown position according to the embedded vector of the base station with the unknown position; and the longitude prediction module is used for predicting the longitude value of the base station according to the embedded vector of the position-unknown base station. And the longitude and latitude prediction accuracy of the base station is improved.

Description

Technical Field

[0001] The present application relates to the field of communication technology, and in particular to a base station location prediction method, apparatus, device, and medium. Background Art

[0002] In the communications sector, base station location information is crucial for network optimization, signal coverage analysis, and user location-based services. Accurate base station latitude and longitude data can help operators plan network layouts, improve signal quality, and reduce costs. However, in practice, due to complex data collection, missing data, or errors, some base station latitude and longitude information is incomplete or erroneous, hindering subsequent applications.

[0003] Traditional field measurements to obtain base station latitude and longitude are labor-intensive, time-consuming, and challenging to implement in hard-to-reach areas like remote mountainous areas and densely populated areas. Therefore, developing easy-to-implement, low-cost methods for completing base station latitude and longitude has become a critical issue urgently needed in the communications industry. Summary of the Invention

[0004] In order to solve the problems of the prior art, the present application proposes a base station location prediction method, device, equipment and medium, which uses known base station location data to train a base station location prediction model, and then predicts the location of unknown base stations. Compared with the field measurement method, the error of the location prediction is within an acceptable range, the implementation difficulty is small and the cost is low.

[0005] In a first aspect, a base station location prediction method is provided, comprising:

[0006] Obtaining a measurement data file and a base station table, wherein the measurement data file includes a user identifier and a series of base station identifier switching records of the user, and the base station table includes base station identifiers of known locations and corresponding latitude and longitude information;

[0007] Determine the comprehensive switching diagram corresponding to all users based on the measurement data file;

[0008] Determine the embedding vector of the base station with unknown location based on the comprehensive handover map and the base station table;

[0009] The embedding vector of the unknown base station is input into the trained base station location prediction model to obtain the latitude and longitude prediction values ​​of the unknown base station. The trained base station location prediction model includes:

[0010] The latitude prediction module is used to predict the latitude value of a base station with unknown location based on its embedding vector;

[0011] The longitude prediction module is used to predict the longitude value of a base station with unknown location based on its embedding vector.

[0012] Furthermore, based on the measurement data file, a comprehensive switching diagram corresponding to all users is determined, including:

[0013] Using the measurement data file of each user, a corresponding user handoff graph is constructed based on the base station identification sequence, where the nodes of the user handoff graph represent the base station identifications, and the edges represent the handoff relationships and corresponding weights.

[0014] The switching graphs of all users are merged to obtain a comprehensive switching graph corresponding to all users, wherein the weights of the edges in the comprehensive switching graph integrate the switching information of multiple users.

[0015] Furthermore, based on the comprehensive handover graph and the base station table, determining the embedding vectors of the base stations with known locations and the embedding vectors of the base stations with unknown locations includes:

[0016] Use the Node2Vec graph embedding algorithm to reduce the dimensionality of the comprehensive handover graph and extract the effective feature vectors of the base stations;

[0017] The base station table and the extracted valid feature vectors are used to determine the embedding vectors of the base stations with known locations and the embedding vectors of the base stations with unknown locations.

[0018] Furthermore, the Node2Vec graph embedding algorithm is used to reduce the dimensionality of the comprehensive switching graph and extract the effective feature vectors of the base stations, including:

[0019] Initialize the parameters of the Node2Vec algorithm;

[0020] Perform random walks on the comprehensive switching graph to generate node sequences;

[0021] The generated node sequence is used for training to obtain the effective feature vector of the location base station.

[0022] Furthermore, the base station location prediction model is trained in the following way:

[0023] Obtaining a measurement data file and a base station table, wherein the measurement data file includes a user identifier and a series of base station identifier switching records of the user, and the base station table includes base station identifiers of known locations and corresponding latitude and longitude information;

[0024] Determine the comprehensive switching diagram corresponding to all users based on the measurement data file;

[0025] Determine the embedding vectors of base stations with known locations based on the comprehensive handoff graph and the base station table;

[0026] The base station location prediction model is trained using the embedding vectors of the base stations with known locations and the longitude and latitude information of the base stations with known locations to obtain a trained base station location prediction model.

[0027] In a second aspect, a base station location prediction device is provided, comprising:

[0028] An acquisition unit, configured to acquire a measurement data file and a base station table, wherein the measurement data file includes a user identifier and a series of base station identifier switching records of the user, and the base station table includes base station identifiers of known locations and corresponding latitude and longitude information;

[0029] a switching diagram determining unit, configured to determine a comprehensive switching diagram corresponding to all users based on the measurement data file;

[0030] an embedding vector determining unit, configured to determine an embedding vector of a base station of unknown location based on the comprehensive handover map and the base station table;

[0031] The prediction unit is configured to input the embedding vector of the unknown base station into a trained base station location prediction model to obtain a predicted longitude and latitude value of the unknown base station, wherein the trained base station location prediction model includes:

[0032] The latitude prediction module is used to predict the latitude value of a base station with unknown location based on its embedding vector;

[0033] The longitude prediction module is used to predict the longitude value of a base station with unknown location based on its embedding vector.

[0034] Furthermore, the switching graph determination unit is specifically used to use the measurement data file of each user to construct a corresponding user switching graph based on the base station identification sequence, wherein the user switching graph nodes represent base station identifications, and the edges represent switching relationships and corresponding weights; the switching graphs of all users are merged to obtain a comprehensive switching graph corresponding to all users, wherein the weights of the edges in the comprehensive switching graph integrate the switching information of multiple users.

[0035] Furthermore, the embedding vector determination unit is specifically used to use the Node2Vec graph embedding algorithm to reduce the dimensionality of the comprehensive switching graph and extract the effective feature vectors of the base stations; and use the base station table and the extracted effective feature vectors to determine the embedding vectors of the base stations with known locations and the embedding vectors of the base stations with unknown locations.

[0036] Furthermore, the Node2Vec graph embedding algorithm is used to reduce the dimensionality of the comprehensive switching graph and extract the effective feature vectors of the base stations, including:

[0037] Initialize the parameters of the Node2Vec algorithm;

[0038] Perform random walks on the comprehensive switching graph to generate node sequences;

[0039] The generated node sequence is used for training to obtain the effective feature vector of the location base station.

[0040] Furthermore, it also includes: a training module, which is used to train the base station location prediction model in the following manner: obtaining a measurement data file and a base station table, wherein the measurement data file contains a user identifier and a series of base station identifier switching records of the user, and the base station table contains a base station identifier with a known location and the corresponding longitude and latitude information; according to the measurement data file, determining a comprehensive switching map corresponding to all users; based on the comprehensive switching map and the base station table, determining the embedding vector of the base station with a known location; using the embedding vector of the base station with a known location and the longitude and latitude information of the base station with a known location, training the base station location prediction model to obtain a trained base station location prediction model.

[0041] According to a third aspect, an electronic device is provided, comprising a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to perform the above method.

[0042] In a fourth aspect, a readable storage medium is provided, wherein the readable storage medium stores computer program instructions, and when the computer program instructions are read and executed by a processor, the above method is executed.

[0043] Compared with the prior art, the advantages of the present invention are:

[0044] 1. With the help of big data analysis and base station location prediction models, the accuracy of base station latitude and longitude completion has been improved (the average base station error is <320 meters), providing more reliable data support for communication network optimization and related applications.

[0045] 2. No need for on-site measurement, which greatly saves manpower, material resources and time costs.

[0046] 3. Good scalability and versatility, able to adapt to communication network data of different scales and complexities. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 This is a flow chart of a base station location prediction method in an embodiment of the present invention. DETAILED DESCRIPTION

[0048] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be derived from these drawings without inventive effort.

[0049] The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. The embodiments cannot be described one by one here, but the embodiments of the present invention are not limited to the following embodiments.

[0050] like Figure 1 FIG. 1 is a flow chart of a method for predicting a base station location provided by the present invention, comprising the following steps:

[0051] Step 101: Obtain a measurement data file and a base station table, wherein the measurement data file includes a user identifier and a series of base station identifier switching records of the user, and the base station table includes base station identifiers with known locations and corresponding latitude and longitude information.

[0052] Step 101 is the data preparation phase: first, obtain a measurement data file and a base station table file. Read the measurement data and base station table information from them, obtain a large number of user base station switching records, and extract information related to known base stations from the base station table, such as base station identifiers and corresponding latitude and longitude information.

[0053] Step 102: Determine a comprehensive handover map corresponding to all users based on the measurement data file;

[0054] Specifically, this step 102 can be implemented using the following steps 102-1 and 102-2:

[0055] Step 102-1: Using the measurement data file of each user, construct a corresponding user handoff graph based on the base station identification sequence, where the nodes of the user handoff graph represent base station identifications, and the edges represent handoff relationships and corresponding weights;

[0056] Specifically, we can group measurement data by user ID, sort each data set by time, and filter out data without base station ID changes. We then construct a user handoff graph based on the user's base station ID sequence. If a handoff relationship exists between two base station IDs, we add an edge to the graph, with the weight of the edge increasing based on the number of handoffs.

[0057] This step 102-1 is the user handoff map construction stage: the measurement data of each user is processed. Taking user A as an example, its measurement data contains a series of base station ID handoff records. After sorting and filtering, the user handoff map of user A is constructed based on the base station ID sequence.

[0058] Step 102 - 2 : Merge the switching graphs of all users to obtain a comprehensive switching graph corresponding to all users, wherein the weights of the edges in the comprehensive switching graph integrate the switching information of multiple users.

[0059] This step 102-1 is the user switching graph merging stage: the switching graphs of all users (such as users A, B, C, etc.) are merged to form a complete comprehensive switching graph. The edge weights in the comprehensive switching graph integrate the switching information of multiple users. For example, when merging, if the same edge exists in different user graphs, their weights are accumulated.

[0060] Step 103: Determine the embedding vectors of the base stations with known locations and the embedding vectors of the base stations with unknown locations based on the comprehensive handover map and the base station table;

[0061] This step 103 is the graph embedding stage: the comprehensive handover graph is processed using a graph embedding algorithm, and after relevant parameters are set, an embedding vector corresponding to each base station is obtained.

[0062] This step 103 specifically includes:

[0063] Step 103-1: Use the Node2Vec graph embedding algorithm to reduce the dimension of the comprehensive handover graph and extract the effective feature vectors of the base stations;

[0064] Step 103 - 2 : Using the base station table and the extracted valid feature vectors, determine the embedding vectors of the base stations with known locations and the embedding vectors of the base stations with unknown locations.

[0065] The Node2Vec algorithm is used to reduce the dimension of the comprehensive switching graph and extract the effective feature vectors of the base station, including:

[0066] Initialize the parameters of the Node2Vec algorithm (network node embedding algorithm based on random walk), such as dimension, random walk length, number of random walks, etc.

[0067] Random walks are performed on the comprehensive switching graph to generate node sequences, which are then used to train and obtain effective feature vectors of base stations.

[0068] Step 104: Input the embedding vector of the unknown base station into the trained base station location prediction model to obtain the longitude and latitude prediction values ​​of the unknown base station.

[0069] Among them, the trained base station location prediction model includes:

[0070] The latitude prediction module is used to predict the latitude value of a base station with unknown location based on its embedding vector;

[0071] The longitude prediction module is used to predict the longitude value of a base station with unknown location based on its embedding vector.

[0072] This step 104 is the stage of predicting the coordinates of the unknown base station: extracting the embedding vector of the unknown base station and inputting it into the trained prediction model to obtain the predicted longitude and latitude of the unknown base station.

[0073] The following describes the training process of the base station location prediction model. In summary, the process involves filtering out the embedding vectors of known base stations, matching them with the corresponding longitude and latitude information, and training the latitude prediction model and the longitude prediction model. For example, after multiple iterations of training, the prediction model can accurately learn the relationship between the embedding vector and the longitude and latitude. The prediction model here can be an existing prediction model. The contribution of this application lies in the application of the prediction model in the field of base station geographic location prediction.

[0074] Specifically, the training process of the station location prediction model includes the following steps:

[0075] Step 201: Obtain a measurement data file and a base station table, wherein the measurement data file includes a user identifier and a series of base station identifier switching records of the user, and the base station table includes the identifiers of base stations with known locations and corresponding latitude and longitude information;

[0076] Step 202: Determine a comprehensive handover map corresponding to all users based on the measurement data file;

[0077] Step 203: Determine the embedding vector of the base station with known location based on the comprehensive handover map and the base station table;

[0078] Step 204: Using the embedding vector of the base station with known location and the longitude and latitude information of the base station with known location, a base station location prediction model is trained to obtain a trained base station location prediction model.

[0079] Specifically, in step 204, the embedding vector of the base station with a known location can be used as an input feature, and the latitude and longitude thereof can be used as target variables for training.

[0080] Specifically, the latitude prediction value of the base station with a known location can be determined by the dimension prediction module according to the embedding vector of the base station with a known location, and the latitude prediction module parameters can be optimized based on the minimum mean square error between the latitude prediction value and the known base station latitude until the minimum mean square error reaches a set value;

[0081] The network parameters of the longitude prediction module are determined according to the embedding vectors of the base stations with known locations. Based on the minimum mean square error between the longitude prediction value and the known base station longitude, the parameters of the longitude prediction module are optimized until the minimum mean square error reaches the set value.

[0082] Based on the above method, the present application also provides a base station location prediction device, comprising: an acquisition unit, a switching map determination unit, an embedding vector determination unit, and a prediction unit; wherein:

[0083] An acquisition unit, configured to acquire a measurement data file and a base station table, wherein the measurement data file includes a user identifier and a series of base station identifier switching records of the user, and the base station table includes base station identifiers of known locations and corresponding latitude and longitude information;

[0084] a switching diagram determining unit, configured to determine a comprehensive switching diagram corresponding to all users based on the measurement data file;

[0085] an embedding vector determining unit, configured to determine an embedding vector of a base station of unknown location based on the comprehensive handover map and the base station table;

[0086] The prediction unit is configured to input the embedding vector of the unknown base station into a trained base station location prediction model to obtain a predicted longitude and latitude value of the unknown base station, wherein the trained base station location prediction model includes:

[0087] The latitude prediction module is used to predict the latitude value of a base station with unknown location based on its embedding vector;

[0088] The longitude prediction module is used to predict the longitude value of a base station with unknown location based on its embedding vector.

[0089] Furthermore, the switching graph determination unit is specifically used to use the measurement data file of each user to construct a corresponding user switching graph based on the base station identification sequence, wherein the user switching graph nodes represent base station identifications, and the edges represent switching relationships and corresponding weights; the switching graphs of all users are merged to obtain a comprehensive switching graph corresponding to all users, wherein the weights of the edges in the comprehensive switching graph integrate the switching information of multiple users.

[0090] Furthermore, the embedding vector determination unit is specifically used to use the Node2Vec graph embedding algorithm to reduce the dimensionality of the comprehensive switching graph and extract the effective feature vectors of the base stations; and use the base station table and the extracted effective feature vectors to determine the embedding vectors of the base stations with known locations and the embedding vectors of the base stations with unknown locations.

[0091] The present invention also discloses an electronic device, which includes a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the method of the present application.

[0092] The present invention also discloses a readable storage medium, in which computer program instructions are stored. When the computer program instructions are read and executed by a processor, the method of the present application is executed.

[0093] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A base station location prediction method, characterized in that: include: Obtaining a measurement data file and a base station table, wherein the measurement data file includes a user identifier and a series of base station identifier switching records of the user, and the base station table includes base station identifiers of known locations and corresponding latitude and longitude information; Determine the comprehensive switching diagram corresponding to all users based on the measurement data file; Determine the embedding vector of the base station with unknown location based on the comprehensive handover map and the base station table; The embedding vector of the unknown base station is input into the trained base station location prediction model to obtain the latitude and longitude prediction values ​​of the unknown base station. The trained base station location prediction model includes: The latitude prediction module is used to predict the latitude value of a base station with unknown location based on its embedding vector; The longitude prediction module is used to predict the longitude value of a base station with unknown location based on its embedding vector.

2. The method according to claim 1, characterized in that Based on the measurement data file, determine the comprehensive switching diagram corresponding to all users, including: Using the measurement data file of each user, a corresponding user handoff graph is constructed based on the base station identification sequence, where the nodes of the user handoff graph represent the base station identifications, and the edges represent the handoff relationships and corresponding weights. The switching graphs of all users are merged to obtain a comprehensive switching graph corresponding to all users, wherein the weights of the edges in the comprehensive switching graph integrate the switching information of multiple users.

3. The method according to claim 1 or 2, characterized in that Determining embedding vectors of base stations with known locations and embedding vectors of base stations with unknown locations based on the comprehensive handover graph and the base station table, including: Use the Node2Vec graph embedding algorithm to reduce the dimensionality of the comprehensive handover graph and extract the effective feature vectors of the base stations; The base station table and the extracted valid feature vectors are used to determine the embedding vectors of the base stations with known locations and the embedding vectors of the base stations with unknown locations.

4. The method according to claim 3, characterized in that The Node2Vec graph embedding algorithm is used to reduce the dimensionality of the comprehensive handover graph and extract the effective feature vectors of the base stations, including: Initialize the parameters of the Node2Vec algorithm; Perform random walks on the comprehensive switching graph to generate node sequences; The generated node sequence is used for training to obtain the effective feature vector of the location base station.

5. The method according to claim 1, characterized in that The base station location prediction model is trained in the following way: Obtaining a measurement data file and a base station table, wherein the measurement data file includes a user identifier and a series of base station identifier switching records of the user, and the base station table includes base station identifiers of known locations and corresponding latitude and longitude information; Determine the comprehensive switching diagram corresponding to all users based on the measurement data file; Determine the embedding vectors of base stations with known locations based on the comprehensive handoff graph and the base station table; The base station location prediction model is trained using the embedding vectors of the base stations with known locations and the longitude and latitude information of the base stations with known locations to obtain a trained base station location prediction model.

6. A base station location prediction device, characterized in that: include: An acquisition unit, configured to acquire a measurement data file and a base station table, wherein the measurement data file includes a user identifier and a series of base station identifier switching records of the user, and the base station table includes base station identifiers of known locations and corresponding latitude and longitude information; a switching diagram determining unit, configured to determine a comprehensive switching diagram corresponding to all users based on the measurement data file; an embedding vector determining unit, configured to determine an embedding vector of a base station of unknown location based on the comprehensive handover map and the base station table; The prediction unit is configured to input the embedding vector of the unknown base station into a trained base station location prediction model to obtain a predicted longitude and latitude value of the unknown base station, wherein the trained base station location prediction model includes: The latitude prediction module is used to predict the latitude value of a base station with unknown location based on its embedding vector; The longitude prediction module is used to predict the longitude value of a base station with unknown location based on its embedding vector.

7. The device according to claim 6, characterized in that The handover graph determination unit is specifically configured to use the measurement data file of each user to construct a corresponding user handover graph based on the base station identification sequence, wherein the user handover graph nodes represent base station identifications, and the edges represent handover relationships and corresponding weights; the handover graphs of all users are merged to obtain a comprehensive handover graph corresponding to all users, wherein the weights of the edges in the comprehensive handover graph integrate the handover information of multiple users.

8. The device according to claim 6 or 7, characterized in that The embedding vector determination unit is specifically used to use the Node2Vec graph embedding algorithm to reduce the dimension of the comprehensive switching graph and extract the effective feature vectors of the base stations; and use the base station table and the extracted effective feature vectors to determine the embedding vectors of the base stations with known locations and the embedding vectors of the base stations with unknown locations.

9. An electronic device, characterized in that: The electronic device includes a memory and a processor, the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to perform the method according to any one of claims 1 to 5.

10. A readable storage medium, characterized in that: The readable storage medium stores computer program instructions, and when the computer program instructions are read and executed by a processor, the method according to any one of claims 1 to 5 is executed.

Images