Parking lot site selection method and device based on space-time tensor enhancement, terminal equipment and storage medium

Abstract

This application relates to the field of urban transportation facility planning and optimization technology, and provides a parking lot site selection method, device, terminal equipment, and storage medium based on spatiotemporal tensor augmentation. The method includes: acquiring sparse traffic flow data and multi-source heterogeneous geographic data of a target area; wherein the target area contains multiple candidate addresses; performing data completion on the sparse traffic flow data to obtain completed target traffic flow data; constructing a feature set corresponding to each candidate address based on the target traffic flow data and multi-source heterogeneous geographic data; constructing a comprehensive objective function corresponding to each candidate address based on the feature set and a preset objective; solving each comprehensive objective function to obtain the solution result; and determining the target parking lot site from the multiple candidate addresses based on the multiple solution results. This method can output an optimal site selection scheme that balances service coverage, construction cost, and traffic convenience, meeting the needs of scientific site selection.

Description

Technical Field

[0001] This application belongs to the field of urban transportation facility planning and optimization technology, and in particular relates to a parking lot site selection method, device, terminal equipment and storage medium based on spatiotemporal tensor enhancement. Background Technology

[0002] With the acceleration of urbanization and the surge in motor vehicle ownership, scientific planning of parking lot site selection is crucial for optimizing spatial resource allocation and improving the operational efficiency of intelligent transportation systems. This decision requires comprehensive consideration of factors such as road network structure, surrounding land use, and dynamic traffic flow characteristics. Among these factors, traffic flow speed data is key to assessing regional traffic capacity and parking demand.

[0003] Existing technologies mostly rely on coarse-grained statistical indicators such as daily average traffic flow for single-objective parking lot site selection planning. However, in the actual traffic data collection process, uncontrollable factors such as sensor failure and extreme weather interference can lead to a large amount of continuous missing traffic flow and speed data, which damages the integrity of the data. Furthermore, traditional methods ignore the dynamic changes of traffic data in the time dimension and cannot effectively capture the weekday patterns (such as weekdays and weekends) and tidal phenomena (such as morning and evening rush hours) of urban traffic, making it difficult to meet the needs of scientific site selection. Summary of the Invention

[0004] This application provides a parking lot site selection method, apparatus, terminal equipment, and storage medium based on spatiotemporal tensor augmentation, which can output an optimal site selection scheme that takes into account service coverage, construction cost, and traffic convenience, thus meeting the needs of scientific site selection.

[0005] In a first aspect, embodiments of this application provide a parking lot site selection method based on spatiotemporal tensor augmentation, including: Acquire sparse traffic flow data and multi-source heterogeneous geographic data of the target area; the target area contains multiple candidate addresses; Complete the sparse traffic flow data to obtain the completed target traffic flow data; Construct a feature set for each candidate address based on target traffic flow data and multi-source heterogeneous geographic data; A comprehensive objective function is constructed for each candidate address based on the feature set and the preset objective; wherein, the preset objective represents a pre-defined optimization direction and evaluation criterion; Solve each of the aforementioned integrated objective functions to obtain the solution results for each of the aforementioned integrated objective functions; The target location address corresponding to the parking lot is determined from the candidate addresses based on the multiple solution results.

[0006] In this embodiment, sparse traffic flow data and multi-source heterogeneous geographic data of the target area are first acquired. The sparse traffic flow data is then supplemented to obtain complete and accurate target traffic flow data. A feature set is then constructed for each candidate address by combining the two types of data. A comprehensive objective function is established for each candidate address based on a preset optimization objective and solved. Finally, the target location of the parking lot is determined based on the solution results. The problem of sparse and incomplete traffic flow data is solved by data supplementation. At the same time, the characteristics of the candidate addresses are comprehensively characterized by the integration of multi-source heterogeneous geographic information. The objective selection is achieved through a quantified comprehensive objective function, which avoids the bias caused by subjective judgment and can significantly improve the accuracy, scientificity and practicality of parking lot location selection, making the location results more in line with actual traffic needs and geographical environment.

[0007] In one possible implementation of the first aspect, sparse traffic flow data is augmented to obtain augmented target traffic flow data, including: The sparse traffic flow data is spatiotemporally mapped to obtain three-dimensional tensor data; A first objective function is constructed based on the aforementioned three-dimensional tensor data and the preset core tensor and latent feature matrix; The first objective function is expressed as:

[0008] in, This indicates that the three-dimensional tensor data X is at position. Observations at; These are the elements corresponding to the latent feature matrix A obtained by decomposing the three-dimensional tensor data; These are the elements corresponding to the latent feature matrix B obtained by decomposing the three-dimensional tensor data; These are the elements corresponding to the latent feature matrix C obtained by decomposing the three-dimensional tensor data; The core tensor obtained by decomposing the three-dimensional tensor data. Element; For elements The rank of the corresponding dimension; For elements The rank of the corresponding dimension; For elements The rank of the corresponding dimension; The regularization coefficient is used. As an indicator variable, it is used to characterize the sparse structure of the three-dimensional tensor data X; Wherein, the first objective function is used to minimize the deviation between the target traffic flow data and the sparse traffic flow data; the core tensor is used to characterize the global distribution characteristics and cross-modal interaction of the sparse traffic flow data in the third-order tensor space; the latent feature matrix is ​​used to characterize the spatiotemporal local evolution law and inherent attribute characteristics of the sparse traffic flow data under different modes. The sparse traffic flow data is completed according to the first objective function to obtain the completed target traffic flow data.

[0009] In this embodiment, sparse traffic flow data is spatiotemporally mapped into three-dimensional tensor data. A first objective function for minimizing the completion deviation is constructed by combining a preset core tensor and a preset latent feature matrix, and data completion is completed accordingly. This not only accurately captures the spatiotemporal dimension features of traffic flow data in tensor form, but also ensures the accuracy and authenticity of the completed data by using the objective function for minimizing the deviation as a guide. This effectively solves the problem of missing or incomplete original sparse traffic flow data, and provides high-precision and high-reliability dynamic traffic data support for the subsequent construction of feature sets for parking lot site selection. This improves the scientificity and accuracy of parking lot site selection analysis from the data source.

[0010] In one possible implementation of the first aspect, the step of performing data completion on the sparse traffic flow data according to the first objective function to obtain the completed target traffic flow data includes: The target core tensor and the target hidden feature matrix are obtained by updating the preset core tensor and the hidden feature matrix according to the first objective function. The sparse traffic flow data is completed by using the target core tensor and the target latent feature matrix to obtain the target traffic flow data.

[0011] In this embodiment, the core tensor representing the global spatiotemporal pattern of traffic flow and the target core tensor and target hidden feature matrix representing the local patterns of each dimension are obtained by iteratively updating the preset tensor data and hidden feature matrix through the first objective function. The sparse traffic flow data is then completed by relying on two types of parameters. This not only accurately mines the spatiotemporal distribution pattern of traffic flow data, but also achieves accurate data completion guided by the pattern features. This effectively improves the authenticity and fit of the target traffic flow data after completion, solves the problem of sparsity and missing original data, and provides high-precision dynamic traffic data support for the feature analysis of subsequent parking lot site selection. This ensures the scientificity and accuracy of site selection decisions from the data source.

[0012] In one possible implementation of the first aspect, the step of updating the preset core tensor and the latent feature matrix according to the first objective function to obtain the target core tensor and the target latent feature matrix includes: Construct a system of linear equations based on the first objective function; Solve the system of linear equations to obtain the updated step size; The core tensor and the latent feature matrix in the first objective function are updated according to the update step size until the first objective function converges. When the first objective function converges, the currently updated core tensor is determined as the target core tensor, and the currently updated latent feature matrix is ​​determined as the target latent feature matrix.

[0013] In this embodiment, by constructing a system of linear equations, solving to update the step size, and iteratively updating the preset core tensor and latent feature matrix until the function converges, it not only ensures the complete capture of the three-dimensional global laws of traffic flow by the core tensor, but also achieves fine mining of local laws in each dimension. This provides optimal core tensor and latent feature matrix support that conforms to actual traffic laws for subsequent accurate completion of sparse traffic flow data, ensuring the accuracy and authenticity of the completed data from the parameter level, and solving the problem of sparsity and missing original data.

[0014] In one possible implementation of the first aspect, a feature set corresponding to each candidate address is constructed based on target traffic flow data and multi-source heterogeneous geographic data, including: For each candidate address, calculate the average traffic flow data of the candidate address within a preset time period based on the target traffic flow data; Calculate the point of interest density index of candidate addresses within a preset range based on multi-source heterogeneous geographic data; The candidate addresses are assigned a first weighting coefficient based on multi-source heterogeneous geographic data; The betweenness centrality of candidate addresses is determined based on multi-source heterogeneous geographic data; where betweenness centrality is used to characterize the degree of coreness of candidate addresses in the road network topology. The feature set of candidate addresses is obtained by combining average traffic flow data, point of interest density index, first weight coefficient, and betweenness centrality.

[0015] In this embodiment, the average traffic flow data for a preset time period under the target traffic flow data is calculated for each candidate address. The point of interest density index is calculated by combining multi-source heterogeneous geographic data, a first weight coefficient is assigned, and the betweenness centrality in the road network topology is determined. Then, the four types of indicators are combined to construct a feature set, which integrates dynamic traffic operation characteristics and static geospatial attributes. The core characteristics of the candidate address are characterized from multiple dimensions such as traffic demand, facility support, land use attributes, and road network hub function. This not only ensures the comprehensiveness and relevance of the feature set, but also realizes the quantification and standardization of the site selection indicators. It provides a structured, comparable, and high-quality data foundation for the subsequent multi-objective optimization analysis of parking lot site selection, effectively avoiding the problems of single and subjective traditional site selection indicators. From the feature construction level, it ensures the scientificity and accuracy of site selection decisions.

[0016] In one possible implementation of the first aspect, a comprehensive objective function corresponding to each candidate address is constructed based on a feature set and a preset target, including: For each candidate address, a second objective function is constructed based on at least one preset objective; the second objective function can be any of the following:

[0017] in, For the weight vector, For the first n Candidate site selection points Construction status variables; The candidate site selection point The scale of construction; The candidate site selection point Feature set;

[0018] in, For parking demand points, A collection of parking demand points Any parking demand point;

[0019] in, To fix startup costs, For construction costs; A comprehensive objective function is constructed based on multiple secondary objective functions and preset constraints; the comprehensive objective function is:

[0020] Among these constraints are land capacity limitations. Minimum allowed distance threshold between candidate address points Total project budget ;in, For the first 10 candidate address points.

[0021] In this embodiment, a second objective function is constructed for each candidate address in conjunction with a preset objective. All second objective functions are then integrated with preset constraints to construct a comprehensive objective function. Based on the comprehensive objective function of each candidate address, the target parking lot location is determined. By constructing multiple objective functions and integrating constraints, the multi-dimensional requirements for parking lot location selection are transformed into a quantitative mathematical optimization model. This achieves both accurate quantification of location requirements and consideration of multiple objectives, while also ensuring the practicality of the location selection results through constraints. This transforms location decision-making from subjective experience-based judgment to data-driven quantitative analysis, effectively improving the scientific, accurate, and rational nature of parking lot location selection.

[0022] In one possible implementation of the first aspect, the target location address corresponding to the parking lot is determined from multiple candidate addresses based on multiple solution results, including: For each solution result, the solution result is weighted and summed according to multiple preset second weight coefficients to obtain the comprehensive score of each candidate address; The highest score is selected from the comprehensive scores corresponding to each of the multiple candidate addresses; The candidate address corresponding to the highest score is determined as the target address.

[0023] In this embodiment, the solution is obtained by solving the comprehensive objective function of each candidate address. The results are then weighted and summed using a preset second weighting coefficient to obtain a comprehensive score. The highest score is selected and its corresponding candidate address is determined as the target parking lot location. The standardized process of quantitative solution, weighted scoring, and extreme value selection achieves precise implementation of the location decision. The weighted summation takes into account the differences in importance of the various dimensions of the location selection objectives, and the highest score selection intuitively locks in the optimal solution, making the parking lot location decision more objective and operable. It effectively avoids the bias of subjective judgment and ultimately determines the parking lot location with the best comprehensive performance under the multi-dimensional location selection objectives, ensuring the scientific nature and adaptability of the location selection results.

[0024] Secondly, embodiments of this application provide a parking lot location selection device based on spatiotemporal tensor augmentation, comprising: The data acquisition module is used to acquire sparse traffic flow data and multi-source heterogeneous geographic data of the target area; the target area contains multiple candidate addresses. The data completion module is used to complete sparse traffic flow data to obtain the completed target traffic flow data. The feature construction module is used to construct a feature set corresponding to each candidate address based on the target traffic flow data and multi-source heterogeneous geographic data. The target address determination module is used to determine the target address corresponding to the parking lot from multiple candidate addresses based on multiple feature sets.

[0025] Thirdly, embodiments of this application provide a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the parking lot location method based on spatiotemporal tensor augmentation as described in any of the first aspects above.

[0026] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the parking lot location method based on spatiotemporal tensor augmentation as described in any of the first aspects above.

[0027] Fifthly, embodiments of this application provide a computer program product that, when run on a terminal device, causes the terminal device to execute the parking lot location selection method based on spatiotemporal tensor augmentation as described in any of the first aspects above.

[0028] It is understood that the beneficial effects of the second to fifth aspects mentioned above can be found in the relevant descriptions in the first aspect mentioned above, and will not be repeated here. Attached Figure Description

[0029] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0030] Figure 1 This is a flowchart illustrating the parking lot site selection method based on spatiotemporal tensor augmentation provided in the embodiments of this application; Figure 2 This is a schematic diagram of the data completion process provided in the embodiments of this application. Figure 1 ; Figure 3 This is a schematic diagram of the data completion process provided in the embodiments of this application. Figure 2 ; Figure 4 This is a flowchart illustrating the process of determining the target core tensor and the target latent feature matrix provided in an embodiment of this application; Figure 5 This is a schematic diagram of the process for determining a candidate address feature set provided in an embodiment of this application; Figure 6 This is a schematic diagram of the process for constructing the target synthesis function provided in an embodiment of this application; Figure 7This is a flowchart illustrating the process of determining the target location address provided in an embodiment of this application; Figure 8 This is a general structural diagram of the parking lot location selection method based on spatiotemporal tensor augmentation provided in the embodiments of this application; Figure 9 This is a structural block diagram of the parking lot location selection device based on spatiotemporal tensor augmentation provided in the embodiments of this application; Figure 10 This is a schematic diagram of the structure of the terminal device provided in the embodiments of this application. Detailed Implementation

[0031] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0032] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0033] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0034] As used in this application specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if detected [the described condition or event]" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once detected [the described condition or event]," or "in response to detection [the described condition or event]."

[0035] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0036] References to "one embodiment" or "some embodiments" in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized.

[0037] With the acceleration of urbanization and the surge in motor vehicle ownership, scientific planning of parking lot site selection is crucial for optimizing spatial resource allocation and improving the operational efficiency of intelligent transportation systems. This decision requires comprehensive consideration of factors such as road network structure, surrounding land use, and dynamic traffic flow characteristics. Among these factors, traffic flow speed data is key to assessing regional traffic capacity and parking demand.

[0038] Existing technologies mostly rely on coarse-grained statistical indicators such as daily average traffic flow for single-objective parking lot site selection planning. However, in the actual traffic data collection process, uncontrollable factors such as sensor failure and extreme weather interference can lead to a large amount of continuous missing traffic flow and speed data, which damages the integrity of the data. Furthermore, traditional methods ignore the dynamic changes of traffic data in the time dimension and cannot effectively capture the weekday patterns (such as weekdays and weekends) and tidal phenomena (such as morning and evening rush hours) of urban traffic, making it difficult to meet the needs of scientific site selection.

[0039] To address the aforementioned technical issues, this application provides a parking lot site selection method based on spatiotemporal tensor augmentation. This method involves data preprocessing (mapping a 3D spatiotemporal tensor and labeling missing data), sparse data completion based on Tucker decomposition and Hessian-free optimization, multi-source data fusion, multi-objective optimization modeling (maximizing service coverage, minimizing walking distance and overall cost, including multiple constraints), and solving for the Pareto optimal solution set using a non-dominated sorting genetic algorithm with an elitist strategy. The final output is the optimal site selection scheme, which can be applied to multiple fields such as urban public parking lot planning and new energy vehicle charging station site selection.

[0040] See Figure 1 This is a flowchart illustrating the parking lot site selection method based on spatiotemporal tensor augmentation provided in an embodiment of this application. It is intended as an example and not a limitation. The method may include the following steps: S101, acquire sparse traffic flow data and multi-source heterogeneous geographic data of the target area; the target area contains multiple candidate addresses.

[0041] In this application embodiment, the basic data preparation work before parking lot site selection needs to obtain two types of key data of the target area (including multiple parking lot candidate addresses): one is sparse traffic flow data reflecting the regional traffic dynamics, and the other is multi-source heterogeneous geographic data reflecting the regional geographic and functional attributes, and it is clear that the target area contains multiple parking lot candidate addresses.

[0042] For example, historical traffic flow and speed data for different road sections, time periods, and days in the target area can be collected through a traffic monitoring system (some data may be missing due to sensor malfunctions, extreme weather, etc., i.e., sparse traffic flow data); basic data can be obtained through channels such as Geographic Information Systems (GIS), government databases, and Internet map application programming interfaces (APIs), such as through local traffic management departments and Internet APIs. Specifically, this includes: (1) Obtain the road network topology G=(V, E) of the target area, including r (1) Main road sections; (2) Collect data on the corresponding area for a continuous period of time. t The average speed data of traffic flow is set to 15 minutes, that is, the whole day is divided into 96 time slices; (3) Obtain land use data (commercial, residential and industrial), point of interest (POI) data (coordinates of shopping malls, hospitals, schools, etc.), and multiple candidate site selection point sets, etc., multi-source heterogeneous geographic data; at the same time, identify multiple parking lot candidate addresses preset in the target area and complete the preliminary collection of the two types of data and candidate address information.

[0043] S102, perform data completion on the sparse traffic flow data to obtain the completed target traffic flow data.

[0044] In this embodiment of the application, due to factors such as sensor failure and extreme weather, some values ​​of the collected traffic flow data are missing. Therefore, it is necessary to fill in the missing parts of the sparse traffic flow data (due to sensor failure, environmental interference, etc.) through specific technical methods, so as to finally obtain complete and usable target traffic flow data.

[0045] In one embodiment, see Figure 2 This is a schematic diagram of the data completion process provided in the embodiments of this application. Figure 1 ,like Figure 2 As shown, step S102 includes: S201 involves spatiotemporal mapping of sparse traffic flow data to obtain three-dimensional tensor data.

[0046] In this embodiment of the application, the original traffic flow data is spatiotemporally mapped to define a three-dimensional traffic flow tensor. ,in The number of road segments Represents the number of time slices within a single day. Represents the number of days observed. Each element in the tensor Indicates the first Heavenly The time slice in the first The average vehicle speed on each road segment. During the mapping process, the following settings are configured. As an indicator variable, it is used to characterize the sparse structure of X. If data is missing due to sensor failure or environmental interference, then... ,on the contrary .

[0047] For example, firstly, clarify the basic information related to traffic flow in the target area, and determine the three dimensions of data mapping: take the number of road segments as the first dimension ( The number of time slices divided within a single day is used as the second dimension. (e.g., 15 minutes is a time slice, with a total of 96 slices per day), the number of days observed is used as the third dimension ( Subsequently, the average vehicle speed information of each road segment in the sparse traffic flow data at different dates and time slices is extracted, and each "first" is... Heavenly The first time slice The average vehicle speed of each road segment corresponds to an element in a three-dimensional tensor. In the process, the data observation status is marked (valid data is retained, and missing data caused by sensor failure or environmental interference is marked separately), ultimately forming structured three-dimensional tensor data that fully carries the spatiotemporal distribution characteristics of traffic flow.

[0048] S202, a first objective function is constructed based on three-dimensional tensor data and multiple preset core tensors and latent feature matrices; the first objective function is expressed as:

[0049] in, Represents the three-dimensional tensor data In position Observations at; These are the elements corresponding to the latent feature matrix A obtained by decomposing the three-dimensional tensor data; These are the elements corresponding to the latent feature matrix B obtained by decomposing the three-dimensional tensor data; These are the elements corresponding to the latent feature matrix C obtained by decomposing the three-dimensional tensor data; The core tensor obtained by decomposing the three-dimensional tensor data. of Yuan; For elements The rank of the corresponding dimension; For elements The rank of the corresponding dimension; For elements The rank of the corresponding dimension; This is the regularization coefficient. As an indicator variable, it is used to characterize the sparse structure of the three-dimensional tensor data X.

[0050] The first objective function is used to minimize the deviation between the target traffic flow data and the sparse traffic flow data; the core tensor is used to characterize the global distribution characteristics and cross-modal interaction of the sparse traffic flow data in the third-order tensor space; and the latent feature matrix is ​​used to characterize the spatiotemporal local evolution law and inherent attribute characteristics of the sparse traffic flow data under different modes.

[0051] In this embodiment, a first objective function is constructed using three-dimensional tensor data (including traffic flow spatiotemporal information) and a preset core tensor and latent feature matrix. The core objective is to minimize the deviation between the completed target traffic flow data and the original sparse traffic flow data, providing direction for subsequent optimization solutions for data completion. The core tensor is used to describe the global distribution of sparse traffic flow data in the three-dimensional tensor space and the correlation between different modes, while the latent feature matrix is ​​used to characterize the spatiotemporal local variation patterns and inherent characteristics of the data under different modes.

[0052] For example, for the sparse tensor X containing missing values ​​constructed above, a second-order latent feature analysis model based on Tucker decomposition is established. Tucker decomposition can decompose the original data tensor into a core tensor. and three hidden feature matrices , and Therefore, reconstruct the tensor It is expressed as follows: (1.1) in, Indicates the first Modal tensor product operation.

[0053] Considering the presence of missing data in the original tensor X, a learning objective for the tensor latent feature model based on Tucker decomposition is constructed according to the density-guided principle. Due to the uneven distribution of data in the original tensor, Tikhonov regularization is used to constrain the optimization process and prevent overfitting. Therefore, the first objective function is defined as follows: (1.2) in, Represents the original tensor In position Observations at; The elements corresponding to the latent feature matrices A, B, and C in the three dimensions of the Tucker decomposition; For core tensor Element; These represent the ranks of the corresponding dimensions; The regularization coefficient is used. As an indicator variable, used to characterize The sparse structure.

[0054] S203, perform data completion on the sparse traffic flow data according to the first objective function to obtain the completed target traffic flow data.

[0055] In the embodiments of this application, the first objective function is to "minimize the deviation between the completed data and the original sparse data". Through the relevant optimization solution process, the missing parts in the sparse traffic flow data are filled in, and finally the complete target traffic flow data is obtained.

[0056] The above method maps sparse traffic flow data spatiotemporally into three-dimensional tensor data. It then constructs a first objective function to minimize the completion bias by combining a preset core tensor and a latent feature matrix, thereby completing the data completion. This approach not only accurately captures the spatiotemporal dimensional features of traffic flow data in tensor form but also ensures the accuracy and authenticity of the completed data by using the objective function to minimize the bias. This effectively solves the problem of missing or incomplete original sparse traffic flow data and provides high-precision, high-reliability dynamic traffic data support for the subsequent construction of feature sets for parking lot site selection. It improves the scientific rigor and accuracy of parking lot site selection analysis from the data source.

[0057] In one embodiment, see Figure 3 This is a schematic diagram of the data completion process provided in the embodiments of this application. Figure 2 ,like Figure 3 As shown, step S203 includes: S301, update the preset core tensor and latent feature matrix according to the first objective function to obtain the target core tensor and target latent feature matrix.

[0058] In this embodiment, the first objective function that minimizes the deviation between the completed data and the original sparse data optimizes and iteratively updates the preset core tensor and latent feature matrix, ultimately obtaining the target core tensor representing the global laws of traffic flow and the target latent feature matrix representing the local laws of each dimension, providing key parameter support for data completion.

[0059] In one embodiment, see Figure 4This is a flowchart illustrating the process of determining the target core tensor and the target latent feature matrix provided in an embodiment of this application, as shown below. Figure 4 As shown, step S301 includes: S401, Construct a system of linear equations based on the first objective function.

[0060] In this embodiment, to address the issues of low efficiency and slow convergence when directly solving the objective function, Hessian-Free Optimization (HF) is introduced to iteratively update the objective function. This method avoids explicitly calculating the Hessian matrix, thereby achieving robust inference and high-precision reconstruction of traffic data at missing locations. The method first employs a Newtonian-style second-order optimization approach, transforming the optimization problem of the first objective function into solving a system of linear equations constructed from the gradient and the Hessian matrix. The unknowns in this system are hypervectors composed of the latent feature matrix and the core tensor. Considering the non-convex nature of traffic flow data, the original Hessian matrix may be non-positive definite, leading to difficulty in convergence. Therefore, a Gaussian-Newton matrix is ​​used to replace the Hessian matrix, ultimately forming a system of linear equations with stronger numerical stability.

[0061] Specifically, to address the slow convergence speed of the first-order gradient descent method, a second-order optimization method using a Newton-type strategy is employed. This method calculates the increment by solving a system of linear equations constructed from the gradient and the Hessian matrix. The system of linear equations is expressed as: (1.3) in, Representing the latent feature matrix , , and core tensor The supervector.

[0062] Considering the non-convex nature of traffic flow data, the original Hessian matrix H may contain negative eigenvalues, leading to non-positive definiteness and preventing optimization from converging. Using a Gaussian-Newton matrix... Replacement Hessian matrix This effectively solves the saddle point problem in nonconvex optimization and enhances the numerical stability of the solution process. Equation (1.3) is transformed into: (1.4) S402, solve the system of linear equations to obtain the updated step size.

[0063] In this embodiment, for the linear equation system that has been transformed and constructed from the Gaussian-Newton matrix (replacing the original non-positive definite Hessian matrix) and the gradient of the objective function, the conjugate gradient method is used for efficient numerical solution. During the solution process, relying on the numerical characteristics of the linear equation system, the complex direct matrix inversion operation is avoided, reducing the computational complexity to the linear level. Finally, the update step size of the target core tensor and the target latent feature matrix is ​​accurately obtained. This step size will be used for the next round of iterative adjustment of the preset core tensor and latent feature matrix, driving the objective function to converge in the direction of "minimizing the deviation between the completed data and the original sparse data".

[0064] Specifically, the conjugate gradient (CG) method is used to approximate the solution of the above linear equations. In each iteration of CG, only the product of the Gaussian-Newton matrix and any vector v needs to be calculated. This reduces the computational complexity from Reduce to the linear level.

[0065] (1.5) in Indicates in Let U be the Jacobian matrix on the surface, and U be the identity matrix. Then... It can be calculated based on the directional derivative of vector v: (1.6) Approximate optimal step size output by the CG algorithm Update model parameters: (1.7) S403, update the preset core tensor and latent feature matrix in the first objective function according to the update step size until the first objective function converges.

[0066] In this embodiment, the preset core tensor and latent feature matrix in the first objective function are continuously iteratively adjusted using the update step size obtained by solving the linear equation system until the objective function reaches a convergent state, and finally the optimal core tensor and latent feature matrix are obtained, providing support for traffic flow data completion.

[0067] For example, the update step size obtained from solving the linear equation system is first applied to the preset core tensor and latent feature matrix in the first objective function to iteratively update various parameters; after each round of updates, the current value of the objective function is calculated (measuring the combined result of the deviation between the completed data and the original sparse data and the regularization constraint term), and it is determined whether the convergence condition is met (e.g., the convergence threshold is less than 10). -5If convergence is not achieved, repeat the process of "calculating gradient based on current parameters → constructing a system of linear equations → solving and updating step size → updating parameters" to continuously iterate and optimize until the objective function value is stable and no longer changes significantly (meeting the convergence criterion), then stop iterating.

[0068] S404, when the first objective function converges, the currently updated core tensor is determined as the target core tensor, and the currently updated latent feature matrix is ​​determined as the target latent feature matrix.

[0069] In this embodiment of the application, when the first objective function satisfies the convergence condition (e.g., the convergence threshold is less than 10), -5 When the combined result of the deviation between the completed data and the original sparse data and the regularization constraint term tends to stabilize, the iterative update of the core tensor and the latent feature matrix is ​​stopped; at this time, the core tensor obtained by the current update iteration is determined as the target core tensor, and the latent feature matrix obtained by the current update iteration is determined as the target latent feature matrix.

[0070] Among them, the target core tensor carries the overall correlation pattern of sparse traffic flow data in the three-dimensional tensor space of road segment, time slice, and number of days, i.e., three-dimensional spatiotemporal space; the target latent feature matrix corresponds to the independent local regularity features of the road segment dimension, time slice dimension, and number of days dimension, respectively. The two types of parameters together constitute the complete optimal parameter set required for data completion, laying the foundation for subsequent reconstruction of complete traffic flow data.

[0071] In the above method, by constructing a system of linear equations from the first objective function, solving and updating the step size, and iteratively updating the preset core tensor and latent feature matrix until the function converges, it not only ensures that the core tensor fully captures the three-dimensional global laws of traffic flow, but also achieves fine mining of local laws in each dimension. This provides optimal core tensor and latent feature matrix support that fits the actual traffic laws for subsequent accurate completion of sparse traffic flow data, ensuring the accuracy and authenticity of the completed data from the parameter level, and solving the problem of sparsity and missing original data.

[0072] S302, based on the target core tensor and the target latent feature matrix, the sparse traffic flow data is completed to obtain the target traffic flow data.

[0073] In this embodiment, the target core tensor and target latent feature matrix obtained after the convergence of the first objective function are used to fill in the missing parts of the sparse traffic flow data through tensor reconstruction, and finally obtain complete and high-precision target traffic flow data.

[0074] For example, after obtaining the optimal target core tensor and target latent feature matrix (corresponding to road segment, time slice, and day dimension), through the first... nModal tensor product operation fuses the target core tensor with the target latent feature matrix to reconstruct a complete three-dimensional traffic flow tensor. The reconstruction formula is shown in Formula 1.1 above.

[0075] This reconstruction process fully preserves the periodic characteristics of traffic flow data in the time domain (such as morning and evening peak hours, weekday / weekend differences) and the correlation characteristics in the spatial domain (such as traffic flow correlation between different road segments). It can accurately infer and fill in missing entries in the original sparse data caused by sensor failure and environmental interference, and finally output complete target traffic flow data that conforms to actual traffic patterns, providing high-quality data support for subsequent multi-source data integration and multi-target parking lot site selection.

[0076] In the above method, the preset core tensor and latent feature matrix are iteratively updated through the first objective function to obtain the target core tensor representing the global spatiotemporal pattern of traffic flow and the target latent feature matrix representing the local pattern of each dimension. Based on two types of parameters, sparse traffic flow data is completed. This method not only accurately mines the spatiotemporal distribution pattern of traffic flow data through the global and local partitioning of the preset core tensor and latent feature matrix, but also achieves accurate data completion guided by the pattern features. This effectively improves the authenticity and fit of the target traffic flow data after completion, solves the problem of sparsity and missing original data, and provides high-precision dynamic traffic data support for the feature analysis of subsequent parking lot site selection. It ensures the scientificity and accuracy of site selection decisions from the data source.

[0077] S103, construct the feature set corresponding to each candidate address based on the target traffic flow data and multi-source heterogeneous geographic data.

[0078] In this embodiment of the application, the target traffic flow data after being fused and supplemented is combined with multi-source heterogeneous geographic data to construct a comprehensive feature set for each parking lot candidate address in the target area, which includes dynamic traffic characteristics, static geographic attributes and cost constraints, providing a data foundation for subsequent multi-objective site selection optimization.

[0079] In one embodiment, see Figure 5 This is a flowchart illustrating the process of determining a candidate address feature set provided in an embodiment of this application, such as... Figure 5 As shown, step S103 includes: S501, for each candidate address, calculate the average traffic flow data of the candidate address within a preset time period based on the target traffic flow data.

[0080] In this embodiment of the application, for each parking lot candidate address Using the completed target traffic flow data, calculate its average traffic flow data within a preset time period (such as morning and evening rush hours), denoted as... This study extracts dynamic access characteristics from candidate addresses, providing key dynamic indicators for subsequent feature set construction and multi-objective site selection optimization. For example, firstly, the preset time period is defined as a representative period of traffic flow (such as the morning and evening rush hours from 7:00 to 9:00 and 17:00 to 19:00); then, based on the completed target traffic flow data (in three-dimensional tensor form, including vehicle speed information in terms of road segments, time slices, and number of days), the set of road segments corresponding to each candidate address is located; traffic flow data (average vehicle speed) for all time slices (e.g., 15 minutes / slice) and different observation days of these road segments within the preset time period is extracted, and the average traffic flow data of the candidate address within the preset time period is calculated by statistical averaging. This is used to quantify its dynamic passage efficiency, serving as one of the core dynamic indicators of the candidate address feature set.

[0081] S502, calculates the point of interest density index of candidate addresses within a preset range based on multi-source heterogeneous geographic data.

[0082] In this embodiment of the application, based on multi-source heterogeneous geographic data, the point of interest (POI) density index of each candidate address within a preset spatial range is calculated and denoted as follows: This quantifies the degree of clustering of functional facilities around candidate addresses, providing key static indicators for subsequent construction of candidate address feature sets and assessment of parking demand potential.

[0083] For example, firstly, the preset range is defined as a fixed spatial search radius centered on the candidate site (e.g., 500 meters in the document embodiment); then, Points of Interest (POI) data (including coordinates and category information of locations with specific functions such as shopping malls, hospitals, and schools) within this preset range are extracted from multi-source heterogeneous geographic data; the total number of POIs in the spatial neighborhood is counted, and the POI distribution density per unit area is calculated to obtain the POI density index corresponding to the candidate site. This index can objectively reflect the density of surrounding functional facilities and indirectly reflect the potential parking demand in the area. As an important component of the comprehensive feature set of candidate sites, its POI density index... The calculation formula is: (1.8) in, Located at candidate point Search radius R The set of POIs within the spatial neighborhood. Represents a set The total number of elements in the neighborhood is the total number of POIs in the neighborhood.

[0084] S503 is the first weighting coefficient assigned to candidate addresses based on multi-source heterogeneous geographic data.

[0085] In this embodiment of the application, based on the land use nature of candidate addresses in multi-source heterogeneous geographic data, a corresponding first weighting coefficient is assigned to each candidate address. (For example, the weight of commercial areas is set to 1.2 and that of residential areas is set to 1.0), quantifying the impact of different land use types on parking demand, and providing static weight indicators for the subsequent construction of candidate address feature sets.

[0086] For example, firstly, land use information (such as commercial land, residential land, industrial land, etc.) for each candidate address is extracted from multi-source heterogeneous geographic data; then, weighting rules are set according to the correlation between land use and parking demand (e.g., commercial centers are given a weight of 1.2 due to high parking demand, residential areas are given a basic weight of 1.0, and industrial areas are given a weight of 0.8 due to relatively low demand); finally, each candidate address is matched and assigned a corresponding first weight coefficient according to the rules, thereby distinguishing the priority of the impact of different land use types on parking lot site selection, which serves as an important static parameter of the comprehensive feature set of candidate addresses.

[0087] S504, determine the betweenness centrality of candidate addresses based on multi-source heterogeneous geographic data; where betweenness centrality is used to characterize the degree of coreness of candidate addresses in the road network topology.

[0088] In this embodiment of the application, the betweenness centrality of each candidate address is calculated based on the road network topology information in multi-source heterogeneous geographic data, thereby quantifying the core hub degree of the candidate address in the road network structure and providing key spatial indicators for evaluating traffic convenience and constructing a candidate address feature set.

[0089] For example, firstly, road network topology data G=(V,E) (where V is the set of nodes and E is the set of road segments) of the target area is extracted from multi-source heterogeneous geographic data to clarify the connection relationships and spatial distribution of each road segment; then, taking the road network node corresponding to the candidate address as the core, the betweenness centrality of the node is determined by calculating the frequency (or proportion) of the node on all shortest paths in the road network. A higher betweenness centrality value indicates that the node where the candidate address is located is a necessary stop on more shortest paths in the regional road network, indicating stronger transportation hub attributes and better accessibility. This indicator is ultimately incorporated into the comprehensive feature set of candidate addresses, providing data support for subsequent multi-objective site selection optimization (while also considering transportation convenience). Among these, betweenness centrality... The calculation formula is: (1.9) in, for s arrive t The total number of shortest paths, for s arrive t Passing through the node The number of shortest paths, s , t This represents any two distinct nodes in the road network.

[0090] S505 combines average traffic flow data, point of interest density index, first weight coefficient, and betweenness centrality to obtain a feature set of candidate addresses.

[0091] In this embodiment, the four core indicators (average traffic flow data, point of interest density index, first weighting coefficient, and betweenness centrality) used to calculate / determine candidate addresses are integrated to form a comprehensive feature set specific to the candidate address. Meanwhile, the estimated construction cost and land capacity restrictions This results in a multi-source input dataset that encompasses dynamic traffic demand, static infrastructure support, and land use attributes, providing a data foundation for subsequent multi-objective optimization.

[0092] In the above method, the average traffic flow data for a preset time period under the target traffic flow data is calculated for each candidate address. The point of interest density index is calculated by combining multi-source heterogeneous geographic data, the first weight coefficient is assigned, and the betweenness centrality in the road network topology is determined. Then, the four types of indicators are combined to construct a feature set, which integrates dynamic traffic operation characteristics and static geospatial attributes. It describes the core characteristics of the candidate address from multiple dimensions such as traffic demand, facility support, land use attributes, and road network hub function. This not only ensures the comprehensiveness and relevance of the feature set, but also realizes the quantification and standardization of site selection indicators. It provides a structured, comparable, and high-quality data foundation for subsequent multi-objective optimization analysis of parking lot site selection, effectively avoiding the problems of single and subjective traditional site selection indicators. From the feature construction level, it ensures the scientificity and accuracy of site selection decision-making.

[0093] S104, construct a comprehensive objective function corresponding to each candidate address based on the feature set and the preset objective; wherein, the preset objective represents a pre-defined optimization direction and evaluation criterion.

[0094] In this embodiment, the characteristic information of each candidate address is used, combined with the pre-set optimization direction and evaluation criteria, to establish a quantifiable comprehensive objective function for each candidate address, which is used to objectively evaluate its site selection merits.

[0095] In one embodiment, see Figure 6 This is a flowchart illustrating the construction of the target synthesis function provided in an embodiment of this application, such as... Figure 6 As shown, step S104 includes: S601, for each candidate address, construct a second objective function based on at least one preset objective, wherein the second objective function is any one of the following:

[0096] in, For the weight vector, For the first n Candidate site selection points Construction status variables; The candidate site selection point The scale of construction; The candidate site selection point Feature set;

[0097] in, For parking demand points, The set of parking demand points Any parking demand point; Indicates the first n Construction status variables of candidate site selection points Indicates construction, It indicates construction. Indicate demand points m with candidate site selection n The actual walking distance between them.

[0098]

[0099] in, To fix startup costs, For construction costs; In this embodiment of the application, for each candidate parking lot address, at least one preset objective is set in combination with the core requirements of parking lot site selection. Based on this, a dedicated second objective function is constructed to provide a mathematical model for quantitative analysis of subsequent quantitative evaluation of candidate addresses and selection of the optimal site.

[0100] For example, first, define For the nth candidate site The construction status variable (1 indicates construction, 0 indicates no construction); This refers to the construction scale (number of berths) of the site, of which It is a non-negative integer.

[0101] To effectively improve the scientific nature of parking lot site selection methods while reasonably controlling costs, this problem will construct three secondary objective functions, specifically: (a) Maximize the utility of dynamic service coverage F 1: (1.10) in, This is the weight vector for each attribute. Used to characterize the diminishing marginal returns of construction scale on service efficiency. For the nth candidate site Construction status variables; Candidate site selection points The scale of construction. For site selection The comprehensive feature set.

[0102] (b) Minimize the average walking distance of users F 2: (1.11) in, For parking demand points, The set of parking demand points Any parking demand point; Indicates the first n Construction status variables of candidate site selection points Indicates construction, It indicates construction. Indicate demand points m with candidate site selection n The actual walking distance between them.

[0103] (c) Minimize the total lifecycle cost F 3: (1.12) in, To fix startup costs, For construction costs S602, construct a comprehensive objective function based on multiple second objective functions and preset constraints; the comprehensive objective function is:

[0104] The constraints include land capacity limitations. Minimum allowed distance threshold between candidate address points Total project budget and 0-1 variable constraints for the construction status; among which, For the first 10 candidate address points.

[0105] In this embodiment, a comprehensive objective function is constructed by integrating the second objective function of all candidate addresses and combining it with the actual preset constraints of parking lot site selection. This enables a comprehensive quantitative evaluation of all candidate addresses and provides a unified mathematical optimization model for subsequent selection of the optimal parking lot site.

[0106] For example, the constraints are: Construction scale The land capacity limit calculated in the above steps shall not be exceeded. ,Right now: (1.13) To avoid excessive concentration of parking resources, any two selected candidate points and The distance between them must exceed the minimum allowed distance threshold. : (1.14) The total construction cost shall not exceed the total project budget. .

[0107] (1.15) In summary, considering the multi-objective function and constraints, the final optimization model (i.e., the comprehensive objective function) is as follows: (1.16) The above method transforms the multi-dimensional requirements of parking lot site selection into a quantitative mathematical optimization model by constructing multi-objective functions and integrating constraints. This not only achieves accurate quantification of site selection requirements and consideration of multiple objectives, but also ensures the practicality of site selection results through constraints. This transforms site selection decisions from subjective experience-based judgments to data-driven quantitative analysis, effectively improving the scientific, accurate, and rational nature of parking lot site selection.

[0108] S105, solve for each comprehensive objective function to obtain the solution result for each comprehensive objective function.

[0109] In this embodiment of the application, the comprehensive objective function constructed for each candidate address is numerically solved to obtain the function solution result corresponding to each candidate address. This quantifies the degree to which each candidate address meets the comprehensive requirements for parking lot site selection, providing a comparable quantitative basis for subsequent screening of target site locations.

[0110] Specifically, the comprehensive objective function corresponding to each candidate address is input into a multi-objective evolutionary solver, and a global search is performed using a non-dominated sorting genetic algorithm with an elitist strategy. Through iterative evolution, a set of non-dominated Pareto optimal solutions is generated, each solution representing a site selection combination that balances the objectives of all parties. Finally, a set of site selection schemes that balance social and economic benefits is output, and based on the final decision weights, the optimal parking space layout scheme is recommended from the set. Specifically: First, population initialization and hybrid encoding are performed, with the population size set to M and the maximum number of iterations to be [value missing]. Chromosomes are constructed using a hybrid real-number and binary encoding scheme, with each individual... It consists of two parts: one is a binary vector of length N. Corresponding location status Second, an integer vector of length N. Corresponding construction scale During initialization, M numbers of elements that satisfy the capacity constraints in step S4 are randomly generated. The feasible solutions are identified and subjected to constraint checks. Individuals that do not satisfy constraints (1.14) or (1.15) are naturally eliminated during evolution using a penalty function method, thereby generating the initial population. .

[0111] For the t-th generation population Calculate the fitness value vector of each individual on the multi-objective function defined in step S4. Based on Pareto dominance, a rapid non-dominated ordination of the population is performed, defining individuals... u Dominant Individual v If and only if and The population is divided into several non-overlapping Pareto ranks based on dominance relationships. ,in The layer represents the non-dominated solution set of the current population. To maintain the diversity of solution distribution in the target space, for the same... Rank Calculate the crowding distance for individuals within the layer:

[0112] in These are the maximum and minimum values ​​of the m-th objective function, respectively.

[0113] Genetic operators are used to generate offspring populations. Using a binary tournament selection method, based on non-dominant levels ( Parent individuals are selected based on the smallest possible value (ideally, the smaller the value of CD) and the crowding distance (ideally, the larger the value of CD, the better). Then, a simulated binary crossover operator is used to crossover the Q vector, and a single-point crossover operator is used to crossover the Z vector. Polynomial mutation and bit-flip mutation are used to process integer and binary variables respectively, generating offspring of size M. The father's generation with offspring Merged into a mixed population of size 2M ,right Perform non-dominated sorting and crowding calculation again, according to The top M individuals are selected sequentially from low to high, and from largest to smallest CD within the same stratum, to form the next generation population. This is to preserve the elite strategy.

[0114] Repeat the evaluation, sorting, selection, crossover, and mutation process described above until the number of iterations reaches [number missing]. Output the final population. In Layer individuals as Pareto optimal solution set (i.e., the solution result).

[0115] S106, determine the target location address corresponding to the parking lot from multiple candidate addresses based on multiple solution results.

[0116] In this embodiment of the application, the solution results (scores / evaluations) calculated for each candidate address are compared, and the optimal address is selected as the final target location address for the parking lot.

[0117] In one embodiment, see Figure 7 This is a flowchart illustrating the process of determining the target location address provided in an embodiment of this application, such as... Figure 7 As shown, step S603 includes: S701: For each solution result, the solution result is weighted and summed according to multiple preset second weight coefficients to obtain the comprehensive score of each candidate address.

[0118] In this embodiment, a preset second weighting coefficient is matched to the comprehensive objective function solution results of each candidate address, and the comprehensive score of each candidate address is calculated by weighted summation. The multi-dimensional solution results are transformed into a single quantitative score that can be directly compared horizontally, providing an intuitive basis for site selection decision.

[0119] For example, based on the decision-maker's preset preference weights for service quality and cost control (i.e., the second weight coefficient). The weighting coefficients are set based on actual data and can be adaptively adjusted according to data changes. Then, the solution results of the comprehensive objective function for each candidate address are extracted, and each solution result is multiplied by its corresponding second weighting coefficient to obtain the weighted score for each dimension. Finally, the weighted scores of all dimensions of the candidate address are summed, and the sum is the comprehensive score of the candidate address. All candidate addresses are calculated according to this rule to form a unified scoring result, enabling a direct comparison of the candidate addresses.

[0120] S702 selects the highest score from the comprehensive scores corresponding to multiple candidate addresses.

[0121] In this embodiment of the application, the comprehensive scores of all parking lot candidate addresses are compared numerically, and the candidate address with the highest score is selected. The candidate address corresponding to the highest score is locked as the optimal candidate for parking lot site selection, providing a direct quantitative basis for determining the target site address.

[0122] For example, a comprehensive score obtained by weighted summation of all candidate addresses is collected to form a standardized score list containing candidate address identifiers and corresponding comprehensive scores, ensuring that all scores are on the same scale and can be directly compared horizontally. Subsequently, the comprehensive scores in the list are sorted in descending order or extreme value search is performed to extract the maximum value in the list, which is the highest score among all candidate addresses. At the same time, the unique identifier of the candidate address corresponding to the highest score is recorded, completing the screening of the highest score and providing core quantitative reference for the subsequent final site selection decision.

[0123] S703 determines the candidate address corresponding to the highest score as the target address.

[0124] In this embodiment of the application, the parking lot candidate address corresponding to the highest score in the comprehensive score is directly determined as the target location address of the parking lot, thus completing the entire data-driven parking lot location decision-making process.

[0125] For example, the highest comprehensive score obtained after screening is retrieved, and the unique identifier of the candidate address corresponding to that score is matched. It is confirmed that all the characteristic indicators, function solution results and comprehensive score of the address meet the preset goals and constraints of parking lot site selection. Subsequently, the candidate address is officially determined as the target site for parking lot selection, completing the entire process from sparse traffic flow data completion, multi-source geographic data fusion, feature set construction, multi-objective function optimization, and finally implementing the site selection decision, ensuring that the target site is the optimal solution in terms of parking demand matching, traffic convenience and road network coreness.

[0126] In the above method, the solution is obtained by solving the comprehensive objective function of each candidate address. The results are then weighted and summed using a preset second weighting coefficient to obtain a comprehensive score. The highest score is selected and its corresponding candidate address is determined as the target parking lot location. The standardized process of quantitative solution, weighted scoring, and extreme value selection achieves precise implementation of the location decision. The weighted summation takes into account the differences in importance of the various dimensions of the location selection objectives, and the selection of the highest score directly locks in the optimal solution, making the parking lot location decision more objective and operable. It effectively avoids the bias of subjective judgment and ultimately determines the parking lot location with the best comprehensive performance under the multi-dimensional location selection objectives, ensuring the scientific nature and adaptability of the location selection results.

[0127] See Figure 8 This is a general structural diagram of the parking lot location selection method based on spatiotemporal tensor augmentation provided in the embodiments of this application, as shown below. Figure 8 As shown, it specifically includes: S1 (Data Preprocessing): Obtain road network topology data, traffic flow speed, land use, POI distribution, and coordinates of candidate sites; based on geographical location, sampling time period, and date attribute, map the original sparse traffic flow data into a three-dimensional spatiotemporal tensor model, and mark missing data entries caused by equipment failure or environmental interference.

[0128] S2 (Data Completion): For the sparse tensor with missing values ​​constructed in step S1, a spatiotemporal traffic flow data completion model based on Tucker decomposition and tensor second-order latent feature analysis is established. Based on Tucker decomposition, the original data tensor is decomposed into a core tensor and a latent feature matrix of the corresponding dimension, which fully preserves the periodic features of traffic flow data in the time domain and the correlation features in the spatial domain. Hessian-free optimization (HF) is introduced to iteratively update the objective function, which can achieve robust inference and high-precision reconstruction of traffic data at missing locations without explicitly calculating the Hessian matrix.

[0129] S3 (Multi-source data integration set): The high-precision dynamic traffic flow data completed in step S2 is integrated with the land use weights, POI distribution density and road network spatial constraints obtained in step S1 to establish a multi-source input dataset covering dynamic traffic demand, static infrastructure support and land use attributes.

[0130] S4 (Multi-objective optimization): The construction status and scale of candidate sites are used as decision variables, and the objective functions are to maximize service coverage, minimize the average walking distance of users, and minimize the comprehensive land and construction costs. At the same time, constraints on land use compliance, minimum regional spacing, and total budget are introduced to transform the site selection problem into a high-dimensional combinatorial optimization model and construct a nonlinear combinatorial optimization space.

[0131] S5 (Solving Pareto Optimal Sets): Input the multi-source data from step S3 into the multi-objective optimization solver and perform a global search using a non-dominated sorting genetic algorithm with an elitist strategy; generate a set of non-dominated Pareto optimal sets through iterative evolution.

[0132] S6 (Determine the target location): Each Pareto optimal solution set represents a location combination that takes into account the objectives of all parties. The final output is a set of location schemes that take into account both social and economic benefits. Based on the final decision weight, the optimal parking lot location is recommended from the set.

[0133] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0134] Corresponding to the parking lot site selection method based on spatiotemporal tensor augmentation in the above embodiments, Figure 9 This is a structural block diagram of the parking lot location selection device 9 based on spatiotemporal tensor enhancement provided in the embodiments of this application. For ease of explanation, only the parts related to the embodiments of this application are shown.

[0135] Reference Figure 9 The device 9 includes: The data acquisition module 91 is used to acquire sparse traffic flow data and multi-source heterogeneous geographic data of the target area; wherein, the target area contains multiple candidate addresses; The data completion module 92 is used to complete sparse traffic flow data to obtain the completed target traffic flow data. Feature construction module 93 is used to construct a feature set corresponding to each candidate address based on target traffic flow data and multi-source heterogeneous geographic data; The function construction module 94 is used to construct a comprehensive objective function corresponding to each candidate address based on the feature set and the preset objective; wherein, the preset objective represents a pre-defined optimization direction and evaluation criterion; The function solving module 95 is used to solve each of the comprehensive objective functions and obtain the solution result of each of the comprehensive objective functions; The target address determination module 96 is used to determine the target location address corresponding to the parking lot from multiple candidate addresses based on multiple solution results.

[0136] Optionally, the data completion module 92 is also used for: Spatiotemporal mapping of sparse traffic flow data yields three-dimensional tensor data; A first objective function is constructed based on three-dimensional tensor data and multiple preset core tensors and latent feature matrices; the first objective function is expressed as:

[0137] in, Represents the position of three-dimensional tensor data X. Observations at; These are the elements corresponding to the latent feature matrix A obtained by decomposing the three-dimensional tensor data; These are the elements corresponding to the latent feature matrix B obtained by decomposing the three-dimensional tensor data; These are the elements corresponding to the latent feature matrix C obtained by decomposing the three-dimensional tensor data; The core tensor obtained from the decomposition of three-dimensional tensor data Element; For elements The rank of the corresponding dimension; For elements The rank of the corresponding dimension; For elements The rank of the corresponding dimension; The regularization coefficient is used. As an indicator variable, it is used to characterize the sparse structure of the three-dimensional tensor data X.

[0138] The first objective function is used to minimize the deviation between the target traffic flow data and the sparse traffic flow data; the core tensor is used to characterize the global distribution characteristics and cross-modal interaction of the sparse traffic flow data in the third-order tensor space; the latent feature matrix is ​​used to characterize the spatiotemporal local evolution law and inherent attribute characteristics of the sparse traffic flow data under different modes. The sparse traffic flow data is completed by performing data augmentation based on the first objective function, resulting in the completed target traffic flow data.

[0139] Optionally, the data completion module 92 is also used for: Update the preset core tensor and latent feature matrix according to the first objective function to obtain the target core tensor and target latent feature matrix; Data completion is performed based on the target core tensor and the target latent feature matrix to obtain the target traffic flow data.

[0140] Optionally, the data completion module 92 is also used for: Construct a system of linear equations based on the first objective function; Solve the system of linear equations to obtain the updated step size; Update multiple preset core tensors and latent feature matrices in the first objective function according to the update step size until the first objective function converges; When the first objective function converges, the currently updated core tensor is determined as the target core tensor, and the currently updated latent feature matrix is ​​determined as the target latent feature matrix.

[0141] Optionally, feature building module 93 is also used for: For each candidate address, calculate the average traffic flow data of the candidate address within a preset time period based on the target traffic flow data; Calculate the point of interest density index of candidate addresses within a preset range based on multi-source heterogeneous geographic data; The candidate addresses are assigned a first weighting coefficient based on multi-source heterogeneous geographic data; The betweenness centrality of candidate addresses is determined based on multi-source heterogeneous geographic data; where betweenness centrality is used to characterize the degree of coreness of candidate addresses in the road network topology. The feature set of candidate addresses is obtained by combining average traffic flow data, point of interest density index, first weight coefficient, and betweenness centrality.

[0142] Optionally, function building module 94 is also used for: For each candidate address, a second objective function is constructed based on at least one preset objective; the second objective function can be any of the following:

[0143] in, For the weight vector, For the first n Candidate site selection points Construction status variables; The candidate site selection point The scale of construction; The candidate site selection point Feature set;

[0144] in, For parking demand points, The set of parking demand points Any parking demand point; Indicates the first n Construction status variables of candidate site selection points Indicates construction, It indicates construction. Indicate demand points m with candidate site selection n The actual walking distance between them.

[0145]

[0146] in, To fix startup costs, For construction costs; A comprehensive objective function is constructed based on multiple secondary objective functions and preset constraints; the comprehensive objective function is:

[0147] The constraints include land capacity limitations. Minimum allowed distance threshold between candidate address points Total project budget and 0-1 variable constraints for the construction status; among which, For the first 10 candidate address points.

[0148] Optionally, the target address determination module 96 is also used for: For each solution result, the solution result is weighted and summed according to multiple preset second weight coefficients to obtain the comprehensive score of each candidate address; The highest score is selected from the comprehensive scores corresponding to each of the multiple candidate addresses; The candidate address corresponding to the highest score is determined as the target address.

[0149] It should be noted that the information interaction and execution process between the above-mentioned devices / units are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, and they will not be repeated here.

[0150] in addition, Figure 9 The parking lot location selection device based on spatiotemporal tensor augmentation shown can be a software unit, a hardware unit, or a combination of software and hardware built into existing terminal equipment. It can also be integrated into the terminal equipment as an independent component, or exist as an independent terminal equipment.

[0151] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0152] Figure 10 This is a schematic diagram of the structure of the terminal device provided in the embodiments of this application. For example... Figure 10 As shown, the terminal device 10 of this embodiment includes: at least one processor 100 ( Figure 10 (Only one is shown in the diagram) a processor, a memory 101, and a computer program 102 stored in the memory 101 and capable of running on at least one processor 100. When the processor 100 executes the computer program 102, it implements the steps in any of the above embodiments of the parking lot location method based on spatiotemporal tensor enhancement.

[0153] The terminal device can be a computing device such as a desktop computer, laptop, handheld computer, or cloud server. This terminal device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that... Figure 10This is merely an example of terminal device 10 and does not constitute a limitation on terminal device 10. It may include more or fewer components than shown, or combine certain components, or different components, such as input / output devices, network access devices, etc.

[0154] The processor 100 may be a Central Processing Unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0155] In some embodiments, memory 101 may be an internal storage unit of terminal device 10, such as a hard disk or memory of terminal device 10. In other embodiments, memory 101 may be an external storage device of terminal device 10, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc., equipped on terminal device 10. Furthermore, memory 101 may include both internal and external storage units of terminal device 10. Memory 101 is used to store operating system, applications, boot loader, data, and other programs, such as program code of computer programs. Memory 101 may also be used to temporarily store data that has been output or will be output.

[0156] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, can implement the steps in the above-described method embodiments.

[0157] This application provides a computer program product that, when run on a terminal device, enables the terminal device to implement the steps described in the various method embodiments above.

[0158] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of this application can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. A computer-readable medium can include at least: any entity or device capable of carrying computer program code to a device / terminal equipment, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electrical carrier signals or telecommunication signals.

[0159] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0160] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0161] In the embodiments provided in this application, it should be understood that the disclosed apparatus / terminal devices and methods can be implemented in other ways. For example, the apparatus / terminal device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0162] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0163] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A spatiotemporal tensor augmentation based parking lot siting method, characterized in that, The method includes: Acquire sparse traffic flow data and multi-source heterogeneous geographic data of the target area; wherein, the target area contains multiple candidate addresses; The sparse traffic flow data is augmented to obtain the augmented target traffic flow data; Construct a feature set corresponding to each candidate address based on the target traffic flow data and the multi-source heterogeneous geographic data; A comprehensive objective function is constructed for each candidate address based on the feature set and the preset objective; wherein, the preset objective represents a pre-defined optimization direction and evaluation criterion; Solve each of the aforementioned integrated objective functions to obtain the solution results for each of the aforementioned integrated objective functions; The target location address corresponding to the parking lot is determined from the candidate addresses based on the multiple solution results.

2. The spatiotemporal tensor augmentation based parking lot siting method of claim 1, wherein, The step of completing the sparse traffic flow data to obtain the completed target traffic flow data includes: The sparse traffic flow data is spatiotemporally mapped to obtain three-dimensional tensor data; A first objective function is constructed based on the aforementioned three-dimensional tensor data and the preset core tensor and latent feature matrix; the first objective function is expressed as: in, This indicates that the three-dimensional tensor data X is at position. Observations at; These are the elements corresponding to the latent feature matrix A obtained by decomposing the three-dimensional tensor data; These are the elements corresponding to the latent feature matrix B obtained by decomposing the three-dimensional tensor data; These are the elements corresponding to the latent feature matrix C obtained by decomposing the three-dimensional tensor data; The core tensor obtained by decomposing the three-dimensional tensor data. Element; For elements The rank of the corresponding dimension; For elements The rank of the corresponding dimension; For elements The rank of the corresponding dimension; The regularization coefficient is used. As an indicator variable, it is used to characterize the sparse structure of the three-dimensional tensor data X; Wherein, the first objective function is used to minimize the deviation between the target traffic flow data and the sparse traffic flow data; the core tensor is used to characterize the global distribution characteristics and cross-modal interaction of the sparse traffic flow data in the third-order tensor space; the latent feature matrix is ​​used to characterize the spatiotemporal local evolution law and inherent attribute characteristics of the sparse traffic flow data under different modes. The sparse traffic flow data is completed according to the first objective function to obtain the completed target traffic flow data.

3. The parking lot site selection method based on spatiotemporal tensor augmentation as described in claim 2, characterized in that, The step of performing data completion on the sparse traffic flow data according to the first objective function to obtain the completed target traffic flow data includes: The target core tensor and the target hidden feature matrix are obtained by updating the preset core tensor and the hidden feature matrix according to the first objective function. The sparse traffic flow data is completed by using the target core tensor and the target latent feature matrix to obtain the target traffic flow data.

4. The parking lot site selection method based on spatiotemporal tensor augmentation as described in claim 3, characterized in that, The step of updating the preset core tensor and the latent feature matrix according to the first objective function to obtain the target core tensor and the target latent feature matrix includes: Construct a system of linear equations based on the first objective function; Solve the system of linear equations to obtain the updated step size; The core tensor and the latent feature matrix in the first objective function are updated according to the update step size until the first objective function converges. When the first objective function converges, the currently updated core tensor is determined as the target core tensor, and the currently updated latent feature matrix is ​​determined as the target latent feature matrix.

5. The parking lot site selection method based on spatiotemporal tensor augmentation as described in claim 4, characterized in that, The step of constructing a feature set corresponding to each candidate address based on the target traffic flow data and the multi-source heterogeneous geographic data includes: For each candidate address, the average traffic flow data of the candidate address within a preset time period is calculated based on the target traffic flow data; Calculate the point of interest density index of the candidate address within a preset range based on the multi-source heterogeneous geographic data; A first weighting coefficient is assigned to the candidate address based on the multi-source heterogeneous geographic data; The betweenness centrality of the candidate addresses is determined based on the multi-source heterogeneous geographical data; wherein, the betweenness centrality is used to characterize the degree of coreness of the candidate addresses in the road network topology; The feature set of the candidate address is obtained by combining the average traffic flow data, the point of interest density index, the first weighting coefficient, and the betweenness centrality.

6. The parking lot site selection method based on spatiotemporal tensor augmentation as described in claim 5, characterized in that, The step of constructing a comprehensive objective function corresponding to each candidate address based on the feature set and a preset target includes: For each candidate address, a second objective function is constructed based on at least one preset objective; the second objective function can be any of the following: in, For the weight vector, For the first n Candidate site selection points Construction status variables; For the candidate site selection point The scale of construction; For the candidate site selection point Feature set; in, For parking demand points, The set of parking demand points Any parking demand point; Indicates the first n Construction status variables of candidate site selection points Indicates construction, It indicates construction. Indicate demand points m with candidate site selection n The actual walking distance between them. in, To fix startup costs, For construction costs; The comprehensive objective function is constructed based on multiple second objective functions and preset constraints; the comprehensive objective function is: The constraints include land capacity limitations. Minimum allowed distance threshold between candidate address points Total project budget and 0-1 variable constraints for the construction status; among which, For the first 10 candidate address points.

7. The parking lot site selection method based on spatiotemporal tensor augmentation as described in claim 5, characterized in that, The step of determining the target location address corresponding to the parking lot from the multiple candidate addresses based on the multiple solution results includes: For each solution result, the solution results are weighted and summed according to a plurality of preset second weight coefficients to obtain a comprehensive score for each candidate address; The highest score is selected from the comprehensive scores corresponding to each of the multiple candidate addresses; The candidate address corresponding to the highest score is determined as the target address.

8. A parking lot location selection device based on spatiotemporal tensor augmentation, characterized in that, include: The data acquisition module is used to acquire sparse traffic flow data and multi-source heterogeneous geographic data of the target area; wherein, the target area contains multiple candidate addresses; The data completion module is used to complete the sparse traffic flow data to obtain the completed target traffic flow data. The feature construction module is used to construct a feature set corresponding to each candidate address based on the target traffic flow data and the multi-source heterogeneous geographic data; The function construction module is used to construct a comprehensive objective function corresponding to each candidate address based on the feature set and the preset objective; wherein, the preset objective represents a pre-defined optimization direction and evaluation criterion; The function solving module is used to solve each of the comprehensive objective functions and obtain the solution result of each of the comprehensive objective functions; The target address determination module is used to determine the target location address corresponding to the parking lot from multiple candidate addresses based on multiple solution results.

9. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 7.

Images