Urban comprehensive park space-time network construction method and system based on crowd trajectory data

Abstract

The invention discloses an urban comprehensive park space-time network construction method and system based on crowd trajectory data. The method comprises the following steps: firstly, collecting, cleaning and dividing crowd trajectory data in a comprehensive park range; secondly, scenic spots in the comprehensive park are clustered, the number of optimal clusters is determined through clustering analysis, and all the clusters serve as all functional activity districts; then building a trajectory model through space projection and a shortest path algorithm, recovering a user moving trajectory path according to the crowd trajectory data, generating path node sequence data according to the timestamp of each trajectory point, and finally, taking the centroid of each functional activity area as each node of the space-time network, and combining each time period to obtain the user moving trajectory path. And constructing an undirected space-time network and a directed transfer network. The method has the advantages of efficient data processing, high modeling precision and perfect privacy protection, dynamic monitoring and space optimization analysis of the park people flow can be realized, and reliable data support and decision basis are provided for urban comprehensive park planning management.

Description

Technical Field

[0001] This invention belongs to the field of data analysis technology, and in particular relates to a method and system for constructing a spatiotemporal network of urban integrated parks based on crowd trajectory data. Background Technology

[0002] With the rapid development of spatiotemporal big data technology, the widespread adoption of mobile internet and location services has brought about an unprecedented volume of population trajectory data. This data, characterized by high frequency, continuity, and broad coverage, provides new technological support for understanding human spatiotemporal behavior. Based on this, behavioral modeling methods based on population movement trajectories have gradually matured and are widely applied in various fields of urban management and spatial optimization.

[0003] In terms of research methods, existing work mainly revolves around three major directions: cluster analysis of trajectory data, behavioral pattern mining, and path prediction. On the one hand, unsupervised cluster analysis is conducted on hotspots or typical paths of population activity using indicators such as trajectory point density, trajectory morphology features, and access frequency. On the other hand, time series analysis, machine learning models, or graph neural networks are used to identify and predict behavioral patterns, uncovering the spatial movement patterns of different population groups in different time periods and scenarios. In addition, an increasing number of studies are beginning to focus on how to structurally represent trajectory data by constructing topological graph structures and spatial feature matrices, and integrating external semantic information such as POI semantics, traffic facilities, and weather to enhance the explanatory power and generalization ability of behavioral models.

[0004] At the application level, these trajectory-based big data analysis methods have achieved significant results in scenarios such as urban traffic organization optimization, subway and bus route adjustments, commercial site selection strategies, peak passenger flow prediction and management, and urban functional zone delineation. For example, transportation departments can use passenger flow transfer paths during peak hours to optimize signal timing; commercial enterprises can select the most densely populated shop areas through trajectory heat analysis; and urban planning can use trajectory clustering characteristics to dynamically identify potential residential, work, or leisure areas.

[0005] However, existing research generally relies on large-scale trajectory data in urban open spaces, and its modeling concepts and data processing strategies tend to focus on extracting behavioral features at the macro level. When dealing with enclosed micro-spaces such as urban parks, there is still a challenge of insufficient adaptability.

[0006] (1) Lacking the ability to model the evolution of crowd trajectories in enclosed spaces, it is difficult to accurately capture the real path relationships and tour behavior sequences between different attractions;

[0007] (2) Trajectory analysis objects usually stay at the level of trajectory points or clustering units, lacking a systematic expression of the "path-node-time" ternary relationship, making it difficult to support the construction and temporal analysis of spatial behavior networks;

[0008] (3) Existing models are mostly based on high-intensity mobile behaviors such as urban commuting and traffic flow, and are poorly adapted to low-intensity sightseeing behaviors such as short-term stays and free wandering in parks. Summary of the Invention

[0009] Purpose of the invention: This invention provides a method and system for constructing a spatiotemporal network for urban integrated parks based on crowd trajectory data, aiming to solve the problem.

[0010] Technical solution: This invention provides a method for constructing a spatiotemporal network for urban integrated parks based on crowd trajectory data, including:

[0011] (1) Collect crowd trajectory data, the content of which includes anonymous encrypted mobile terminal UUID, timestamp of trajectory point when mobile phone positioning occurs, and latitude and longitude coordinates of trajectory point when mobile phone positioning occurs;

[0012] (2) Clean the crowd trajectory data, retain only the crowd trajectory data within the comprehensive park area, and delete duplicate trajectory data of the same UUID and UUID and its trajectory data that appear only once;

[0013] (3) Divide the cleaned crowd trajectory data into time periods, and merge the crowd trajectory data of each time period according to the timestamp to generate a basic database for each time period and a comprehensive database for the whole day.

[0014] (4) Cluster the attractions in the comprehensive park and determine the optimal number of clusters through cluster analysis. Each cluster is used as a functional activity area. Based on the coordinates of the attractions in each cluster, construct the minimum bounding geometry of each functional activity area and extract the centroid of the minimum bounding geometry as the centroid of each functional activity area.

[0015] (5) Build a trajectory model and reconstruct the user's movement trajectory path based on the crowd trajectory data;

[0016] (6) Use the UUID field as an identifier to obtain the restored user movement trajectory path and spatially map it to each functional activity area. Extract the timestamps of each trajectory point in the user movement trajectory path according to the basic database of each time period and the comprehensive database of the whole day, and generate path node sequence data.

[0017] (7) Using the centroid of each functional activity area as the nodes of the spatiotemporal network, and combining each time period, construct an undirected spatiotemporal network and a directed spatiotemporal network.

[0018] Furthermore, the method for obtaining the optimal number of clusters through cluster analysis includes:

[0019] The K-means clustering algorithm is used to randomly select k scenic spots, each serving as the centroid of a cluster. The Euclidean distance from each scenic spot to the centroid of its cluster is iteratively calculated, and each scenic spot is clustered into the cluster with the smallest Euclidean distance. The centroids of each cluster are then updated until the clustering algorithm converges. The Euclidean distance formula is:

[0020]

[0021] Among them, C i Let C be the centroid of the i-th cluster, with coordinates (Ci, C ...). i1 C i2 ), where x is the coordinate vector of the scenic spot, and m represents the coordinate dimension;

[0022] The number of clusters k is determined by the elbow rule and the actual situation of the comprehensive park, which establishes a candidate interval for the number of clusters. The silhouette coefficient method is then used to determine the optimal number of clusters within this candidate interval. The elbow rule is established by calculating the change in the sum of squared errors (SSE) under different numbers of clusters k, using the following formula:

[0023]

[0024] Where k represents the number of clusters, and x represents the number of clusters belonging to C. i The coordinate vectors of the attractions in the cluster; when the rate of decrease of SSE slows down significantly with the increase of k, the value of k corresponding to this inflection point is the optimal number of clusters determined by the elbow rule; the clustering interval is determined based on the optimal number of clusters determined by the elbow rule and the actual situation of the comprehensive park.

[0025] The contour coefficient method calculates the contour coefficient of all attractions within the candidate interval for each cluster size k and averages them to obtain the total contour coefficient for that cluster size. By calculating the total contour coefficient for each cluster size within the clustering interval, the cluster size corresponding to the maximum total contour coefficient is taken as the optimal cluster size. The formula for the contour coefficient is:

[0026]

[0027] Where s(i) represents the contour coefficient, a(i) represents the average distance between attraction i and other attractions in the same cluster, b(i) represents the minimum average distance between attraction i and all attractions in other adjacent clusters, and n represents the number of attractions.

[0028] Furthermore, the method of building a trajectory model and reconstructing the user's movement trajectory path based on the crowd trajectory data includes: constructing a road network base map within the comprehensive park, covering roads and travel paths; identifying trajectory points in the cleaned crowd trajectory data using the UUID field, and projecting the user's trajectory points onto the nearest road or travel path on the road network base map using a spatial matching algorithm; based on the projection results, using a shortest path analysis algorithm, calculating the shortest travel path between two adjacent trajectory points in the road network, and stitching together all the user's shortest travel path segments in chronological order to build the user's trajectory model, thereby obtaining the reconstructed user movement trajectory path.

[0029] Furthermore, the method of generating path node sequence data includes: using the UUID field as an identifier, spatially overlaying the restored user movement trajectory path with each functional activity area to identify the functional activity areas traversed by the user movement trajectory path; based on the basic database for each time period and the comprehensive database for the whole day, extracting the timestamp sequence of trajectory points in the user's crowd trajectory data to obtain the order in which the user passes through each functional activity area, and constructing a scenic spot access path with time series characteristics, i.e., path node sequence data.

[0030] Furthermore, the feature is that the construction of the undirected spatiotemporal network and the directed spatiotemporal network includes:

[0031] The centroids of the functional activity zones obtained from clustering are defined as nodes in the spatiotemporal network, and a basic node graph is constructed. The complete trajectory paths reconstructed by all users are spatially overlaid with the nodes. The connection frequencies between each functional activity zone are statistically analyzed to construct an undirected node relationship matrix, generating an undirected weighted network. The network centrality is then calculated using the following formula:

[0032]

[0033] Among them, w ij This represents the number of times node i and node j appear consecutively in the user's trajectory, where n represents the number of nodes.

[0034] Based on the undirected weighted network, an undirected spatiotemporal network is constructed by combining different time periods;

[0035] Simultaneously, based on the basic database for each time period, the user trajectory paths are sorted chronologically on the node-based graph, and the visit sequences of the population between various attractions are extracted. Based on large-scale user path data, a Markov chain transition model is established, the transition probability matrix of each node within different time periods is calculated, and a directed transition network, i.e., a directed spatiotemporal network, is constructed. The state transition probability formula of the Markov chain is:

[0036]

[0037] Where, n ij This represents the total number of times a user moves from node i to node j. P represents the total number of transitions from node i to all other nodes. ij This represents the probability of transitioning from node i to node j.

[0038] This invention also provides a spatiotemporal network construction system for urban integrated parks based on crowd trajectory data, comprising:

[0039] The data collection module is used to collect crowd trajectory data, which includes the anonymous encrypted mobile terminal UUID, the timestamp of the trajectory point when the mobile phone location occurred, and the latitude and longitude coordinates of the trajectory point when the mobile phone location occurred.

[0040] The data cleaning module is used to clean crowd trajectory data, retaining only the crowd trajectory data within the comprehensive park area, and deleting duplicate trajectory data of the same UUID and UUID and its trajectory data that appear only once;

[0041] The data segmentation module is used to divide the cleaned crowd trajectory data into time periods and merge the crowd trajectory data of each time period according to the timestamp to generate the basic database for each time period and the comprehensive database for the whole day.

[0042] The attraction clustering module is used to cluster attractions in the comprehensive park, determine the optimal number of clusters through cluster analysis, and use each cluster as a functional activity area. Based on the coordinates of attractions in each cluster, the minimum bounding geometry of each functional activity area is constructed, and the centroid of the minimum bounding geometry is extracted as the centroid of each functional activity area.

[0043] The trajectory restoration module is used to build a trajectory model and restore the user's movement trajectory path based on the crowd trajectory data;

[0044] The node sequence generation module is used to obtain the restored user movement trajectory path using the UUID field as the identifier, and spatially map it to each functional activity area. Based on the basic database of each time period and the comprehensive database of the whole day, the timestamps of each trajectory point in the user movement trajectory path are extracted to generate path node sequence data.

[0045] The spatiotemporal network construction module is used to construct undirected and directed spatiotemporal networks by using the centroid of each functional activity area as the nodes of the spatiotemporal network and combining different time periods.

[0046] Furthermore, the feature is that, in the scenic spot clustering module, the clustering analysis to obtain the optimal number of clusters includes:

[0047] The K-means clustering algorithm is used to randomly select k scenic spots, each serving as the centroid of a cluster. The Euclidean distance from each scenic spot to the centroid of its cluster is iteratively calculated, and each scenic spot is clustered into the cluster with the smallest Euclidean distance. The centroids of each cluster are then updated until the clustering algorithm converges. The Euclidean distance formula is:

[0048]

[0049] Among them, C i Let C be the centroid of the i-th cluster, with coordinates (Ci, C ...). i1 C i2 ), where x is the coordinate vector of the scenic spot, and m represents the coordinate dimension;

[0050] The number of clusters k is determined by the elbow rule and the actual situation of the comprehensive park, which establishes a candidate interval for the number of clusters. The silhouette coefficient method is then used to determine the optimal number of clusters within this candidate interval. The elbow rule is established by calculating the change in the sum of squared errors (SSE) under different numbers of clusters k, using the following formula:

[0051]

[0052] Where k represents the number of clusters, and x represents the number of clusters belonging to C. i The coordinate vectors of the attractions in the cluster; when the rate of decrease of SSE slows down significantly with the increase of k, the value of k corresponding to this inflection point is the optimal number of clusters determined by the elbow rule; the clustering interval is determined based on the optimal number of clusters determined by the elbow rule and the actual situation of the comprehensive park.

[0053] The contour coefficient method calculates the contour coefficient of all attractions within the candidate interval for each cluster size k and averages them to obtain the total contour coefficient for that cluster size. By calculating the total contour coefficient for each cluster size within the clustering interval, the cluster size corresponding to the maximum total contour coefficient is taken as the optimal cluster size. The formula for the contour coefficient is:

[0054]

[0055] Where s(i) represents the contour coefficient, a(i) represents the average distance between attraction i and other attractions in the same cluster, b(i) represents the minimum average distance between attraction i and all attractions in other adjacent clusters, and n represents the number of attractions.

[0056] Furthermore, the feature is that, in the trajectory restoration module, the step of building a trajectory model and restoring the user's movement trajectory path based on the crowd trajectory data includes: constructing a road network base map inside the comprehensive park, covering roads and travel paths; identifying the trajectory points in the cleaned crowd trajectory data using the UUID field, and projecting the user's trajectory points onto the nearest road or travel path on the road network base map using a spatial matching algorithm; based on the projection results, using a shortest path analysis algorithm, calculating the shortest travel path between two adjacent trajectory points in the road network, and stitching together all the user's shortest travel path segments in chronological order to build the user's trajectory model, thereby obtaining the restored user movement trajectory path.

[0057] Furthermore, the feature is that, in the node sequence generation module, the generation of path node sequence data includes: using the UUID field as an identifier, spatially overlaying the restored user movement trajectory path with each functional activity area to identify the functional activity areas traversed by the user movement trajectory path; based on the basic database for each time period and the comprehensive database for the whole day, extracting the timestamp sequence of trajectory points in the user's crowd trajectory data to obtain the order in which the user passes through each functional activity area, and constructing a scenic spot access path with time sequence characteristics, i.e., path node sequence data.

[0058] Furthermore, the feature is that, in the spatiotemporal network construction module, the construction of the undirected spatiotemporal network and the directed spatiotemporal network includes:

[0059] The centroids of the functional activity zones obtained from clustering are defined as nodes in the spatiotemporal network, and a basic node graph is constructed. The complete trajectory paths reconstructed by all users are spatially overlaid with the nodes. The connection frequencies between each functional activity zone are statistically analyzed to construct an undirected node relationship matrix, generating an undirected weighted network. The network centrality is then calculated using the following formula:

[0060]

[0061] Among them, w ij This represents the number of times node i and node j appear consecutively in the user's trajectory, where n represents the number of nodes.

[0062] Based on the undirected weighted network, an undirected spatiotemporal network is constructed by combining different time periods;

[0063] Simultaneously, based on the basic database for each time period, the user trajectory paths are sorted chronologically on the node-based graph, and the visit sequences of the population between various attractions are extracted. Based on large-scale user path data, a Markov chain transition model is established, the transition probability matrix of each node within different time periods is calculated, and a directed transition network, i.e., a directed spatiotemporal network, is constructed. The state transition probability formula of the Markov chain is:

[0064]

[0065] Where, n ij This represents the total number of times a user moves from node i to node j. P represents the total number of transitions from node i to all other nodes. ij This represents the probability of transitioning from node i to node j.

[0066] Beneficial effects: This invention provides a method and system for constructing a spatiotemporal network of urban integrated parks based on crowd trajectory data, which has significant advantages such as high data processing efficiency, high modeling accuracy, and comprehensive privacy protection, specifically reflected in:

[0067] 1) Compared with existing methods that focus on crowd behavior analysis at the scale of open blocks or cities, this invention constructs a customized trajectory modeling process for closed spaces such as urban parks, which can truly reflect the tour paths, relationships between attractions and behavioral evolution within the park.

[0068] 2) By using the shortest path analysis method, the user trajectory is spatially repaired and completed, avoiding problems such as sparse and broken distribution of original trajectory points, and improving the continuity and structured expression of the population trajectory.

[0069] 3) This invention extracts the scenic spots that users pass through inside the park and integrates them with time series to construct a dynamic network structure, thereby achieving a refined expression of crowd behavior at the micro-spatial scale.

[0070] 4) This invention takes trajectory structure modeling as its core, does not rely on personal identity information, improves behavioral interpretability through path structure analysis, and supports spatial optimization design while ensuring data privacy. Attached Figure Description

[0071] Figure 1 This is an overall flowchart of the method of the present invention.

[0072] Figure 2 This is a base map of the activity area nodes inside Xuanwu Lake Park in an embodiment of the present invention.

[0073] Figure 3 This is a flowchart illustrating the shortest path restoration of the real trajectory and the construction of the temporal path in this embodiment of the invention.

[0074] Figure 4 This is a spatiotemporal network diagram of the Xuanwu Lake Park in an embodiment of the present invention.

[0075] Figure 5 This is a diagram of the internal transfer network of Xuanwu Lake Park in an embodiment of the present invention. Detailed Implementation

[0076] The present invention will be further illustrated below with reference to the accompanying drawings and specific embodiments. It should be understood that the following specific embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. After reading the present invention, any modifications of the present invention in various equivalent forms by those skilled in the art will fall within the scope defined by the appended claims.

[0077] Example 1

[0078] like Figure 1 As shown, this invention provides a method for constructing a spatiotemporal network for urban integrated parks based on crowd trajectory data, comprising the following steps:

[0079] (1) Collect crowd trajectory data, the content of which includes anonymous encrypted mobile terminal UUID, timestamp of trajectory point when mobile phone positioning occurs, and latitude and longitude coordinates of trajectory point when mobile phone positioning occurs;

[0080] (2) Clean the crowd trajectory data, retain only the crowd trajectory data within the comprehensive park area, and delete duplicate trajectory data of the same UUID and UUID and its trajectory data that appear only once;

[0081] (3) Divide the cleaned crowd trajectory data into time periods, and merge the crowd trajectory data of each time period according to the timestamp to generate a basic database for each time period and a comprehensive database for the whole day.

[0082] (4) Cluster the attractions in the comprehensive park and determine the optimal number of clusters through cluster analysis. Each cluster is used as a functional activity area. Based on the coordinates of the attractions in each cluster, construct the minimum bounding geometry of each functional activity area and extract the centroid of the minimum bounding geometry as the centroid of each functional activity area.

[0083] (5) Build a trajectory model and reconstruct the user's movement trajectory path based on the crowd trajectory data;

[0084] (6) Use the UUID field as an identifier to obtain the restored user movement trajectory path and spatially map it to each functional activity area. Extract the timestamps of each trajectory point in the user movement trajectory path according to the basic database of each time period and the comprehensive database of the whole day, and generate path node sequence data.

[0085] (7) Using the centroid of each functional activity area as the nodes of the spatiotemporal network, and combining each time period, construct an undirected spatiotemporal network and a directed spatiotemporal network.

[0086] In this embodiment, the crowd trajectory data in step (1) is obtained from mobile phone location data, and the anonymous encrypted trajectory data is obtained. The content of the crowd trajectory data includes the anonymous encrypted mobile phone terminal UUID, the timestamp of the mobile phone location, the latitude and longitude coordinates when the mobile phone location occurred, etc., where the UUID is used to uniquely identify the user; the data space range is Nanjing city area, the time is locked on May 1, 2024, and the daily average data volume is about 160,000 data.

[0087] Step (2) First, based on the boundary vector data of Xuanwu Lake Park, perform spatial filtering on the crowd trajectory data, retaining only the trajectory point data within the Xuanwu Lake Park area to ensure that the analysis focuses on the target area; second, for duplicate records with the same latitude and longitude coordinates, timestamps and UUIDs, retain only one valid data to eliminate redundancy caused by differences in data collection frequency or scenario settings; finally, delete the UUID and its trajectory record that appears only once in the dataset, as such records cannot form a valid trajectory sequence and have no analytical value.

[0088] The cleaned crowd trajectory data meets the following conditions: all trajectory points are located within the Xuanwu Lake Park area; the same user has data records at two or more time points; the data has trajectory change characteristics; the data has continuity and differences; and it can support subsequent path modeling and behavior analysis.

[0089] Step (3) divide the cleaned crowd trajectory data into early morning, morning, afternoon and evening according to typical time periods, and merge the crowd trajectory data of each time period to generate a basic database for each time period and a comprehensive database for the whole day.

[0090] Step (4) Regarding the distribution of attractions within Xuanwu Lake Park, the activity area is first identified and extracted. Data on registered points of interest (POIs) within Xuanwu Lake Park is obtained through the "Planning Cloud" platform. Combined with the results of on-site surveys, and supplemented by the "Two Steps" App's point-marking function, missing landscape facilities such as pavilions, towers, and covered bridges are added. A relatively complete database of attractions within the park is then integrated to construct a comprehensive map of attractions in Xuanwu Lake Park.

[0091] Subsequently, the K-means clustering algorithm was used to perform cluster analysis on the spatial locations of attractions in the comprehensive attraction distribution map. By randomly selecting k attractions as the centroids of clusters, the Euclidean distance from each attraction in the comprehensive park to the centroid of each cluster was iteratively calculated. Each attraction was then clustered into the cluster with the smallest Euclidean distance, and the centroids of each cluster were updated until the clustering algorithm converged. The Euclidean distance formula is:

[0092]

[0093] Among them, C iLet C be the centroid of the i-th cluster, with coordinates (Ci, C ...). i1 C i2 ), where x is the coordinate vector of the scenic spot, and m represents the coordinate dimension;

[0094] The number of clusters, k, is determined by the elbow method and the actual situation of the integrated park, with a candidate interval for the number of clusters. The silhouette coefficient method is then used to determine the optimal number of clusters within the candidate interval. The elbow method calculates and evaluates the change in the sum of squared errors (SSE) under different numbers of clusters, k. When the rate of decrease in SSE slows significantly with increasing k, the k value corresponding to this inflection point is the optimal number of clusters determined by the elbow method, as shown in the formula:

[0095]

[0096] Here, the size of SSE is used to measure the quality of clustering results, k represents the number of clusters, and x represents the number of clusters belonging to C. i The coordinate vectors of the attractions in the clusters; considering that the research object is at the park scale, the candidate range for the number of clusters is 20 to 40.

[0097] The silhouette coefficient method calculates the silhouette coefficient of all attractions within the cluster interval for each cluster size k and averages them to obtain the total silhouette coefficient for that cluster size. By calculating the total silhouette coefficient for each cluster size within the cluster interval, the optimal cluster size is determined by the cluster size corresponding to the maximum total silhouette coefficient. The formula for the silhouette coefficient is:

[0098]

[0099] Where s(i) represents the contour coefficient, a(i) represents the average distance between attraction i and other attractions in the same cluster, b(i) represents the minimum average distance between attraction i and all attractions in other adjacent clusters, and n represents the number of attractions.

[0100] The optimal number of clusters was determined to be k=32, resulting in 32 functional activity zones.

[0101] Finally, based on the spatial distribution of attractions in each functional activity zone, a minimum bounding geometry is constructed, and the centroid of this geometry is extracted as the centroid of the functional activity zone for subsequent trajectory path mapping and spatiotemporal network analysis, such as... Figure 2 As shown.

[0102] Step (5) First, construct a road network base map inside Xuanwu Lake Park, covering the main roads and secondary traffic paths, to provide a spatial reference framework for trajectory modeling.

[0103] Secondly, the cleaned crowd trajectory data is identified by the UUID field. The user's trajectory points are projected onto the nearest road node or segment on the road network base map using the spatial matching algorithm in the ArcGIS platform to eliminate positioning errors and ensure that all trajectories match the actual path conditions.

[0104] Next, based on the matching results, the shortest path analysis algorithm is used to calculate the shortest travel path in the road network between two adjacent trajectory points. For a user trajectory containing n valid trajectory points, a total of n-1 shortest path segments are generated. All the user's shortest travel path segments are spliced ​​together in chronological order to build the user's trajectory model and obtain the restored user movement trajectory path.

[0105] Step (6) First, using the UUID field as an identifier, the restored user movement trajectory path is spatially superimposed with the functional activity areas defined in the Xuanwu Lake Park Comprehensive Scenic Spot Distribution Map to identify the functional activity areas traversed by the user movement trajectory path.

[0106] Next, based on the basic database for each time period and the comprehensive database for the whole day, the timestamp sequence of trajectory points in the user's crowd trajectory data is extracted to obtain the order in which the user passes through each functional activity area, and a scenic spot access path with time series characteristics is constructed. For a user movement trajectory path containing n trajectory points, the scenic spots visited are numbered and integrated according to the time sequence, and finally, a complete path node sequence data with time series structure is generated for the user, such as... Figure 3 As shown.

[0107] Step (7) First, the centroids of the 32 functional activity areas obtained by clustering are defined as nodes of the spatiotemporal network. A node base graph is constructed, with each node corresponding to a specific functional activity area. The lines between nodes represent the movement paths that users may form during the tour.

[0108] Subsequently, the complete trajectories reconstructed by all users are spatially overlaid with the nodes. If a trajectory passes through both node i and node j, a connection is recorded between i and j. The connection frequency between each functional activity area is counted, and an undirected node relationship matrix is ​​constructed to quantify the association strength between nodes. An undirected weighted network is generated, and structural indicators such as network centrality are calculated based on this matrix to reflect the status and importance of each functional activity area in the overall spatial network. The formula for calculating network centrality is as follows:

[0109]

[0110] Among them, w ijThis represents the number of consecutive occurrences of node i and node j in the user's trajectory. This metric measures the connection strength of nodes in the overall network, where n represents the number of nodes. Since this analysis does not incorporate the direction of movement, the resulting network relationship is an undirected weighted network structure. Based on this undirected weighted network, and combining different time periods, an undirected spatiotemporal network is constructed, such as... Figure 4 As shown.

[0111] On the node-based graph, based on the basic database for each time period, the time sequence of user trajectory paths is arranged, and the visit sequences of people between various attractions are extracted to construct directional path links. Based on large-scale user path data, a Markov chain transition model is established to calculate the transition probability matrix of each node in different time periods, so as to characterize the dynamic transition characteristics of people between different functional activity areas, thereby constructing a directed transition network, i.e., a directed spatiotemporal network; the state transition probability formula of the Markov chain is:

[0112]

[0113] Where, n ij This represents the total number of times a user moves from node i to node j. This represents the total number of transitions from node i to all other nodes. Since this matrix has directionality, P... ij This represents the probability of transitioning from node i to node j. The resulting network relationship is a directed transition network that reflects the behavioral migration pattern, such as... Figure 5 As shown.

[0114] The spatiotemporal network can intuitively reflect the actual behavioral paths of users within the integrated park. In addition, since different paths and nodes also have different crowd sizes in the spatiotemporal network, it can be seen from the spatiotemporal network which specific nodes or paths users form clusters on.

[0115] Example 2

[0116] like Figure 1 As shown in Embodiment 1, the present invention also provides a spatiotemporal network construction system for urban integrated parks based on crowd trajectory data, comprising:

[0117] The data collection module is used to collect crowd trajectory data, which includes the anonymous encrypted mobile terminal UUID, the timestamp of the trajectory point when the mobile phone location occurred, and the latitude and longitude coordinates of the trajectory point when the mobile phone location occurred.

[0118] The data cleaning module is used to clean crowd trajectory data, retaining only the crowd trajectory data within the comprehensive park area, and deleting duplicate trajectory data of the same UUID and UUID and its trajectory data that appear only once;

[0119] The data segmentation module is used to divide the cleaned crowd trajectory data into time periods and merge the crowd trajectory data of each time period according to the timestamp to generate the basic database for each time period and the comprehensive database for the whole day.

[0120] The attraction clustering module is used to cluster attractions in the comprehensive park, determine the optimal number of clusters through cluster analysis, and use each cluster as a functional activity area. Based on the coordinates of attractions in each cluster, the minimum bounding geometry of each functional activity area is constructed, and the centroid of the minimum bounding geometry is extracted as the centroid of each functional activity area.

[0121] The trajectory restoration module is used to build a trajectory model and restore the user's movement trajectory path based on the crowd trajectory data;

[0122] The node sequence generation module is used to obtain the restored user movement trajectory path using the UUID field as the identifier, and spatially map it to each functional activity area. Based on the basic database of each time period and the comprehensive database of the whole day, the timestamps of each trajectory point in the user movement trajectory path are extracted to generate path node sequence data.

[0123] The spatiotemporal network construction module is used to construct undirected and directed spatiotemporal networks by using the centroid of each functional activity area as the nodes of the spatiotemporal network and combining different time periods.

[0124] In this embodiment, the data collection module obtains anonymously encrypted trajectory data from mobile phone location data. The content of the crowd trajectory data includes information such as the anonymously encrypted mobile terminal UUID, the timestamp of the mobile phone location, and the latitude and longitude coordinates at the time of the mobile phone location. The UUID is used to uniquely identify the user. The data space range is Nanjing city area, the time is locked to May 1, 2024, and the average daily data volume is about 160,000 data entries.

[0125] The data cleaning module first performs spatial filtering on the crowd trajectory data based on the boundary vector data of Xuanwu Lake Park, retaining only trajectory point data within the park area to ensure that the analysis focuses on the target area. Second, for duplicate records with the same latitude and longitude coordinates, timestamps, and UUIDs, only one valid record is retained to eliminate redundancy caused by differences in data collection frequency or context settings. Finally, UUIDs and their trajectory records that appear only once in the dataset are deleted, as these records cannot form a valid trajectory sequence and have no analytical value.

[0126] The cleaned crowd trajectory data meets the following conditions: all trajectory points are located within the Xuanwu Lake Park area; the same user has data records at two or more time points; the data has trajectory change characteristics; the data has continuity and differences; and it can support subsequent path modeling and behavior analysis.

[0127] The data segmentation module divides the cleaned crowd trajectory data into typical time periods: early morning, morning, afternoon, and evening. It then merges the crowd trajectory data from each time period to generate a basic database for each time period and a comprehensive database for the entire day.

[0128] The scenic spot clustering module targets the distribution of scenic spots within Xuanwu Lake Park. First, it identifies and extracts activity areas, obtains registered points of interest (POI) data within the Xuanwu Lake area through the "Planning Cloud" platform, combines the results of field surveys, and supplements the missing pavilions, bridges, and other landscape facilities with the "Two Steps" App's point-marking function. This integrates to form a relatively complete database of scenic spot information within the park, and constructs a comprehensive scenic spot distribution map of Xuanwu Lake Park.

[0129] Subsequently, the K-means clustering algorithm was used to perform cluster analysis on the spatial locations of attractions in the comprehensive attraction distribution map. By randomly selecting k attractions as the centroids of clusters, the Euclidean distance from each attraction in the comprehensive park to the centroid of each cluster was iteratively calculated. Each attraction was then clustered into the cluster with the smallest Euclidean distance, and the centroids of each cluster were updated until the clustering algorithm converged. The Euclidean distance formula is:

[0130]

[0131] Among them, C i Let C be the centroid of the i-th cluster, with coordinates (Ci, C ...). i1 C i2 ), where x is the coordinate vector of the scenic spot, and m represents the coordinate dimension;

[0132] The number of clusters, k, is determined by the elbow method and the actual situation of the integrated park, with a candidate interval for the number of clusters. The silhouette coefficient method is then used to determine the optimal number of clusters within the candidate interval. The elbow method calculates and evaluates the change in the sum of squared errors (SSE) under different numbers of clusters, k. When the rate of decrease in SSE slows significantly with increasing k, the k value corresponding to this inflection point is the optimal number of clusters determined by the elbow method, as shown in the formula:

[0133]

[0134] Here, the size of SSE is used to measure the quality of clustering results, k represents the number of clusters, and x represents the number of clusters belonging to C. i The coordinate vectors of the attractions in the clusters; considering that the research object is at the park scale, the candidate range for the number of clusters is 20 to 40.

[0135] The silhouette coefficient method calculates the silhouette coefficient of all attractions within the cluster interval for each cluster size k and averages them to obtain the total silhouette coefficient for that cluster size. By calculating the total silhouette coefficient for each cluster size within the cluster interval, the optimal cluster size is determined by the cluster size corresponding to the maximum total silhouette coefficient. The formula for the silhouette coefficient is:

[0136]

[0137] Where s(i) represents the contour coefficient, a(i) represents the average distance between attraction i and other attractions in the same cluster, b(i) represents the minimum average distance between attraction i and all attractions in other adjacent clusters, and n represents the number of attractions.

[0138] The optimal number of clusters was determined to be k=32, resulting in 32 functional activity zones.

[0139] Finally, based on the spatial distribution of attractions in each functional activity zone, a minimum bounding geometry is constructed, and the centroid of this geometry is extracted as the centroid of the functional activity zone for subsequent trajectory path mapping and spatiotemporal network analysis, such as... Figure 2 As shown.

[0140] The trajectory restoration module first constructs a road network base map inside Xuanwu Lake Park, covering major roads and secondary traffic paths, to provide a spatial reference framework for trajectory modeling.

[0141] Secondly, the cleaned crowd trajectory data is identified by the UUID field. The user's trajectory points are projected onto the nearest road node or segment on the road network base map using the spatial matching algorithm in the ArcGIS platform to eliminate positioning errors and ensure that all trajectories match the actual path conditions.

[0142] Next, based on the matching results, the shortest path analysis algorithm is used to calculate the shortest travel path in the road network between two adjacent trajectory points. For a user trajectory containing n valid trajectory points, a total of n-1 shortest path segments are generated. All the user's shortest travel path segments are spliced ​​together in chronological order to build the user's trajectory model and obtain the restored user movement trajectory path.

[0143] The node sequence generation module first uses the UUID field as an identifier to spatially overlay the restored user movement trajectory path with the functional activity areas defined in the Xuanwu Lake Park comprehensive scenic spot distribution map, and identifies the functional activity areas traversed by the user movement trajectory path.

[0144] Next, based on the basic database for each time period and the comprehensive database for the whole day, the timestamp sequence of trajectory points in the user's crowd trajectory data is extracted to obtain the order in which the user passes through each functional activity area, and a scenic spot access path with time series characteristics is constructed. For a user movement trajectory path containing n trajectory points, the scenic spots visited are numbered and integrated according to the time sequence, and finally, a complete path node sequence data with time series structure is generated for the user, such as... Figure 3 As shown.

[0145] The spatiotemporal network construction module first defines the centroids of the 32 functional activity areas obtained by clustering as nodes of the spatiotemporal network, constructs a node base graph, with each node corresponding to a specific functional activity area, and the lines between nodes representing the movement paths that users may form during the tour.

[0146] Subsequently, the complete trajectories reconstructed by all users are spatially overlaid with the nodes. If a trajectory passes through both node i and node j, a connection is recorded between i and j. The connection frequency between each functional activity area is counted, and an undirected node relationship matrix is ​​constructed to quantify the association strength between nodes. An undirected weighted network is generated, and structural indicators such as network centrality are calculated based on this matrix to reflect the status and importance of each functional activity area in the overall spatial network. The formula for calculating network centrality is as follows:

[0147]

[0148] Among them, w ij This represents the number of consecutive occurrences of node i and node j in the user's trajectory. This metric measures the connection strength of nodes in the overall network, where n represents the number of nodes. Since this analysis does not incorporate the direction of movement, the resulting network relationship is an undirected weighted network structure. Based on this undirected weighted network, and combining different time periods, an undirected spatiotemporal network is constructed, such as... Figure 4 As shown.

[0149] On the node-based graph, based on the basic database for each time period, the time sequence of user trajectory paths is arranged, and the visit sequences of people between various attractions are extracted to construct directional path links. Based on large-scale user path data, a Markov chain transition model is established to calculate the transition probability matrix of each node in different time periods, so as to characterize the dynamic transition characteristics of people between different functional activity areas, thereby constructing a directed transition network, i.e., a directed spatiotemporal network; the state transition probability formula of the Markov chain is:

[0150]

[0151] Where, n ij This represents the total number of times a user moves from node i to node j. This represents the total number of transitions from node i to all other nodes. Since this matrix has directionality, P... ij This represents the probability of transitioning from node i to node j. The resulting network relationship is a directed transition network that reflects the behavioral migration pattern, such as... Figure 5 As shown.

[0152] The spatiotemporal network can intuitively reflect the actual behavioral paths of users within the integrated park. In addition, since different paths and nodes also have different crowd sizes in the spatiotemporal network, it can be seen from the spatiotemporal network which specific nodes or paths users form clusters on.

Claims

1. A method for constructing a spatiotemporal network for urban integrated parks based on crowd trajectory data, characterized in that, include: (1) Collect crowd trajectory data, the content of which includes anonymous encrypted mobile terminal UUID, timestamp of trajectory point when mobile phone positioning occurs, and latitude and longitude coordinates of trajectory point when mobile phone positioning occurs; (2) Clean the crowd trajectory data, retain only the crowd trajectory data within the comprehensive park area, and delete duplicate trajectory data of the same UUID and UUID and its trajectory data that appear only once; (3) Divide the cleaned crowd trajectory data into time periods, and merge the crowd trajectory data of each time period according to the timestamp to generate a basic database for each time period and a comprehensive database for the whole day. (4) Cluster the attractions in the comprehensive park and determine the optimal number of clusters through cluster analysis. Each cluster is used as a functional activity area. Based on the coordinates of the attractions in each cluster, construct the minimum bounding geometry of each functional activity area and extract the centroid of the minimum bounding geometry as the centroid of each functional activity area. (5) Build a trajectory model and reconstruct the user's movement trajectory path based on the crowd trajectory data; (6) Use the UUID field as an identifier to obtain the restored user movement trajectory path and spatially map it to each functional activity area. Extract the timestamps of each trajectory point in the user movement trajectory path according to the basic database of each time period and the comprehensive database of the whole day, and generate path node sequence data. (7) Using the centroid of each functional activity area as the nodes of the spatiotemporal network, and combining each time period, construct an undirected spatiotemporal network and a directed spatiotemporal network.

2. The method for constructing a spatiotemporal network of an urban integrated park based on crowd trajectory data according to claim 1, characterized in that, The cluster analysis obtains the optimal number of clusters, including: The K-means clustering algorithm is used to randomly select k scenic spots, each serving as the centroid of a cluster. The Euclidean distance from each scenic spot to the centroid of its cluster is iteratively calculated, and each scenic spot is clustered into the cluster with the smallest Euclidean distance. The centroids of each cluster are then updated until the clustering algorithm converges. The Euclidean distance formula is: Among them, C i Let C be the centroid of the i-th cluster, with coordinates (Ci, C ...). i1 C i2 ), where x is the coordinate vector of the scenic spot, and m represents the coordinate dimension; The number of clusters k is determined by the elbow rule and the actual situation of the comprehensive park, which establishes a candidate interval for the number of clusters. The silhouette coefficient method is then used to determine the optimal number of clusters within this candidate interval. The elbow rule is established by calculating the change in the sum of squared errors (SSE) under different numbers of clusters k, using the following formula: Where k represents the number of clusters, and x represents the number of clusters belonging to C. i The coordinate vectors of the attractions in the cluster; when the rate of decrease of SSE slows down significantly with the increase of k, the value of k corresponding to this inflection point is the optimal number of clusters determined by the elbow rule; the clustering interval is determined based on the optimal number of clusters determined by the elbow rule and the actual situation of the comprehensive park. The contour coefficient method calculates the contour coefficient of all attractions within the candidate interval for each cluster size k and averages them to obtain the total contour coefficient for that cluster size. By calculating the total contour coefficient for each cluster size within the clustering interval, the cluster size corresponding to the maximum total contour coefficient is taken as the optimal cluster size. The formula for the contour coefficient is: Where s(i) represents the contour coefficient, a(i) represents the average distance between attraction i and other attractions in the same cluster, b(i) represents the minimum average distance between attraction i and all attractions in other adjacent clusters, and n represents the number of attractions.

3. The method for constructing a spatiotemporal network of an urban integrated park based on crowd trajectory data according to claim 1, characterized in that, The process of building a trajectory model and reconstructing the user's movement trajectory path based on crowd trajectory data includes: constructing a road network base map within the comprehensive park, covering roads and travel paths; identifying trajectory points in the cleaned crowd trajectory data using the UUID field, and projecting the user's trajectory points onto the nearest road or travel path on the road network base map using a spatial matching algorithm; based on the projection results, using a shortest path analysis algorithm, calculating the shortest travel path between two adjacent trajectory points in the road network, and stitching together all the user's shortest path segments in chronological order to build the user's trajectory model, thereby obtaining the reconstructed user movement trajectory path.

4. The method for constructing a spatiotemporal network of an urban integrated park based on crowd trajectory data according to claim 3, characterized in that, The generated path node sequence data includes: using the UUID field as an identifier, spatially overlaying the restored user movement trajectory path with each functional activity area to identify the functional activity areas traversed by the user movement trajectory path; based on the basic database for each time period and the comprehensive database for the whole day, extracting the timestamp sequence of trajectory points in the user's crowd trajectory data to obtain the order in which the user passes through each functional activity area, and constructing a scenic spot access path with time series characteristics, i.e., path node sequence data.

5. The method for constructing a spatiotemporal network of an urban integrated park based on crowd trajectory data according to claim 1, characterized in that, The construction of undirected and directed spatiotemporal networks includes: The centroids of the functional activity zones obtained from clustering are defined as nodes in the spatiotemporal network, and a basic node graph is constructed. The complete trajectory paths reconstructed by all users are spatially overlaid with the nodes. The connection frequencies between each functional activity zone are statistically analyzed to construct an undirected node relationship matrix, generating an undirected weighted network. The network centrality is then calculated using the following formula: Among them, w ij This represents the number of times node i and node j appear consecutively in the user's trajectory, where n represents the number of nodes. Based on the undirected weighted network, an undirected spatiotemporal network is constructed by combining different time periods; Simultaneously, based on the basic database for each time period, the user trajectory paths are sorted chronologically on the node-based graph, and the visit sequences of the population between various attractions are extracted. Based on large-scale user path data, a Markov chain transition model is established, the transition probability matrix of each node within different time periods is calculated, and a directed transition network, i.e., a directed spatiotemporal network, is constructed. The state transition probability formula of the Markov chain is: Where, n ij This represents the total number of times a user moves from node i to node j. P represents the total number of transitions from node i to all other nodes. ij This represents the probability of transitioning from node i to node j.

6. A spatiotemporal network construction system for urban integrated parks based on crowd trajectory data, characterized in that, include: The data collection module is used to collect crowd trajectory data, which includes the anonymous encrypted mobile terminal UUID, the timestamp of the trajectory point when the mobile phone location occurred, and the latitude and longitude coordinates of the trajectory point when the mobile phone location occurred. The data cleaning module is used to clean crowd trajectory data, retaining only the crowd trajectory data within the comprehensive park area, and deleting duplicate trajectory data of the same UUID and UUID and its trajectory data that appear only once; The data segmentation module is used to divide the cleaned crowd trajectory data into time periods and merge the crowd trajectory data of each time period according to the timestamp to generate the basic database for each time period and the comprehensive database for the whole day. The attraction clustering module is used to cluster attractions in the comprehensive park, determine the optimal number of clusters through cluster analysis, and use each cluster as a functional activity area. Based on the coordinates of attractions in each cluster, the minimum bounding geometry of each functional activity area is constructed, and the centroid of the minimum bounding geometry is extracted as the centroid of each functional activity area. The trajectory restoration module is used to build a trajectory model and restore the user's movement trajectory path based on the crowd trajectory data; The node sequence generation module is used to obtain the restored user movement trajectory path using the UUID field as the identifier, and spatially map it to each functional activity area. Based on the basic database of each time period and the comprehensive database of the whole day, the timestamps of each trajectory point in the user movement trajectory path are extracted to generate path node sequence data. The spatiotemporal network construction module is used to construct undirected and directed spatiotemporal networks by using the centroid of each functional activity area as the nodes of the spatiotemporal network and combining different time periods.

7. The urban integrated park spatiotemporal network construction system based on crowd trajectory data according to claim 6, characterized in that, In the scenic spot clustering module, the clustering analysis to obtain the optimal number of clusters includes: The K-means clustering algorithm is used to randomly select k scenic spots, each serving as the centroid of a cluster. The Euclidean distance from each scenic spot to the centroid of its cluster is iteratively calculated, and each scenic spot is clustered into the cluster with the smallest Euclidean distance. The centroids of each cluster are then updated until the clustering algorithm converges. The Euclidean distance formula is: Among them, C i Let C be the centroid of the i-th cluster, with coordinates (Ci, C ...). i1 C i2 ), where x is the coordinate vector of the scenic spot, and m represents the coordinate dimension; The number of clusters k is determined by the elbow rule and the actual situation of the comprehensive park, which establishes a candidate interval for the number of clusters. The silhouette coefficient method is then used to determine the optimal number of clusters within this candidate interval. The elbow rule is established by calculating the change in the sum of squared errors (SSE) under different numbers of clusters k, using the following formula: Where k represents the number of clusters, and x represents the number of clusters belonging to C. i The coordinate vectors of the attractions in the cluster; when the rate of decrease of SSE slows down significantly with the increase of k, the value of k corresponding to this inflection point is the optimal number of clusters determined by the elbow rule; the clustering interval is determined based on the optimal number of clusters determined by the elbow rule and the actual situation of the comprehensive park. The contour coefficient method calculates the contour coefficient of all attractions within the candidate interval for each cluster size k and averages them to obtain the total contour coefficient for that cluster size. By calculating the total contour coefficient for each cluster size within the clustering interval, the cluster size corresponding to the maximum total contour coefficient is taken as the optimal cluster size. The formula for the contour coefficient is: Where s(i) represents the contour coefficient, a(i) represents the average distance between attraction i and other attractions in the same cluster, b(i) represents the minimum average distance between attraction i and all attractions in other adjacent clusters, and n represents the number of attractions.

8. The urban integrated park spatiotemporal network construction system based on crowd trajectory data according to claim 6, characterized in that, In the trajectory restoration module, the process of building a trajectory model and restoring the user's movement trajectory path based on crowd trajectory data includes: constructing a road network base map within the comprehensive park, covering roads and travel paths; identifying trajectory points in the cleaned crowd trajectory data using the UUID field, and projecting the user's trajectory points onto the nearest road or travel path on the road network base map using a spatial matching algorithm; based on the projection results, using a shortest path analysis algorithm, calculating the shortest travel path between two adjacent trajectory points in the road network, and stitching together all of the user's shortest travel path segments in chronological order to build the user's trajectory model and obtain the restored user movement trajectory path.

9. The urban integrated park spatiotemporal network construction system based on crowd trajectory data according to claim 8, characterized in that, In the node sequence generation module, the generation of path node sequence data includes: using the UUID field as an identifier, spatially overlaying the restored user movement trajectory path with each functional activity area to identify the functional activity areas traversed by the user movement trajectory path; based on the basic database for each time period and the comprehensive database for the whole day, extracting the timestamp sequence of trajectory points in the user's crowd trajectory data to obtain the order in which the user passes through each functional activity area, and constructing a scenic spot access path with time series characteristics, i.e., path node sequence data.

10. The urban integrated park spatiotemporal network construction system based on crowd trajectory data according to claim 6, characterized in that, In the spatiotemporal network construction module, the construction of undirected and directed spatiotemporal networks includes: The centroids of the functional activity zones obtained from clustering are defined as nodes in the spatiotemporal network, and a basic node graph is constructed. The complete trajectory paths reconstructed by all users are spatially overlaid with the nodes. The connection frequencies between each functional activity zone are statistically analyzed to construct an undirected node relationship matrix, generating an undirected weighted network. The network centrality is then calculated using the following formula: Among them, w ij This represents the number of times node i and node j appear consecutively in the user's trajectory, where n represents the number of nodes. Based on the undirected weighted network, an undirected spatiotemporal network is constructed by combining different time periods; Simultaneously, based on the basic database for each time period, the user trajectory paths are sorted chronologically on the node-based graph, and the visit sequences of the population between various attractions are extracted. Based on large-scale user path data, a Markov chain transition model is established, the transition probability matrix of each node within different time periods is calculated, and a directed transition network, i.e., a directed spatiotemporal network, is constructed. The state transition probability formula of the Markov chain is: Where, n ij This represents the total number of times a user moves from node i to node j. P represents the total number of transitions from node i to all other nodes. ij This represents the probability of transitioning from node i to node j.

Images