Mobile signaling trajectory data generation method and system based on diffusion process

Abstract

The invention provides a mobile signaling trajectory data generation method and system based on a diffusion process, and the method comprises the steps: obtaining original mobile signaling data in a preset time period, the original mobile signaling data comprising mobile signaling data corresponding to each base station, and carrying out the preprocessing of the original mobile signaling data; a base station relation graph is determined based on the mobile signaling data of each base station, the base station relation graph comprises a switching relation between the base stations, and an initial data set of each base station is constructed based on the switching relation of each base station and the feature data of the base station; constructing a representation vector corresponding to each base station on the basis of the initial data set of each base station, combining the representation vectors of the base stations into an input vector, inputting the input vector into a pre-trained diffusion model, outputting a track generation vector by the diffusion model, and enabling the track generation vector to be composed of a plurality of track generation sub-vectors corresponding to users; and determining trajectory data of the user in a preset time period based on the trajectory generation vector.

Description

Technical Field

[0001] The present invention relates to the technical field of trajectory generation, and in particular to a method and system for generating mobile signaling trajectory data based on a diffusion process. Background Art

[0002] The generation of trajectory data depends on a variety of sensors, algorithms, and application scenarios. It is a complex and sophisticated process. The real trajectory data is mainly obtained by sampling the movement of the moving object in the spatiotemporal environment. The sampling points contain key information such as the position, timestamp, and speed of the moving object. Connecting the sampling points according to the sampling order constitutes the trajectory data.

[0003] However, existing trajectory generation technologies are often designed for continuous trajectory data generated by positioning devices such as GPS. However, trajectory data based on GPS requires user authorization and has a low coverage rate, which limits its scope of application. Mobile signaling data is collected by mobile operators and usually has the characteristics of a large user base, wide coverage, authenticity, objectivity, and strong real-time performance. It has generated huge application value in the fields of population monitoring, public safety, and smart transportation. However, due to personal privacy and data security risks, the availability of real individual mobile trajectory data based on mobile signaling in practical applications is still greatly limited. Summary of the invention

[0004] In view of this, an embodiment of the present invention provides a method for generating mobile signaling trajectory data based on a diffusion process to eliminate or improve one or more defects existing in the prior art.

[0005] One aspect of the present invention provides a method for generating mobile signaling trajectory data based on a diffusion process, the method comprising the steps of:

[0006] Acquire original mobile signaling data of a preset time period, the original mobile signaling data including mobile signaling data corresponding to each base station, and pre-process the original mobile signaling data;

[0007] Determine a base station relationship graph based on mobile signaling data of each base station, wherein the base station relationship graph includes a switching relationship between base stations, and construct an initial data set for each base station based on the switching relationship of each base station and feature data of the base station;

[0008] constructing a representation vector corresponding to each base station based on an initial data set of each base station, combining the representation vectors of each base station into an input vector and inputting it into a pre-trained diffusion model, wherein the diffusion model outputs a trajectory generation vector, and the trajectory generation vector is composed of a plurality of trajectory generation sub-vectors corresponding to the user;

[0009] The trajectory data of the user in a preset time period is determined based on the trajectory generation vector.

[0010] The above scheme is adopted. This scheme takes the mobile signaling data of the base station as the basis. First, the mobile signaling data is preprocessed to delete invalid data such as duplicate data, and a base station relationship graph including the switching relationship between base stations is constructed. The switching relationship between base stations is the relationship in which the user switches from the connection of one base station to the connection of another base station, and a representation vector corresponding to each base station is constructed. Then, the trajectory generation vector is output through the diffusion model, and the user's trajectory data is obtained. This scheme can obtain a coarse-grained user mobile trajectory based on the mobile signaling data of the base station.

[0011] In some implementations of the present invention, in the step of preprocessing the original mobile signaling data:

[0012] Dividing the original mobile signaling data of a preset time period into mobile signaling data of a plurality of sub-time periods;

[0013] The base station to which the user is connected in each sub-time period is determined based on the mobile signaling data in the sub-time period.

[0014] In some embodiments of the present invention, in the step of determining the base station to which the user is connected in each sub-time period based on the mobile signaling data in the sub-time period, the base station with the longest connection time for each user in the sub-time period is counted as the base station to which the user is connected in the sub-time period.

[0015] In some embodiments of the present invention, in the step of preprocessing the original mobile signaling data, the preprocessing includes sorting according to time, removing ping-pong data, removing duplicate data and removing noise data.

[0016] In some embodiments of the present invention, in the step of determining the base station relationship graph based on the mobile signaling data of each base station, each base station is regarded as a node in the base station relationship graph, and if there is a switching relationship between two base stations, an edge between the two base stations is constructed.

[0017] In some embodiments of the present invention, the characteristic data of the base station includes information such as the location area identification code, the cell identification code, and the longitude and latitude of the base station. In the step of constructing a representation vector corresponding to each base station based on the initial data set of each base station:

[0018] Combining the characteristic data of the base stations in the initial data set with the switching relationship of each base station and encoding them;

[0019] Get the representation vector corresponding to each base station.

[0020] In some embodiments of the present invention, the value of each dimension in the trajectory generation vector corresponds to a base station, and in the step of determining the trajectory data of the user in a preset time period based on the trajectory generation vector, the trajectory data of the user for the base station is determined based on the order of the dimensions in the trajectory generation vector.

[0021] In some embodiments of the present invention, the method further comprises pre-training the diffusion model, in which:

[0022] Calculate the prediction noise loss, restoration loss and base station switching constraint loss based on the output trajectory generation vector and label data;

[0023] A total loss function is calculated based on the predicted noise loss, the restoration loss and the base station switching constraint loss, and the diffusion model is pre-trained based on the total loss function.

[0024] In some embodiments of the present invention, in the step of calculating the predicted noise loss, the restoration loss and the base station switching constraint loss based on the output trajectory generation vector and the label data:

[0025] Calculate the mean square error between the predicted noise and white noise as the prediction noise loss;

[0026] Calculate the mean square error between the output trajectory generation vector and the input vector as the recovery loss;

[0027] Traverse the adjacent base stations in the trajectory generation vector. If two adjacent base stations have an edge in the base station relationship graph, the loss value is 0; if two adjacent base stations do not have an edge in the base station relationship graph, the loss value is 1. The loss values ​​are accumulated to obtain the base station switching constraint loss.

[0028] The second aspect of the present invention also provides a mobile signaling trajectory data generation system based on a diffusion process, the system comprising a computer device, the computer device comprising a processor and a memory, the memory storing computer instructions, the processor being used to execute the computer instructions stored in the memory, and when the computer instructions are executed by the processor, the system implements the steps implemented by the method described above.

[0029] The third aspect of the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps implemented by the aforementioned method for generating mobile signaling trajectory data based on a diffusion process.

[0030] Additional advantages, purposes, and features of the present invention will be described in part in the following description, and will become apparent to those skilled in the art after studying the following, or may be learned from the practice of the present invention. The purposes and other advantages of the present invention can be specifically pointed out and obtained in the specification and the accompanying drawings.

[0031] Those skilled in the art will appreciate that the objectives and advantages that can be achieved with the present invention are not limited to the above specific description, and the above and other objectives that can be achieved by the present invention will be more clearly understood from the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] The drawings described herein are used to provide a further understanding of the present invention, constitute a part of this application, and do not constitute a limitation of the present invention.

[0033] Figure 1 A schematic diagram of an implementation of a method for generating mobile signaling trajectory data based on a diffusion process of the present invention;

[0034] Figure 2 Schematic diagram of the processing architecture of the method for generating mobile signaling trajectory data based on the diffusion process of the present invention. DETAILED DESCRIPTION

[0035] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments and the accompanying drawings. Here, the illustrative embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention.

[0036] It should also be noted that, in order to avoid obscuring the present invention due to unnecessary details, only structures and / or processing steps closely related to the solutions according to the present invention are shown in the accompanying drawings, while other details that are not closely related to the present invention are omitted.

[0037] In the specific implementation process, mobile signaling data usually has the characteristics of a large user base, wide coverage of the population, authenticity and objectivity, and strong real-time performance, and has generated huge application value in the fields of population monitoring, public safety, and smart transportation. However, due to personal privacy and data security risks, the availability of real individual mobile trajectory data based on mobile signaling in practical applications is still very limited. At the same time, due to the instability of communication, storage, and equipment performance, the original mobile signaling trajectory data may have a large amount of redundancy, missing data, and noise, which also seriously affects the usability of the data. In this context, constructing a mobile signaling trajectory data generation model to generate synthetic data that is close to the real data at the statistical level and can replace the real data at the application level is one of the effective solutions to the above problems. However, most of the existing technologies are only for GPS trajectory data, and there is currently no trajectory data generation technology for mobile signaling. In response to the above problems, the present invention proposes a trajectory data generation method for mobile signaling.

[0038] like Figure 1 and 2 As shown, the present invention proposes a method for generating mobile signaling trajectory data based on a diffusion process, the steps of the method comprising:

[0039] Step S100, obtaining original mobile signaling data of a preset time period, the original mobile signaling data including mobile signaling data corresponding to each base station, and preprocessing the original mobile signaling data;

[0040] In the specific implementation process, mobile signaling data is the data generated during the communication between mobile phone users and base stations, mainly including information such as user location, call start and stop, text message sending and receiving, etc. These data have broad application value in many fields such as network planning, user behavior analysis, and traffic flow monitoring.

[0041] Step S200, determining a base station relationship diagram based on mobile signaling data of each base station, wherein the base station relationship diagram includes a switching relationship between base stations, and constructing an initial data set for each base station based on the switching relationship of each base station and feature data of the base station;

[0042] In a specific implementation process, the characteristic data of the base station includes information such as the location area identification code LAC (Local Area Code), the cell identification code CI (Cell Identity), and the longitude and latitude of the base station.

[0043] Step S300, constructing a representation vector corresponding to each base station based on the initial data set of each base station, combining the representation vectors of each base station into an input vector and inputting it into a pre-trained diffusion model, wherein the diffusion model outputs a trajectory generation vector, and the trajectory generation vector is composed of a plurality of trajectory generation sub-vectors corresponding to the user;

[0044] In the specific implementation process, the diffusion model includes a forward noise addition module and a reverse noise removal module:

[0045] Forward Noise Adding Module: Use the diffusion model to gradually add noise to the trajectory representation vector to obtain the noise vector after adding noise;

[0046] Reverse denoising module: First, the noise vector obtained by the forward diffusion denoising module is used to predict the noise through denoising network learning, and then the initial noise vector is restored based on the predicted noise and the noise vector obtained by the forward diffusion module.

[0047] In the specific implementation process, given any time step, a predicted noise is generated according to the trained denoising network, and then the cosine similarity between the predicted noise and the base station representation vector is calculated, and the generated trajectory generation vector is obtained using the argmax operator.

[0048] The argmax operator is a concept widely used in mathematical optimization, statistical analysis, and machine learning. Its core function is to find and return the independent variable value or index that can make a given function reach the maximum value. When processing a series of numerical or function values, the argmax operator accurately locates the position of the maximum value by comparing the sizes of these values. In the context of arrays or lists, the argmax operator returns the index with the largest element, which is particularly important for fast retrieval and analysis of data. In the scenario of function optimization, the argmax operator helps researchers lock in the parameter combination that makes the objective function achieve the extreme value, thereby promoting decision making and model optimization. It is worth noting that the argmax operator does not directly provide the maximum value itself, but indicates the location of the maximum value. This feature makes it irreplaceable in the fields of data science and machine learning.

[0049] In the specific implementation process, in the processing steps of the forward noise addition module:

[0050] First, set the initial values ​​of the model parameters, such as the diffusion step T, the noise β t , where β t It follows a standard normal distribution with a mean of 0 and a variance of 1.

[0051] Suppose the initial input (i.e., input vector) of a trajectory T is represented as x0, randomly sample a diffusion step t from 1 to T, add noise to x0, and generate noise data x according to the following formula t :

[0052]

[0053] where ∈ tis white noise sampled from a standard normal distribution N(0,I), where I is the identity matrix with the same dimensions as the input x0, and α t =1-β t

[0054]

[0055] In the processing steps of the inverse denoising module:

[0056] From the noisy data x T At the beginning, the noise is gradually removed by T time steps. At each step t, the denoising network DN learns to predict the noise ε θ (x t ,t), that is: ε θ (x t ,t)=DN θ (x T ), where θ is a learnable parameter. The present invention uses a model architecture related to sequence generation such as Diffwave as a denoising network.

[0057] According to the predicted noise ε θ (x t ,t) and the current noise data x t Let's try to gradually restore x t-1 , let the restored x t-1 Expressed as It can be obtained by the following formula:

[0058]

[0059] The recovered x0 is

[0060] Calculate the recovered The cosine similarity with the base station representation vector is obtained by using argmax to generate the base station trajectory as a sequence

[0061] Step S400: determining the trajectory data of the user in a preset time period based on the trajectory generation vector.

[0062] The above scheme is adopted. This scheme takes the mobile signaling data of the base station as the basis. First, the mobile signaling data is preprocessed to delete invalid data such as duplicate data, and a base station relationship graph including the switching relationship between base stations is constructed. The switching relationship between base stations is the relationship in which the user switches from the connection of one base station to the connection of another base station, and a representation vector corresponding to each base station is constructed. Then, the trajectory generation vector is output through the diffusion model, and the user's trajectory data is obtained. This scheme can obtain a coarse-grained user mobile trajectory based on the mobile signaling data of the base station.

[0063] In some implementations of the present invention, in the step of preprocessing the original mobile signaling data:

[0064] Dividing the original mobile signaling data of a preset time period into mobile signaling data of a plurality of sub-time periods;

[0065] The base station to which the user is connected in each sub-time period is determined based on the mobile signaling data in the sub-time period.

[0066] In some embodiments of the present invention, in the step of determining the base station to which the user is connected in each sub-time period based on the mobile signaling data in the sub-time period, the base station with the longest connection time for each user in the sub-time period is counted as the base station to which the user is connected in the sub-time period.

[0067] In some embodiments of the present invention, the preset time period is divided into H time periods, H=288, that is, one sub-time period every 5 minutes, and the base station where the user stays the longest in 5 minutes is extracted as the position in the time period.

[0068] In some embodiments of the present invention, in the step of preprocessing the original mobile signaling data, the preprocessing includes sorting according to time, removing ping-pong data, removing duplicate data and removing noise data.

[0069] In some embodiments of the present invention, ping-pong data refers to data generated by frequent switching of a mobile phone between two or more base stations due to changes in base station signal strength in a mobile communication network. This type of data appears to indicate that the user's location has changed, but in fact the user's location has not moved at all, thus interfering with data analysis and needing to be identified and eliminated.

[0070] The method of removing ping-pong data usually includes the following steps:

[0071] Data preprocessing: Preprocess the collected signaling data, including data cleaning, format conversion, etc., to ensure the accuracy and consistency of the data.

[0072] Identify the ping-pong effect: By analyzing the base station switching records in the signaling data, identify the data segments with the ping-pong effect. This can usually be achieved by calculating the time interval and frequency of switching between adjacent base stations. If the switching frequency is too high (such as exceeding a certain threshold), it is considered that the ping-pong effect exists.

[0073] Eliminate ping-pong data: For the identified ping-pong data segments, you can choose to directly eliminate them or perform smoothing. Direct elimination means deleting the data segments with ping-pong effects from the data set; smoothing means correcting the ping-pong data through an algorithm to obtain data that is closer to the user's actual movement trajectory.

[0074] In some embodiments of the present invention, in the step of determining the base station relationship graph based on the mobile signaling data of each base station, each base station is regarded as a node in the base station relationship graph, and if there is a switching relationship between two base stations, an edge between the two base stations is constructed.

[0075] In a specific implementation process, the base station relationship diagram can be represented by an adjacency matrix. Specifically, the topological relationship can be represented by an adjacency matrix A. In the adjacency matrix, if there is an edge between two base stations, the values ​​at the corresponding positions of the two base stations are set to 1, otherwise they are set to 0.

[0076] In some embodiments of the present invention, the characteristic data of the base station includes information such as the location area identification code, the cell identification code, and the longitude and latitude of the base station. In the step of constructing a representation vector corresponding to each base station based on the initial data set of each base station:

[0077] Combining the characteristic data of the base stations in the initial data set with the switching relationship of each base station and encoding them;

[0078] Get the representation vector corresponding to each base station.

[0079] In the specific implementation process, the unique hot encoding of the base station location area identification code LAC is used to represent the data of the base station location area identification code. The dimension of the base station feature data can be the number of different LACs, and other information such as the latitude and longitude information address of the base station can also be added as a supplement to the base station features.

[0080] In the specific implementation process, in the step of combining the feature data of the base stations in the initial data set with the switching relationship of each base station and encoding them, a preset encoder is used for encoding. Specifically, the encoder can be a graph autoencoder, a graph variational autoencoder, and a graph attention autoencoder.

[0081] In the specific implementation process, unsupervised learning graph neural networks such as Graph Auto-Encoders GAE (Graph Auto-Encoders), Graph Variational Auto-Encoders VGAE (Variational Graph Auto-Encoders), and Graph Attention Auto-Encoders GATE (Graph Attention Auto-Encoders) are used to learn the vector of each base station and convert the base station into a continuous representation vector. Where d is the dimension of the feature.

[0082] In some embodiments of the present invention, the value of each dimension in the trajectory generation vector corresponds to a base station, and in the step of determining the trajectory data of the user in a preset time period based on the trajectory generation vector, the trajectory data of the user for the base station is determined based on the order of the dimensions in the trajectory generation vector.

[0083] In some embodiments of the present invention, the method further comprises pre-training the diffusion model, in which:

[0084] Calculate the prediction noise loss, restoration loss and base station switching constraint loss based on the output trajectory generation vector and label data;

[0085] A total loss function is calculated based on the predicted noise loss, the restoration loss and the base station switching constraint loss, and the diffusion model is pre-trained based on the total loss function.

[0086] In a specific implementation process, in the step of calculating the total loss function based on the predicted noise loss, restoration loss and base station switching constraint loss, the sum of the predicted noise loss, restoration loss and base station switching constraint loss is calculated as the total loss function.

[0087] In some embodiments of the present invention, in the step of calculating the predicted noise loss, the restoration loss and the base station switching constraint loss based on the output trajectory generation vector and the label data:

[0088] Calculate the mean square error between the predicted noise and white noise as the prediction noise loss;

[0089] Calculate the mean square error between the output trajectory generation vector and the input vector as the recovery loss;

[0090] Traverse the adjacent base stations in the trajectory generation vector. If two adjacent base stations have an edge in the base station relationship graph, the loss value is 0; if two adjacent base stations do not have an edge in the base station relationship graph, the loss value is 1. The loss values ​​are accumulated to obtain the base station switching constraint loss.

[0091] In the specific implementation process, since there is a switching spatial relationship between base stations, in order to ensure that the adjacent base stations in the generated trajectory data still retain the original switching spatial relationship, the base station switching spatial relationship constraint is added. The model loss is divided into three parts: prediction noise loss Recovering losses and base station switching constraint loss

[0092] in is the predicted noise ε θ (x t ,t) and white noise∈ t The mean square error between them is expressed as:

[0093]

[0094] For recovery The mean square error with the original x0 is expressed as:

[0095]

[0096] For the generated base station trajectory sequence, traverse the adjacent base stations and If two adjacent and If there is no edge in the original base station relationship graph G, that is, there is no switching relationship, then the loss value is 1, otherwise the loss value is 0, and all the loss values ​​are accumulated and summed to obtain

[0097] The loss function of the final model is

[0098] In some embodiments of the invention, in the processing of the diffusion model:

[0099] First generate a noise that conforms to the standard normal distribution, given a diffusion step T, and generate the prediction noise ε according to the trained network DN T , and then generate according to the following formula To noise

[0100]

[0101] Calculating prediction noise The cosine similarity with the base station representation vector is also obtained by using argmax to obtain the generated base station trajectory sequence.

[0102] An embodiment of the present invention also provides a mobile signaling trajectory data generation system based on a diffusion process, the system comprising a computer device, the computer device comprising a processor and a memory, the memory storing computer instructions, the processor being used to execute the computer instructions stored in the memory, and when the computer instructions are executed by the processor, the system implements the steps implemented by the method described above.

[0103] The embodiment of the present invention also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps implemented by the mobile signaling trajectory data generation method based on the diffusion process are implemented. The computer-readable storage medium can be a tangible storage medium, such as a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a floppy disk, a hard disk, a removable storage disk, a CD-ROM, or any other form of storage medium known in the technical field.

[0104] It should be understood by those skilled in the art that the exemplary components, systems and methods described in conjunction with the embodiments disclosed herein can be implemented in hardware, software or a combination of the two. Whether it is performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention. When implemented in hardware, it can be, for example, an electronic circuit, an application specific integrated circuit (ASIC), appropriate firmware, a plug-in, a function card, etc. When implemented in software, the elements of the present invention are programs or code segments used to perform the required tasks. The program or code segment can be stored in a machine-readable medium, or transmitted on a transmission medium or a communication link via a data signal carried in a carrier.

[0105] It should be clear that the present invention is not limited to the specific configuration and processing described above and shown in the figures. For the sake of simplicity, a detailed description of the known method is omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present invention is not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order between the steps after understanding the spirit of the present invention.

[0106] In the present invention, features described and / or illustrated for one embodiment may be used in the same or similar manner in one or more other embodiments, and / or combined with features of other embodiments or replace features of other embodiments.

[0107] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the embodiments of the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for generating mobile signaling trajectory data based on a diffusion process, characterized in that: The steps of the method include: Acquire original mobile signaling data of a preset time period, the original mobile signaling data including mobile signaling data corresponding to each base station, and pre-process the original mobile signaling data; Determine a base station relationship graph based on mobile signaling data of each base station, wherein the base station relationship graph includes a switching relationship between base stations, and construct an initial data set for each base station based on the switching relationship of each base station and feature data of the base station; constructing a representation vector corresponding to each base station based on an initial data set of each base station, combining the representation vectors of each base station into an input vector and inputting it into a pre-trained diffusion model, wherein the diffusion model outputs a trajectory generation vector, and the trajectory generation vector is composed of a plurality of trajectory generation sub-vectors corresponding to the user; The trajectory data of the user in a preset time period is determined based on the trajectory generation vector.

2. The method for generating mobile signaling trajectory data based on diffusion process according to claim 1, characterized in that: In the step of preprocessing the original mobile signaling data: Dividing the original mobile signaling data of a preset time period into mobile signaling data of a plurality of sub-time periods; The base station to which the user is connected in each sub-time period is determined based on the mobile signaling data in the sub-time period.

3. The method for generating mobile signaling trajectory data based on diffusion process according to claim 2, characterized in that: In the step of determining the base station connected to the user in each sub-time period based on the mobile signaling data in the sub-time period, the base station with the longest connection time of each user in the sub-time period is counted as the base station connected to the user in the sub-time period.

4. The method for generating mobile signaling trajectory data based on diffusion process according to claim 1, characterized in that: In the step of preprocessing the original mobile signaling data, the preprocessing includes sorting according to time, removing ping-pong data, removing duplicate data and removing noise data.

5. The method for generating mobile signaling trajectory data based on diffusion process according to claim 1, characterized in that: In the step of determining the base station relationship graph based on the mobile signaling data of each base station, each base station is regarded as a node in the base station relationship graph, and if there is a switching relationship between two base stations, an edge between the two base stations is constructed.

6. The method for generating mobile signaling trajectory data based on diffusion process according to claim 1, characterized in that: The characteristic data of the base station includes information such as the location area identification code, cell identification code, and longitude and latitude of the base station. In the step of constructing a representation vector corresponding to each base station based on the initial data set of each base station: Combining the characteristic data of the base stations in the initial data set with the switching relationship of each base station and encoding them; Get the representation vector corresponding to each base station.

7. The method for generating mobile signaling trajectory data based on diffusion process according to claim 1, characterized in that: The value of each dimension in the trajectory generation vector corresponds to a base station. In the step of determining the trajectory data of the user in a preset time period based on the trajectory generation vector, the trajectory data of the user for the base station is determined based on the order of the dimensions in the trajectory generation vector.

8. The method for generating mobile signaling trajectory data based on a diffusion process according to any one of claims 1 to 7, characterized in that: The method further comprises pre-training the diffusion model, in which: Calculate the prediction noise loss, restoration loss and base station switching constraint loss based on the output trajectory generation vector and label data; A total loss function is calculated based on the predicted noise loss, the restoration loss and the base station switching constraint loss, and the diffusion model is pre-trained based on the total loss function.

9. The method for generating mobile signaling trajectory data based on diffusion process according to claim 8, characterized in that: In the step of calculating the prediction noise loss, restoration loss, and base station switching constraint loss based on the output trajectory generation vector and label data: Calculate the mean square error between the predicted noise and white noise as the prediction noise loss; Calculate the mean square error between the output trajectory generation vector and the input vector as the recovery loss; Traverse the adjacent base stations in the trajectory generation vector. If two adjacent base stations have an edge in the base station relationship graph, the loss value is 0; if two adjacent base stations do not have an edge in the base station relationship graph, the loss value is 1. The loss values ​​are accumulated to obtain the base station switching constraint loss.

10. A mobile signaling trajectory data generation system based on a diffusion process, characterized in that: The system includes a computer device, which includes a processor and a memory. The memory stores computer instructions. The processor is used to execute the computer instructions stored in the memory. When the computer instructions are executed by the processor, the system implements the steps implemented by the method as described in any one of claims 1 to 9.

Images