Macroscopic region division method based on urban traffic demand

Abstract

The invention relates to the field of urban traffic management and intelligent traffic, and discloses a macroscopic region division method based on urban traffic demands, which comprises the following steps: acquiring traffic node traffic demands of an urban dynamic road network, and constructing a traffic demand adjacency matrix, a degree matrix and a Laplacian matrix, utilizing a spectral decomposition method to extract aggregation degree features among the nodes; a K-means clustering algorithm is adopted to classify the node feature vectors, and a macroscopic region is preliminarily divided; gPS normalized coordinate features are introduced, node space distribution is modeled by using two-dimensional Gaussian distribution, and a comprehensive distance optimization region division result is calculated in combination with a spectral clustering feature distance; and finally, iteratively optimizing category distribution based on the comprehensive distance to obtain a macroscopic region division scheme with the minimum segmentation loss. According to the method, the rationality of traffic area division can be improved, the macroscopic area coverage range is optimized, the scientificity of traffic management is enhanced, and the urban traffic resource allocation efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the field of urban traffic management and intelligent traffic technology, and in particular to a macro-region division method based on urban traffic demand. Background Art

[0002] With the acceleration of urbanization, the complexity of the traffic network continues to increase. Reasonable traffic area division is crucial for optimizing urban traffic management and resource allocation. Traditional traffic area division methods mainly rely on administrative divisions or empirical division methods, lacking sufficient consideration of dynamic factors such as traffic flow and the strength of connections between nodes, resulting in the difficulty of regional division results to adapt to actual traffic needs. In recent years, data-driven clustering analysis methods have been introduced into the field of traffic area division, among which K-means, hierarchical clustering and other methods have improved the rationality of division to a certain extent. However, these methods often only calculate based on traffic flow or geographical location, ignoring the topological structure characteristics of the traffic network, so that the traffic connection within the same category may be weak, and the traffic interaction between different categories is still significant, which reduces the scientificity and effectiveness of the division.

[0003] On the other hand, spectral clustering methods are widely used in network partitioning problems because they can make full use of graph structure information. However, in urban traffic applications, relying solely on spectral decomposition features for clustering may lead to some areas covering too wide a range, or even overlap between areas, making actual management more difficult. In addition, the complexity of urban road networks leads to significant imbalance in traffic demand. Traditional spectral clustering methods fail to fully consider this factor, resulting in some nodes with larger weights being aggregated into a large single area, while node clusters with smaller weights are isolated and fragmented, affecting the overall partitioning effect.

[0004] In practical applications, the rational division of macro-regions not only needs to consider the topological relationship of the traffic network, but also needs to be combined with geographic spatial characteristics to ensure that the regional scope is moderate, the boundaries are clear, and it is easy to manage and regulate. However, existing technologies make it difficult to effectively constrain the regional spatial scope while ensuring the rationality of traffic connections, making the division results less operational in actual traffic management. Summary of the invention

[0005] In view of the shortcomings of the existing technology, the present invention provides a macro-regional division method based on urban traffic demand, which solves the problem that the existing traffic area division method fails to fully combine traffic topological relationships and spatial distribution characteristics, resulting in unreasonable regional division, blurred boundaries and high management difficulty.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: A macro-region division method based on urban traffic demand includes the following steps: Obtain real-time traffic demand data for multiple traffic nodes in the urban transportation network; Based on the traffic demand data, construct an adjacency matrix, a degree matrix and a Laplace matrix of the traffic network; Performing spectral decomposition on the Laplace matrix to extract spectral feature vectors of traffic nodes; Based on the spectral feature vector, the traffic nodes are initially divided into multiple macro-regions using clustering algorithm; The divided macro-regions are optimized in combination with the spatial location information of the traffic nodes to obtain the final region division result.

[0007] Preferably, the traffic node is an urban road intersection or road section, and the traffic demand data includes at least one of Internet travel platform data, traffic detection equipment data, and bus route data.

[0008] Preferably, the adjacency matrix is ​​constructed as follows: If the traffic node v i With v j If there is a traffic demand association between ij =T(i→j)+T(j→i), otherwise w ij =0; Where T(i→j) represents the node v i to v j of traffic flow.

[0009] Preferably, the Laplace matrix is ​​a normalized Laplace matrix, which is constructed as follows: L norm =D -1 / 2 (DW)D -1 / 2 Among them, W is the adjacency matrix, D is the degree matrix, and D -1 / 2 is the inverse square root of the degree matrix.

[0010] Preferably, the step of performing spectral decomposition on the Laplace matrix to extract characteristic vectors of traffic nodes comprises: Perform eigenvalue decomposition on the normalized Laplace matrix, extract the eigenvectors corresponding to the 2k smallest eigenvalues, and construct a feature matrix.

[0011] Preferably, the clustering algorithm is a K-means algorithm, comprising: Initialize cluster centers; Calculate the Euclidean distance from each node to the cluster center; Iteratively update the node category affiliation until the clustering is stable.

[0012] Preferably, the step of optimizing the divided macro-areas in combination with the spatial location information of the traffic nodes comprises: Initialize geographic spatial region division based on GPS coordinates of traffic nodes; Calculate the mean and variance of GPS coordinates of nodes in each category and fit a two-dimensional Gaussian distribution; Calculate the comprehensive distance of the spectral features and GPS coordinate features of the node to each category; Update the node category according to the comprehensive distance until the category division is stable.

[0013] Preferably, the GPS spatial probability distance is a probability density value of a two-dimensional Gaussian distribution, and the calculation formula is: Where (x, y) is the GPS coordinate of the node, (μ x ,μ y ) represents the mean of GPS coordinates of all nodes in the current category, σ x ,σ y Indicates the standard deviation of the GPS coordinates of all nodes in the current category.

[0014] Preferably, the calculation formula of the comprehensive distance is: d final =d spectral ×P(x,y) Where: d spectral is the Euclidean distance between spectral clustering feature vectors; P(x,y) is the GPS spatial probability distance.

[0015] The macro-region division device based on urban traffic demand of the present invention comprises: A data acquisition module is used to obtain real-time traffic demand data of traffic nodes; Matrix building modules for generating adjacency matrices, degree matrices, and normalized Laplacian matrices; A spectrum decomposition module, used for performing eigenvalue decomposition on the normalized Laplace matrix and extracting spectrum feature vectors; Clustering module, used to divide the preliminary regions by K-means algorithm; The optimization module is used to spatially optimize the regional division results in combination with GPS coordinates.

[0016] The present invention provides a macro-region division method based on urban traffic demand, which has the following beneficial effects: 1. The present invention extracts the structural characteristics of the traffic network through spectral decomposition and classifies the nodes in combination with K-means clustering, so that the nodes in the same category have higher traffic connections, and the connections between nodes in different categories are weaker, ensuring that the regional division conforms to the actual flow law of urban traffic demand.

[0017] 2. The traditional clustering-based division method may cause some areas to span too large or overlap between areas, affecting the effectiveness of management. The present invention introduces GPS normalized coordinate features in the clustering process to ensure that the spatial range of the macro-region is reasonable, avoid cross-regional overlap, and improve management efficiency.

[0018] 3. By calculating the segmentation loss between traffic nodes and combining it with geographic spatial distribution information, the present invention realizes the precise partitioning of the traffic network, provides clear regional boundaries for traffic managers, helps to formulate more scientific traffic planning and flow control strategies, and improves the operability of traffic management.

[0019] 4. Reasonable zoning of traffic demand helps to more accurately layout and optimize traffic infrastructure. The present invention can provide a scientific basis for public transportation line planning, signal control optimization, road resource scheduling, etc. according to traffic flow characteristics and regional distribution, improve resource utilization, and alleviate traffic congestion.

[0020] 5. By using two-dimensional Gaussian distribution to fit the regional GPS spatial distribution, the present invention can intuitively display the traffic distribution conditions in different regions, enabling managers to understand the urban traffic flow pattern more clearly and provide support for the subsequent construction of intelligent transportation systems, regional development planning, etc. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a schematic diagram of the method flow of the present invention; Figure 2 It is a schematic diagram of the structure of the device of the present invention; Figure 3 A schematic diagram of a method flow of an embodiment of the present invention; Figure 4 Schematic diagram of traffic node demand distribution of a road section during the morning peak and evening peak periods in a city according to an embodiment of the present invention; (a) is the morning peak period, and (b) is the evening peak period; Figure 5 Schematic diagram of preliminary area division results of an embodiment of the present invention; (a) is the morning peak, (b) is the evening peak; Figure 6 Schematic diagram of the regional division result after adding GPS normalization features in an embodiment of the present invention; (a) is the morning peak and (b) is the evening peak.

[0022] Among them, 10, data acquisition module; 20, matrix construction module; 30, spectral decomposition module; 40, clustering module; 50, optimization module. DETAILED DESCRIPTION

[0023] The following will be combined with the drawings in the specification of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0024] Please refer to the attached Figure 1 The present invention provides a macro-region division method based on urban traffic demand. By combining the topological relationship, spatial distribution characteristics and spectral clustering technology of urban traffic demand, a dynamic and reasonable division of traffic areas is achieved, thereby optimizing urban traffic management and improving the efficiency of traffic dispatching.

[0025] like Figure 1 As shown, the macro-region division method based on urban traffic demand may include the following steps: S1. Obtain real-time traffic demand data of multiple traffic nodes in the urban traffic network; S2. Based on the traffic demand data, construct the adjacency matrix, degree matrix and Laplace matrix of the traffic network; S3, performing spectral decomposition on the Laplace matrix to extract the characteristic vector of the traffic node; S4, based on the feature vector, the traffic node is divided into multiple macro areas using clustering algorithm; S5. Optimize the divided macro-areas based on the spatial location information of the traffic nodes.

[0026] The following is a detailed description of each step in the method of the present invention, which comprehensively describes the specific implementation principle, technical details and process of each step.

[0027] For step S1, this step mainly obtains the real-time traffic demand data of multiple traffic nodes in the urban traffic network and pre-processes the data to provide reliable data support for subsequent steps. The implementation process includes four aspects: data source, data type, data pre-processing and data storage, ensuring that the traffic demand data can accurately reflect the current traffic conditions and provide effective input for subsequent steps.

[0028] First, traffic demand data can be obtained in the following ways: 1. Internet travel platform data Through the commonly used online car-hailing and shared bicycle platforms in the city, various travel information is collected in real time. The data provided by these platforms include but are not limited to: Departure point and destination coordinates; The start and end times of the trip; Travel type (such as taxi, shared bike, taxi, etc.); Number of people travelling and frequency of travel.

[0029] With this information, the traffic intensity between traffic nodes can be estimated, and the traffic demand of each node can be further calculated.

[0030] 2. Traffic detection equipment data Traditional traffic detection equipment (such as geomagnetic detectors, traffic cameras, radar speed guns, etc.) provides real-time traffic flow information, including: Traffic volume per unit time; traffic speeds and vehicle types; Traffic light status and road congestion.

[0031] This information can reflect the congestion level of traffic nodes in real time and assist in traffic demand analysis.

[0032] 3. Bus route data Through floating car technology (FCD), the trajectory data of buses flowing in the city can be collected in real time. These data include: GPS coordinates of the vehicle; The travel time and path of the vehicle; The length of time the vehicle stayed, where it stayed, and the number of passengers.

[0033] By processing these data, the traffic conditions between different traffic nodes can be estimated and the deficiencies of other data sources can be supplemented.

[0034] The acquisition of data depends on multiple channels and sensors, and data from different sources have different characteristics and accuracy. Therefore, data fusion and processing are required to ensure the integrity and consistency of the data.

[0035] Data preprocessing includes the following aspects: 1. Denoising After receiving the traffic data, the first thing to do is to remove the noise in the data. Common noise includes GPS signal loss, traffic sensor failure data, etc. These data need to be cleaned by a filtering algorithm. Preferably, a Kalman filter algorithm is used to process the trajectory data to reduce errors.

[0036] 2. Time synchronization Since the collection time of each data source may be different, it is necessary to synchronize the data from different sources. The specific method is to match the data at different time nodes through interpolation algorithms to ensure that all data are consistent at the same time point.

[0037] 3. Spatial matching Since the spatial positions reflected by different data sources (such as GPS trajectory data and road sensor data) are not completely consistent, it is necessary to map all data to a unified traffic node coordinate system. This process is achieved through spatial interpolation technology, which maps traffic data from different sources to specific traffic nodes.

[0038] In terms of data storage, the preprocessed data needs to be stored according to traffic nodes. The data of each traffic node includes: Node ID; The traffic demand intensity of the node; Corresponding traffic flow, travel frequency and other values.

[0039] The data can be stored in a database management system (such as MySQL, PostgreSQL) for quick extraction during subsequent calculations.

[0040] Through the above steps, accurate urban traffic demand data can be obtained and preprocessed, providing effective input for traffic network modeling, spectral decomposition and regional division in subsequent steps.

[0041] As for step S2, this step constructs the adjacency matrix, degree matrix and Laplace matrix of the traffic network through the traffic demand data, providing basic data support for subsequent spectral decomposition and clustering analysis.

[0042] Firstly, based on the traffic demand data, the adjacency matrix, degree matrix and Laplace matrix of the transportation network are constructed through the following process.

[0043] The adjacency matrix W in the transportation network is a matrix that represents the interconnection relationship between transportation nodes, where the elements of the matrix w are ij Represents node v i With node v j The traffic demand intensity between . The elements of the adjacency matrix can be calculated by the following formula: w ij =T(i→j)+T(j→i) Where T(i→j) represents the distance from the traffic node v i to v j The traffic flow or demand intensity, T(j→i) represents the traffic flow from the traffic node v j to v i If there is no traffic flow between two nodes, then w ij =0.

[0044] The degree matrix D is a diagonal matrix with the elements D on the diagonal ii Represents node v i The degree of node v iThe total traffic demand intensity connected to other nodes. The elements of the degree matrix can be calculated by the following formula: That is, the diagonal elements of the degree matrix are the sum of the traffic demand intensities of the node and all other nodes.

[0045] The Laplace matrix L is a commonly used matrix in graph theory and is widely used in graph clustering analysis. The Laplace matrix reflects the connection strength and topological structure between traffic nodes. In the present invention, the normalized Laplace matrix L is used. norm , which is defined as: L norm =D -1 / 2 (DW)D -1 / 2 Where D is the degree matrix, W is the adjacency matrix, and D -1 / 2 is the inverse square root of the degree matrix D.

[0046] The normalized Laplace matrix is ​​normalized by introducing the degree matrix D, so that the influence of each node is no longer directly affected by the node degree, ensuring that the connection relationship between nodes can be fairly reflected in the matrix. This matrix provides the basis for spectral clustering analysis.

[0047] Through the above steps, the adjacency matrix, degree matrix and Laplace matrix reflecting the urban traffic network structure and traffic demand can be constructed. It can accurately reflect the connection relationship and traffic distribution between nodes in the urban traffic network, and provide basic data for subsequent spectral decomposition and clustering analysis; at the same time, it ensures the normalization of node degrees during the calculation process, which helps to improve the accuracy and effectiveness of spectral clustering.

[0048] For step S3, this step performs spectral decomposition on the Laplace matrix to extract the solution information with the minimum segmentation loss between nodes, providing a basis for subsequent regional division. Specifically, the spectral decomposition process can divide the traffic nodes according to the structural characteristics of the traffic network, so that the nodes with larger connection weights are divided into the same category, achieving the effect of minimizing the segmentation loss.

[0049] First, the Laplace matrix is ​​spectrally decomposed to extract the solution information with the minimum segmentation loss between nodes and provide a basis for subsequent regional division. The spectral decomposition process divides the traffic nodes based on the structural characteristics of the traffic network, so that nodes with larger connection weights are classified into the same category, thereby minimizing the segmentation loss.

[0050] Given the Laplace matrix L of the transportation network norm , which describes the connection relationship and traffic intensity between traffic nodes. The goal of the spectral decomposition process is to find a segmentation scheme that minimizes the graph segmentation loss through eigenvalue decomposition. Let the row vector c = [c1 ,c 2 ,...,c n ] T Represents the category of each traffic node, where c i Indicates the category of node i, and the connection weight between node i and node j is recorded as w ij The objective function is expressed as: If node i and node j belong to the same category, then c i =c j , in which case the term in the loss function is zero. If nodes i and j belong to different categories, the value of this term increases. Therefore, by minimizing the segmentation loss, the objective function divides nodes with large connection weights that need to be separated into different categories, while nodes with small connection weights are divided into the same category.

[0051] To solve this optimization problem, we need to normalize the Laplace matrix L norm Perform eigenvalue decomposition and get: L norm v i =λ i v i Among them, λ i is the i-th eigenvalue of the Laplace matrix, v i is the corresponding eigenvector. After eigenvalue decomposition, n eigenvalues ​​and their corresponding n-dimensional eigenvectors can be obtained. The relationship between eigenvalues ​​and eigenvectors can reveal the segmentation relationship between nodes. The segmentation scheme corresponding to the eigenvector with a smaller eigenvalue indicates the minimum segmentation loss.

[0052] The feature vector is a continuous value. Therefore, in order to facilitate the subsequent clustering process, it is necessary to select the feature vectors corresponding to the smallest k to 2k feature values ​​as effective features, where k is the number of pre-set clustering categories. These feature vectors correspond to the smallest segmentation loss and can effectively reflect the clustering structure in the traffic network. The selected feature vectors form the feature matrix F, which is used as the input of the K-means clustering algorithm to achieve the final node category division.

[0053] By decomposing the spectral matrix of the Laplace matrix and selecting the eigenvector, the graph segmentation loss can be minimized, so that nodes with larger connection weights are classified into the same class, while nodes with smaller connection weights are classified into different classes. This method effectively extracts the segmentation relationship between nodes through eigenvalue decomposition, providing reliable data support for subsequent clustering analysis. By selecting the eigenvector corresponding to the minimum eigenvalue, the accuracy of the clustering results is improved, so that the final traffic node division can be more in line with actual traffic needs and regional characteristics, providing an optimization solution for urban traffic management.

[0054] For step S4, based on the spectral decomposition results of the previous step, the K-means algorithm is used to cluster the feature vectors to complete the regional division of traffic nodes. The goal of this step is to map the feature vectors to a limited number of categories based on the segmentation relationship between nodes, so that similar nodes are assigned to the same area, and to achieve reasonable regional division based on traffic demand.

[0055] First, the feature matrix F extracted in step S3 is used as the input of K-means clustering, where F consists of k to 2k feature vectors, each of which represents a low-dimensional feature of a traffic node. The goal of the K-means clustering algorithm is to divide traffic nodes into k categories through iterative optimization, so that nodes in the same category have a high traffic connection, while nodes in different categories have a weak traffic connection.

[0056] Define the cluster center matrix C, where C = [c 1 ,c 2 ,...,c k ] represents the center point of k categories. The clustering process first randomly initializes k cluster centers, and then calculates the distance between each node x and i Distance to each cluster center: d(x i ,c j )=∥x i -c j ∥ Among them, x i is the i-th row data in the feature matrix F, representing the feature representation of node i, c j is the cluster center of the current category j. Assign each node to the cluster center c with the smallest distance j Corresponding categories: cluster(x i ) = argmin j d(x i ,c j ) After completing a node assignment, recalculate the new cluster center of each category, that is, calculate the mean of all nodes in the category: Among them, |C j | is the number of nodes in category j, x i is the feature representation of all nodes in this category. After this step is completed, the cluster center is updated and the above calculation is repeated until the cluster center no longer changes significantly or the set maximum number of iterations is reached.

[0057] The iterative termination conditions of K-means clustering include: all cluster centers converge, that is, the difference between the newly calculated center and the previously calculated center is within the set threshold range, or the preset maximum number of iterations is reached. Ultimately, the K-means clustering result will give the category allocation corresponding to each node, that is, the regional division plan of the transportation network.

[0058] For step S5, based on the spectral decomposition and K-means clustering results of the previous steps, the regional spatial range is optimized by introducing the node normalized GPS coordinate features, making the divided macro-region more reasonable, avoiding excessive coverage and overlap of the region, and improving the feasibility of management. The goal of this step is to comprehensively utilize the topological relationship of the node's traffic demand and the spatial distribution characteristics to establish a visual traffic region model and achieve more accurate regional division.

[0059] First, the n nodes are initially classified into k categories using the correspondence between the GPS coordinates of the nodes and the Geohash grid to form a preliminary spatial region division scheme. Geohash encoding is used to map geographic coordinates to a specific spatial grid so that the spatial relationship between adjacent nodes can be maintained. For each category, the mean and variance of the longitude and latitude of all nodes in it are calculated, and a two-dimensional Gaussian distribution is used to fit the GPS spatial distribution of the category: Among them, μ x ,μ y Respectively represent the mean longitude and latitude of the nodes in this category, σ x ,σ y Represent the standard deviation of longitude and latitude respectively. This step is used to visualize the spatial distribution of the region on the map, and the iteration can be stopped when the spatial distribution of the category GPS tends to be stable.

[0060] Then, the mean center of the spectral clustering feature of each category is calculated as the spectral clustering feature center of the category. Define the mean center vector v of the spectral clustering feature c : Among them, C is the node set of the current category, v i is the spectral decomposition feature vector of node i. This center point is used to calculate the clustering distance from the node to the category.

[0061] Next, the comprehensive distance from each node to each category center is calculated, taking into account the influence of GPS coordinate distance and spectral clustering feature distance. Spectral clustering feature distance calculates the Euclidean distance between the spectral decomposition feature of each node and the spectral feature of the category center: d spectral (v i ,v c )=∥vi -v c ∥ Finally, calculate node i to category C j The comprehensive distance is expressed by the product of the spectral clustering feature distance and the GPS spatial probability distance: d final =d spectral ×P(x,y) Among them, d spectral Reflecting the differences in topological features of nodes, P(x,y) represents the spatial belonging probability of a node in a category. This calculation method comprehensively considers the topological structure and spatial position of the node.

[0062] Each node is assigned to the category with the smallest distance, the node category label is updated, and the category spatial distribution calculation step is returned to perform the next round of iteration until the category GPS spatial distribution is stable. Through this method, the spatial coverage of the macro area is more reasonable and the operability of urban traffic management is improved.

[0063] This method makes comprehensive use of traffic topological relationships and spatial distribution characteristics to make regional division more in line with actual traffic needs, optimize traffic flow analysis, improve the level of urban traffic management, and provide a scientific basis for traffic planning and resource allocation.

[0064] The apparatus for dividing a macro region based on urban traffic demand described below and the method for dividing a macro region based on urban traffic demand described above can refer to each other.

[0065] Please refer to the attached Figure 2 The present invention also provides a macro-region division device based on urban traffic demand, comprising: The data acquisition module 10 is used to acquire the real-time traffic demand data of the traffic nodes; A matrix construction module 20, for generating an adjacency matrix, a degree matrix and a normalized Laplace matrix; A spectrum decomposition module 30, used for performing eigenvalue decomposition on the normalized Laplace matrix and extracting eigenvectors; A clustering module 40, used for dividing preliminary regions by using a K-means algorithm; The optimization module 50 is used to perform spatial optimization on the area division result in combination with the GPS coordinates.

[0066] The device of this embodiment can be used to execute the above method embodiment, and its principles and technical effects are similar, which will not be repeated here.

[0067] Example: like Figure 3As shown, the embodiment of the present invention provides a method for dividing urban traffic regions based on the topological relationship of traffic demand and spatial distribution characteristics, with the goal of minimizing the regional segmentation loss and realizing the reasonable management of traffic nodes in the urban dynamic road network. The specific implementation steps are described below with reference to the diagram.

[0068] Step 1: Obtain traffic demand of traffic nodes in the city's dynamic road network The traffic demand of urban traffic nodes is obtained through traffic surveys, road network traffic status data from Internet travel platforms, and flow data collected by traditional traffic detection equipment.

[0069] The granularity of traffic nodes can be selected as intersections, sections or roads, and the node granularity is determined based on data accuracy and regional division requirements.

[0070] Figure 4 (a) and Figure 4 (b) shows the traffic node demand distribution of a certain city during the morning and evening peak hours.

[0071] Step 2: Calculate the traffic demand matrix 1. Construct the adjacency matrix W: Calculate the node pairs connected by traffic demand, and define the connection weight as the sum of the two-way traffic demand between the nodes: w ij =Demand from traffic node i to j + demand from traffic node j to i.

[0072] 2. Calculate the degree matrix D: The total demand weight of each node is calculated by the adjacency matrix W: 3. Calculate the Laplace matrix L: Used to measure the segmentation loss of the graph: 4. Calculate the normalized Laplace matrix L norm : Due to the uneven distribution of traffic demand caused by differences in road grades and locations, the Laplace matrix needs to be normalized: Step 3: Spectral decomposition Compute the eigenvalue decomposition: The normalized Laplace matrix L norm Perform eigenvalue decomposition to obtain several eigenvectors with minimum segmentation loss.

[0073] Let L norm There are n eigenvalues ​​and corresponding eigenvectors, and k to 2k eigenvectors whose values ​​are closest to zero are selected as clustering features.

[0074] Through spectral decomposition, the aggregation characteristics of traffic nodes are obtained. Then, the nodes are preliminarily divided into regions through K-means clustering.

[0075] Step 4: K-means clustering Preliminary clustering based on spectral decomposition features The K-means algorithm is used to cluster the spectral decomposition feature vectors to obtain k preliminary regions.

[0076] The preliminary regional division results are as follows: Figure 5 (a) and Figure 5 As shown in (b), different colors and shapes represent different types of traffic areas.

[0077] Since some macro-regions have a wide coverage area and overlap, they are difficult to manage.

[0078] It is necessary to introduce GPS normalized coordinate features to optimize the regional scope.

[0079] Step 5: Introduce GPS normalized coordinate features to optimize regional division 1. Initialize classification Through Geohash coding, n nodes are divided into k categories according to GPS coordinates as the initial clustering result.

[0080] 2. Calculate GPS statistical features Calculate the latitude and longitude mean and variance of each category, and use a two-dimensional Gaussian distribution to fit the spatial distribution of the category: 3. Calculate the feature center of spectral clustering Compute the mean center of the spectral clustering features for each class: 4. Calculate the distance from the node to the category center Calculate the spectral clustering feature distance: d spectral (v i ,v c )=∥v i -v c ∥ Calculate the comprehensive distance, combining the spectral clustering feature distance and GPS probability distribution: d final =d spectral ×P(x,y) 5. Update node category According to the principle of minimum comprehensive distance, the node category is updated and iterative calculation is performed until the category GPS spatial distribution is stable.

[0081] Figure 6 This shows the regional division results after adding GPS normalization features. Figure 5 , the optimized regional boundaries are clearer, the macro-region span is limited, and the manageability is improved.

[0082] In summary, based on the topological relationship of traffic demand and the normalized characteristics of GPS, the present invention adopts spectral decomposition and K-means clustering methods to achieve efficient urban traffic area division and improve the rationality and scientificity of traffic management.

[0083] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A macro-regional division method based on urban traffic demand, characterized in that: The following steps are involved: Obtain real-time traffic demand data for multiple traffic nodes in the urban transportation network; Based on the traffic demand data, construct an adjacency matrix, a degree matrix and a Laplace matrix of the traffic network; Performing spectral decomposition on the Laplace matrix to extract spectral feature vectors of traffic nodes; Based on the spectral feature vector, the traffic nodes are initially divided into multiple macro-regions using clustering algorithm; The macro-regions after preliminary division are optimized in combination with the spatial location information of the traffic nodes to obtain the final regional division result.

2. The macro-region division method based on urban traffic demand according to claim 1 is characterized in that: The traffic node is an intersection or road section of a city road, and the traffic demand data includes at least one of Internet travel platform data, traffic detection equipment data, and bus route data.

3. The macro-region division method based on urban traffic demand according to claim 1 is characterized in that: The adjacency matrix is ​​constructed as follows: If the traffic node v i With v j If there is a traffic demand association between ij =T(i→j)+T(j→i), otherwise w ij =0; Where T(i→j) represents the node v i to v j of traffic flow.

4. The macro-region division method based on urban traffic demand according to claim 3 is characterized in that: The Laplace matrix is ​​a normalized Laplace matrix, which is constructed as follows: L norm =D -1 / 2 (D-W)D -1 / 2 Among them, W is the adjacency matrix, D is the degree matrix, and D -1 / 2 is the inverse square root of the degree matrix.

5. The macro-region division method based on urban traffic demand according to claim 4 is characterized in that: The step of performing spectral decomposition on the Laplace matrix to extract the characteristic vector of the traffic node comprises: Perform eigenvalue decomposition on the normalized Laplace matrix, extract the eigenvectors corresponding to the 2k smallest eigenvalues, and construct the K-means clustering features.

6. The macro-region division method based on urban traffic demand according to claim 1 is characterized in that: The clustering algorithm is a K-means algorithm, including: Initialize cluster centers; Calculate the Euclidean distance from each node to the cluster center; Iteratively update the node category affiliation until the clustering is stable.

7. The macro-region division method based on urban traffic demand according to claim 6 is characterized in that: The step of optimizing the divided macro-areas in combination with the spatial location information of the traffic nodes comprises: Initialize geographic spatial region division based on GPS coordinates of traffic nodes; Calculate the mean and variance of GPS coordinates of nodes in each category and fit a two-dimensional Gaussian distribution; Calculate the comprehensive distance of the spectral features and GPS coordinate features of the node to each category; Update the node category according to the comprehensive distance until the category division is stable.

8. The macro-region division method based on urban traffic demand according to claim 7 is characterized in that: The GPS spatial probability distance is the probability density value of the two-dimensional Gaussian distribution, and the calculation formula is: Where (x, y) is the GPS coordinate of the node, μ x ,μ y represents the mean of the GPS coordinates of all nodes in the current category, σ x ,σ y Indicates the standard deviation of the GPS coordinates of all nodes in the current category.

9. The macro-region division method based on urban traffic demand according to claim 7 is characterized in that: The calculation formula of the comprehensive distance is: d final =d spectral ×P(x,y) Where: d spectral is the Euclidean distance between spectral clustering feature vectors; P(x,y) is the GPS spatial probability distance.

10. A macro-region division device based on urban traffic demand, used to execute the macro-region division method based on urban traffic demand as claimed in any one of claims 1 to 9, characterized in that: include: A data acquisition module is used to obtain real-time traffic demand data of traffic nodes; Matrix building modules for generating adjacency matrices, degree matrices, and normalized Laplacian matrices; A spectral decomposition module, used to perform eigenvalue decomposition on the normalized Laplace matrix and extract eigenvectors; Clustering module, used to divide the preliminary regions by K-means algorithm; The optimization module is used to spatially optimize the regional division results in combination with GPS coordinates.

Images