Mobile communication big data positioning method, device, equipment and storage medium

Abstract

The present invention discloses a method, device, equipment and storage medium for positioning big data of mobile communications. The positioning method comprises: obtaining data sent by a mobile terminal to a base station to form a measurement report; preprocessing and data filling the measurement report data; establishing a local coordinate system with the main service base station as the origin, and utilizing a single feature in the measurement report data in combination with an improved deep learning model to complete positioning from a high-dimensional signal space to a two-dimensional geographic space; utilizing the mean absolute error and the root mean square error as two indicators to evaluate positioning accuracy, thereby achieving high-precision positioning of the mobile terminal. The present invention effectively fills in missing data, realizes the mapping between high-dimensional signal space and two-dimensional geographic coordinates, and enhances the positioning accuracy of the terminal. A missing value filling method with hierarchical clustering as the core is proposed, and on this basis, accurate positioning of mobile terminals under small sample conditions is achieved.

Description

Technical Field

[0001] The present invention belongs to the field of positioning and navigation technology, and specifically relates to a mobile communication big data positioning method, device, equipment and storage medium. Background Art

[0002] With the continued development of mobile cellular networks, location-based services (LBS) have attracted the attention of numerous practitioners and researchers. Compared to the positioning services provided by GPS, data-driven positioning technology within mobile cellular networks offers significant advantages. All mobile devices can receive cellular signals, providing a ubiquitous global positioning service. This positioning method can be performed on virtually any mobile device without consuming additional energy. However, due to the significant loss of neighboring cell level values ​​and the scarcity of splines containing location information, positioning performance has been poor, resulting in high computational complexity and low positioning accuracy.

[0003] Therefore, it is crucial to develop a fast, reliable, and robust method to accurately locate mobile terminals. Due to the disparity of mobile devices and the complex communication environment, existing research has been unable to effectively repair the large number of missing values ​​in MR (measurement reports), and most observation data in LTE commercial networks only contain signal strength information from the service cell. The existence of this problem greatly limits the positioning accuracy of cellular networks. In addition, due to privacy and other reasons, a large amount of GPS data information in MR data is missing, and only a small amount of data contains coordinates. This makes the data-driven machine learning-based single-point positioning method have certain limitations. How to complete the positioning of all terminals in the service cell with the help of a small number of samples is currently a key issue. Among them, during each session and call, the measurement data related to the mobile device will be collected by the network. The present invention refers to these measurements as measurement reports. Summary of the Invention

[0004] To address the aforementioned issues with the existing technology, the present invention aims to provide a method, apparatus, device, and storage medium for mobile communication big data positioning. These methods effectively fill missing data, reducing the data missing rate to less than 3%, thereby resolving the data dimensionality problem. Furthermore, the present invention improves the Transformer, enabling a mapping between high-dimensional signal space and two-dimensional geographic coordinates, thereby enhancing the positioning accuracy of the terminal. Taking missing values ​​in MR as a starting point, and based on mobile user profiling technology, a missing value filling method centered on hierarchical clustering is proposed. This method, based on this, enables precise positioning of mobile terminals under small sample conditions.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is:

[0006] A mobile communication big data positioning method, comprising:

[0007] Obtaining data sent by the mobile terminal to the base station to form a measurement report;

[0008] Preprocess and fill measurement report data;

[0009] Establish a local coordinate system with the primary service base station as the origin, and use individual features in the measurement report data combined with an improved deep learning model to complete positioning from high-dimensional signal space to two-dimensional geographic space;

[0010] The positioning accuracy is evaluated using the mean absolute error and root mean square error, achieving high-precision positioning of the mobile terminal.

[0011] As a further improvement of the above technical solution:

[0012] When completing the positioning from high-dimensional signal space to two-dimensional geographic space, a mapping relationship from high-dimensional signal space to two-dimensional geographic space is constructed, and a positioning relationship model with path loss and cell local coordinate values ​​as a triplet is combined with an improved deep learning model to complete the precise positioning of the terminal.

[0013] Preprocessing of measurement report data includes data standardization and data compression to reduce data size and computational complexity;

[0014] When filling data, a hierarchical clustering algorithm is introduced to perform self-filling and mutual filling in the splines of similar clusters. After multiple iterations, the mobile user portrait is filled.

[0015] The single feature in the measurement report data used is the reference signal received power value.

[0016] After a clustering is completed, the self-filling method fills the splines with the same user ID in the same class. After a clustering is completed, the mutual filling method fills the splines with different user IDs in a class.

[0017] When establishing a local coordinate system, all base station and spline coordinates are converted, Gaussian plane coordinates are calculated from geodetic coordinates, the coordinates of the main service base station are set as the origin of the two-dimensional plane coordinate system, the distance and angle between the sample point and the origin are calculated, and the two-dimensional plane coordinates of the sample and base station in the local coordinate system are obtained to complete the construction of the local coordinate system.

[0018] The terminal receives and reports signal values ​​from m cells, uses m cells to construct a high-dimensional signal space, characterizes the spline features through the reported multiple reference signal received power values, and finds m mapping relationships.

[0019] The deep learning model used is the Transformer model. The Transformer architecture stacks multiple identical segments of encoders and decoders together. Both encoder and decoder modules are constructed using multi-head self-attention units and position feedforward networks.

[0020] The positioning method comprises the following steps:

[0021] S1. Standardize and compress the measurement report data to reduce the data size and computational complexity.

[0022] S2. Based on the mobile user profile filling method, the problem of missing a large number of level value data in multiple neighboring cell dimensions in the measurement report data is solved;

[0023] S3. Introduce the Chebyshev distance hierarchical clustering algorithm to perform self-filling and mutual filling in the splines of similar clusters, and complete the filling of mobile user portraits through multiple iterations;

[0024] S4. Unify the coordinates of the terminal user and the base station to establish a local coordinate system with the primary base station as the origin;

[0025] S5. Use the reference signal received power value and location-related base station parameters in the measurement report to construct a mapping relationship from the high-dimensional signal space to the two-dimensional geographic space;

[0026] S6. A positioning relationship model based on a triplet of path loss and local cell coordinate values ​​is combined with an improved Transformer model to accurately position the terminal.

[0027] Step S2 includes the following steps:

[0028] S21, self-filling: self-filling the spline of the same user with a non-zero mean value;

[0029] S22. After the self-filling is completed, the missing values ​​remaining are temporarily filled with a manually set value. When the clustering operation is performed later, the value of the partially filled value is returned to be modified.

[0030] Step S3 includes the following steps:

[0031] S31. In the initial stage, a threshold is manually selected for clustering;

[0032] S32, performing self-filling: filling the splines with the same user ID in the same cluster;

[0033] S33, performing mutual filling: filling the splines of different user IDs in the same cluster;

[0034] S34. Enlarge the threshold by 1.05 times, perform hierarchical clustering again, and iterate repeatedly until the termination condition is reached.

[0035] A mobile communication big data positioning device, comprising:

[0036] The mobile terminal is used to upload data including current location information to the base station;

[0037] Base stations, which provide wireless coverage and are used to connect mobile terminals to the Internet;

[0038] The wireless access network element management system reports the measurement report data of the mobile terminal to the wireless access network element management system for storage;

[0039] The server obtains a plurality of measurement report data from a wireless access network element management system and parses the plurality of measurement report data.

[0040] A positioning device includes a memory and a processor, wherein the memory stores a computer program, and the processor is used to execute the computer program to implement the positioning method.

[0041] A storage medium carries at least one computer program. When the computer program is executed by an electronic device, the electronic device implements the positioning method.

[0042] The beneficial effects of the present invention are:

[0043] (1) Effectively fill in missing data, reducing the data missing rate to less than 3%, solving the data dimensionality disaster, and improving the Transformer to achieve the mapping between high-dimensional signal space and two-dimensional geographic coordinates, thereby enhancing the terminal's positioning accuracy.

[0044] (2) Taking the missing values ​​in MR as the starting point, based on the mobile user portrait technology, a missing value filling method with hierarchical clustering as the core is proposed. On this basis, accurate positioning of mobile terminals under small sample conditions is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 This is a principle block diagram of a mobile terminal positioning method based on measurement reports provided by the present invention;

[0046] Figure 2 It is the MR data visualization picture provided by the present invention;

[0047] Figure 3 It is a cluster filling iterative flow chart provided by the present invention;

[0048] Figure 4 It is a schematic diagram of the local coordinate system provided by the present invention;

[0049] Figure 5This is a diagram of the Transformer coding layer structure provided by the present invention;

[0050] Figure 6 This is a performance comparison picture of the positioning error provided by the present invention and other algorithms. DETAILED DESCRIPTION

[0051] The following describes the specific embodiments of the present invention in detail with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only used to illustrate and explain the present invention and are not intended to limit the present invention.

[0052] For ease of description, spatially relative terms such as "above," "above," "on the upper surface of," and "upper" may be used herein to describe the spatial positional relationship of a device or feature to other devices or features as shown in the figures. It should be understood that spatially relative terms are intended to encompass different orientations of the device in use or operation in addition to the orientation depicted in the figures. For example, if a device in a drawing is inverted, a device described as "above" or "on top of" another device or structure would then be positioned as "below" or "below" the other device or structure. Thus, the exemplary term "above" can include both the "above" and "below" orientations. The device may also be positioned in other different ways (rotated 90 degrees or in other orientations), and the spatially relative descriptions used herein should be interpreted accordingly.

[0053] A mobile communication big data positioning method, such as Figure 1 As shown, the following steps are included:

[0054] Step S1: normalize and compress the measurement report (MR) data to reduce the data size and computational complexity.

[0055] The measurement report includes status information of the terminal when it is connected to the network, such as signal strength, connection time, base station parameters, etc.

[0056] The measurement report (MR) may specifically include the following categories: sample measurement time (TimeStamp) of the sample data measurement object, unique identifier of the UE (user) on the S1 interface of the MME (network node) side (MmeUeSIapID), MME group identifier (MmeGroupID), MME code (Mmecode), reference signal received power of the primary service cell (MR.LteScRSRP), reference signal received power of the neighboring cell (MR.LteNcRSRP), timing advance of the primary service cell (MR.LteScTadv), and carrier number of the primary service cell (MR.LteScEarfcn).

[0057] The standardization and compression of raw MR data and the selection of data labels mainly include the extraction of structured neighborhoods for data and data compression to address data redundancy issues, so that the data can be statistically analyzed from a unified perspective, as well as the selection of data feature labels to prepare for subsequent model positioning.

[0058] Regarding the neighboring cell extraction, since the neighboring cells in the original data are stacked, when processing this part of the data, the cell is used as the processing unit, and all the neighboring cells received by the terminal users under the serving cell are extracted to perform data standardization operations.

[0059] Regarding data compression, due to the measurement reporting mechanism and other reasons, the same terminal will report multiple splines within 1 second. Often, only one of the multiple splines has the neighboring area measurement value reported, and the measurement information of the remaining samples is empty, resulting in a large data scale. In this case, it is necessary to ensure that valid data values ​​are not discarded while performing data compression.

[0060] The data feature label selection of the present invention is to fill in data with the goal of improving the positioning accuracy of the user terminal, and the user's geographical location coordinates are used as the label to be predicted.

[0061] In this embodiment, when extracting neighboring cells, the lte_ncell_eNodeBid_Cellid (neighboring cell identifier) ​​column of a sample reported by a single user at a specific moment reports the neighboring cell detected by the terminal at that moment, and the corresponding cell RSRP value (reference signal received power) is included in the subsequent lte_ncell_rsrp (neighboring cell reference signal received power value) column. The RSRP values ​​of all cells in the lte_ncell_eNodeBid_Cellid column and the corresponding cells in the lte_ncell_rsrp column are extracted. For a sample with CellID 506, the reported neighboring cells are CELL_1019 and CELL_626, with corresponding RSRP values ​​of -82 and -87. In this operation, CELL_1019 and CELL_626 are used as column names, and the corresponding RSRP values ​​are placed under the sample.

[0062] The Z-score normalization process is performed on the unique identifiers of users and cells. The conversion function is as follows:

[0063] ;

[0064] Where x is the reference signal received power RSRP. is the mean of the original data, is the standard deviation, and the processed data is visualized as follows Figure 2As shown in the figure, the z-axis is the normalized user ID, the y-axis is the normalized cell ID, and the x-axis is the RSRP value received by the user in the corresponding cell. As can be seen from the figure, the RSRP values ​​are concentrated in a certain value range.

[0065] Taking cell CELL_992 as an example, considering the limited base station coverage and the influence of noise, a distance of 700 meters was selected as the working parameter. Neighboring cells within a 700m radius centered on the primary serving cell were selected as the feature dimension, with 70 neighboring cells within a 700m radius. The intersection of reported neighboring cells and those within the radius yielded 24 neighboring cells. Taking cell CELL_1019 as another example, all neighboring cells reported by users in CELL_1019 within 24 hours were denoted as Set A, and the cells within the coverage area of ​​CELL_1019's base station were denoted as Set B. The intersection of Sets A and B was taken. Set A: 112 neighboring cells reported by users in CELL_1019. Set B: 31 cells within a 700m radius, with the coordinates of CELL_1019's base station as the origin. The intersection of A and B yielded 16 cells.

[0066] Step S2: Based on the mobile user portrait filling method, the problem of missing a large amount of level value data in multiple neighboring area dimensions in the MR data is processed.

[0067] This step is the filling before hierarchical clustering, which can be called preliminary filling.

[0068] The main reasons for data loss include: physical failure of the data receiving end, data loss during the transmission process, and improper collection process; the two stages of data loss refer to the batch or unit stage and the data item stage.

[0069] Filling out mobile user profiles requires selecting user tags. Feature tags are ultimately selected based on communication mechanisms and user-reported MR data to accurately reflect user characteristics.

[0070] In this step, the filling method used includes two steps:

[0071] Step S21: Self-filling. Specifically, each spline reported by a UE only reports a few or a single neighboring area. If clustering is performed directly, using different clustering thresholds under the Euclidean distance for hierarchical clustering, multiple splines reported by the same user at the same geographic location will be clustered in different clusters. Therefore, in this step, before clustering, the splines of the same UE are self-filled with a non-zero mean value.

[0072] Step S22: After the self-filling is completed, temporarily use a manually set value to fill in the missing values. When the clustering operation is performed later, return to modify the value of the partially filled value. In this step, -145 is temporarily used for filling.

[0073] Step S3: Introduce the Chebyshev distance hierarchical clustering algorithm to perform self-filling and mutual filling in the splines of similar clusters, and iterate multiple times to complete the filling of mobile user portraits.

[0074] MR data filling uses a hierarchical clustering iterative algorithm. Filling includes self-filling and mutual filling. Self-filling fills the splines with the same user ID in the same cluster after clustering. Mutual filling fills the splines with different user IDs in the same cluster after clustering. Both filling methods use non-zero mean.

[0075] In hierarchical clustering, considering that MR data may fail in high-dimensional signal space, the study considered a variety of distance metrics and finally selected Chebyshev distance for clustering. The Calinski-Harabasz (CH) index was used to measure the clustering quality. The distance between sample points in the present invention is calculated in high-dimensional signal space. Assuming that there are two data points in high-dimensional signal space, , .

[0076] The Chebyshev distance d and CH can be described by the following expressions:

[0077] ;

[0078] ;

[0079] is the intra-class dispersion, Represents the number of MR samples, is the number of clusters obtained after hierarchical clustering, is the covariance matrix between the categories of each class, It represents the trace of the matrix. The larger the CH value, the more dispersed the clusters are.

[0080] The hierarchical clustering algorithm is used to accurately characterize mobile user features. The specific steps are as follows:

[0081] 1) In the initial stage, a smaller threshold is selected for clustering.

[0082] 2) Perform self-filling: fill the splines with the same user ID in the same cluster.

[0083] 3) Mutual filling: fill the splines of different user IDs in the same cluster.

[0084] 4) Increase the threshold by 1.05 times and perform hierarchical clustering again, iterating back and forth until the termination condition is reached.

[0085] In this example, scipy (scipy is an open source Python algorithm library and mathematical toolkit) is used to perform hierarchical clustering. For this hierarchical clustering, the main parameters that need to be selected are:

[0086] Connection standard: Single (a connection standard in the scipy library) is selected, and the minimum distance between samples of different classes is used as the standard;

[0087] Distance threshold: If the connection distance is greater than this threshold, the two categories will not be merged;

[0088] The measurement standard for calculating connections: select "chebyshev" (Chebyshev distance);

[0089] In the initial stage, we experimented with a small number of splines, using some industrial parameters to adjust the clustering threshold, and used the CH (Calinski-Harabasz) metric to measure clustering quality. The CH metric is used when the true labels of the data are unknown and evaluates the quality of clustering. A higher CH score indicates better clustering. Specifically, this metric is the ratio of the between-class dispersion to the within-class dispersion, calculated as follows:

[0090] ;

[0091] After a clustering is completed, the splines with the same user ID in the same cluster are filled with the non-zero mean of all RSRP values ​​under the same neighborhood dimension (feature).

[0092] After a clustering is completed, the splines of different user IDs in a class are filled. The mutual filling is performed after the sample self-filling is completed. Combining the clustering results and the self-filling data, non-zero mean filling is also performed.

[0093] The threshold is enlarged by 1.05 times, hierarchical clustering is performed again, and missing values ​​are filled. This iteration is repeated until the termination condition is reached, that is, whether the threshold reaches 24. The flowchart of the algorithm is as follows Figure 3 shown.

[0094] In the results of the initial hierarchical clustering using Chebyshev distance, all samples can be clustered into one cluster when the distance threshold is 35dB. After clustering, we found that by continuously iteratively filling in the gaps and reaching a threshold of 24dB, most of the missing data can be basically filled.

[0095] Step S4: Unify the coordinates of the terminal user and the base station, and establish a local coordinate system with the primary service base station as the origin.

[0096] The local coordinate system is independent of the global coordinate system and is a standard coordinate system established within the service range of the primary base station. In addition, the cells of each base station can be associated through Gaussian plane coordinates.

[0097] The terminal receives and reports signal values ​​from multiple cells. The position of the jth terminal is calculated through the mapping relationship between m cells (including the main cell) and the terminal:

[0098] ;

[0099] in, is the coordinates of the base station corresponding to the i-th cell received by the terminal within the coverage of the current primary serving cell. That is, the real coordinates of the terminal. It is The district and The path loss between terminals.

[0100] In the above formula, Represents a mapping function.

[0101] In this embodiment, the local coordinate system is as follows Figure 4 As shown in Figure 1, this local coordinate system consists of an x-axis and a y-axis. In the figure, red represents the base station, and blue represents the spline. Both axes are constructed in meters, and both the terminal coordinates and the base station coordinates are GPS coordinates.

[0102] Convert all base station and spline coordinates from geodetic coordinates , calculate the Gaussian plane coordinates The Gaussian direct calculation formula is as follows:

[0103] ;

[0104] ;

[0105] In the formula is the meridian arc length line from the equator to the projection point;

[0106] is the radius of the Maoyou circle;

[0107] ; For the Sutra difference; is the central meridian longitude.

[0108] 、 And the following used They are the first eccentricity, major radius and minor radius, which are the commonly used geometric parameters in the Earth ellipsoid.

[0109] is the second squared eccentricity term.

[0110] The calculation formula is as follows:

[0111] ;

[0112] in:

[0113] ;

[0114] ;

[0115] After converting to Gaussian plane coordinates, the coordinates of the main service base station are set to the origin (0,0) in the two-dimensional plane coordinate system. The distance and angle between the sample point and the origin are calculated to obtain the two-dimensional plane coordinates of the sample and the base station in the local coordinate system, completing the construction of the local coordinate system.

[0116] Step S5: Use the RSRP (reference signal received power) value in the MR and the location-related base station parameters to construct a mapping relationship from the high-dimensional signal space to the two-dimensional geographic space.

[0117] In this step, the improved neural network model is combined to realize the mapping between high-dimensional signal space and two-dimensional geographic coordinates, thereby improving the positioning accuracy of the terminal.

[0118] In this embodiment, the coordinates of the base station and the corresponding transmit power are known. The i-th path loss in the current primary serving cell is described as:

[0119] ;

[0120] in, is the path loss, Indicates the The reference signal transmission power of the base station corresponding to each cell, Indicates that the terminal receives the Under the premise of ensuring positioning accuracy, a method is proposed to adapt to the current cell coverage. The relationship between the signal propagation distance and the corresponding path loss can be described as follows:

[0121] ;

[0122] in, The terminal receives the first The coordinates of the base station corresponding to each cell, That is, the real coordinates of the terminal.

[0123] ;

[0124] It is The district and The path loss between the terminals is calculated. The triplet of cells is used as input for deep learning models (i.e., Transformer models). The triplet is the basic unit of knowledge representation, in the form of (subject, relation, object), and is used to describe the relationship between two entities (or an entity and an attribute).

[0125] After data cleaning and feature selection, The high-dimensional signal space is constructed by multiple cells, and the spline features are accurately described by multiple reported RSRP values ​​to find the The mapping relationship is used to accurately locate the terminal position. The objective function is defined as follows:

[0126] ;

[0127] minimize spline The mapping error of each cell is calculated to obtain the optimal position solution.

[0128] Step S6: propose a positioning relationship model with path loss and cell local coordinate value as a triplet, and combine it with the improved Transformer model to complete the precise positioning of the terminal.

[0129] The deep learning-based wireless positioning method uses an improved Transformer model. During modeling, only the Encoder part of the original model is used, and the Decoder part is directly changed to a fully connected layer. Finally, the tensor is projected into the [batch_size, output_len] format.

[0130] In this embodiment, the Transformer architecture stacks together N identical fragments of the following two units: encoder and decoder.

[0131] The encoder and decoder modules are mainly constructed using multi-head self-attention units and positional feed-forward networks. The first sub-layer is a multi-head self-attention module that allows the encoder to focus on the most relevant information of the sequence. The MHSA (multi-head self-attention) output vector Add to the position embedding tensor via residual connection and further processed by the normalization layer as shown in the following formula:

[0132] ;

[0133] ;

[0134] Represents layer normalization, normalizing the residual MHSA output vector It is fed into the position feed-forward network and normalized to produce the output of the encoder as follows:

[0135] ;

[0136] ;

[0137] FFN stands for point-wise feed-forward network. Represents the output of the FFN module. Represents the encoder. The Transformer encoder is constructed by repeatedly stacking N MHSA and FFN sublayers, as well as residual connections and normalization. The Transformer encoding layer structure is as follows Figure 5 As shown, Figure 5 middle, 、 、 、 They are the input signals of the encoder respectively. 、 、 、 They are the output signals of the encoder respectively.

[0138] Similar to the encoder, the decoder also consists of N identical blocks connected by three sub-layers. The first sub-layer is the masked Multi-Head Self-Attention (MHSA) module, which operates similarly to the MHSA module discussed previously, except that future positions in the sequence are masked.

[0139] The triplet features are normalized and then fed into the model. This experiment uses Min-Max normalization to process the data, which maps the data values ​​to between [0, 1]. The processing function is as follows:

[0140] ;

[0141] in, is the minimum value of the feature (triplet feature under the current cell dimension), is the maximum value under this feature.

[0142] The traditional SVR (Support Vector Regression Algorithm), KNN (K-Nearest Neighbor Algorithm), AdaBoost (Adaptive Boosting Algorithm) improved regression positioning reconstruction algorithm and CRL (Conditional Random Field Algorithm) algorithm were compared in experiments, using triple features as input. The algorithm comparison results are as follows: Figure 6 shown. Figure 6 Among them, K-nearest neighbor algorithm, support vector regression algorithm, conditional random field algorithm, adaptive enhancement algorithm, and self-attention model algorithm are different existing training algorithms.

[0143] This invention uses a missing value filling method based on hierarchical clustering to reduce the data in the original ultra-high-dimensional signal space to a lower level, thus avoiding the dimensionality curse. It also uses the Chebyshev distance as a measure of inter-cluster distance, using the hierarchical clustering results as a benchmark to fill in the data within clusters. The clustering threshold is then iteratively increased to complete the filling of missing data. This method, based on data-driven analysis and using the RSRP feature in MR data combined with a modified Transformer model, achieves positioning from the high-dimensional signal space to the two-dimensional geographic space in a local coordinate system. Consequently, this invention improves the positioning capabilities of mobile terminals.

[0144] Finally, it is necessary to explain here that the above embodiments are only used to further illustrate the technical solution of the present invention in detail and cannot be understood as limiting the scope of protection of the present invention. Some non-essential improvements and adjustments made by technicians in this field based on the above content of the present invention all fall within the scope of protection of the present invention.

Claims

1. A mobile communication big data positioning method, characterized in that: include: Obtaining data sent by the mobile terminal to the base station to form a measurement report; The measurement report data is preprocessed and filled with data. When filling the data, the Chebyshev distance hierarchical clustering algorithm is introduced to perform self-filling and mutual filling in the splines of the same cluster, and multiple iterations are performed to complete the filling of the mobile user portrait; a local coordinate system with the main service base station as the origin is established, and the single feature in the measurement report data is combined with the improved deep learning model to complete the positioning from the high-dimensional signal space to the two-dimensional geographic space. The deep learning model used is the Transformer model, and the single feature in the measurement report data used is the reference signal received power value. When establishing a local coordinate system, all base station and spline coordinates are converted, Gaussian plane coordinates are calculated from geodetic coordinates, the coordinates of the master base station are set as the origin of the two-dimensional plane coordinate system, the distance and angle between the sample point and the origin are calculated, and the two-dimensional plane coordinates of the sample and base station in the local coordinate system are obtained, completing the construction of the local coordinate system. When completing positioning from high-dimensional signal space to two-dimensional geographic space, a mapping relationship is constructed from high-dimensional signal space to two-dimensional geographic space. A positioning relationship model using path loss and local cell coordinate values ​​as a triplet is combined with an improved deep learning model to accurately locate the terminal. The positioning accuracy is evaluated using the mean absolute error and root mean square error, achieving high-precision positioning of the mobile terminal.

2. The positioning method according to claim 1, wherein: Preprocessing of measurement report data includes data standardization and data compression to reduce data size and computational complexity.

3. The positioning method according to claim 2, wherein: After a clustering is completed, the self-filling method fills the splines with the same user ID in the same class. After a clustering is completed, the mutual filling method fills the splines with different user IDs in a class.

4. The positioning method according to claim 1, wherein: The terminal receives and reports signal values ​​from m cells, uses m cells to construct a high-dimensional signal space, characterizes the spline features through the reported multiple reference signal received power values, and finds m mapping relationships.

5. The positioning method according to claim 2, wherein: The Transformer architecture stacks multiple identical segments of encoders and decoders together, and both encoder and decoder modules are built using multi-head self-attention units and positional feed-forward networks.

6. A mobile communication big data positioning device, used to implement the positioning method according to any one of claims 1 to 5, characterized in that: include: The mobile terminal is used to upload data including current location information to the base station; Base stations, which provide wireless coverage and are used to connect mobile terminals to the Internet; The wireless access network element management system reports the measurement report data of the mobile terminal to the wireless access network element management system for storage; The server obtains a plurality of measurement report data from a wireless access network element management system and parses the plurality of measurement report data.

7. A positioning device, characterized in that: The device comprises a memory and a processor, wherein the memory stores a computer program, and the processor is used to execute the computer program to implement the positioning method according to any one of claims 1 to 5.

8. A storage medium carrying at least one computer program, characterized in that: When the computer program is executed by an electronic device, the electronic device implements the positioning method described in any one of claims 1 to 5.

Images