A building-level public transport accessibility evaluation method based on heterogeneous graph

Abstract

This invention discloses a building-level public transportation convenience assessment method based on heterogeneous graphs, comprising: acquiring building data, street block data, public transportation station and route data, gated residential area boundary data, water system data, and population distribution data of the assessment area; preprocessing the acquired data; constructing a heterogeneous graph model: obtaining building nodes and street nodes based on building data and street block data; assigning weights to building nodes and street nodes, wherein the weights of building nodes are determined based on the population distribution data, and the weights of street nodes are determined based on the associated comprehensive public transportation service capacity; constructing edges connecting each node according to the spatial topological relationships between building nodes, between street nodes, and between building nodes and street nodes; defining walking costs for the constructed edges; and calculating the public transportation convenience index based on the constructed heterogeneous graph model. The aim is to improve the accuracy of urban public transportation convenience assessment and enhance residents' actual travel experience.

Description

Technical Field

[0001] This invention relates to a method for assessing the convenience of public transportation, and more particularly to a building-level method for assessing the convenience of public transportation based on heterogeneous graphs, belonging to the field of urban public transportation planning and assessment technology. Background Technology

[0002] Public transportation accessibility assessment is a core basis for urban transportation planning, diagnosis, and upgrading. Existing assessment methods mainly include macro-statistical methods based on administrative units or grids, micro-scale methods based on potential models, accessibility assessment methods based on cumulative opportunities, and micro-analysis methods based on refined route simulation.

[0003] However, the following shortcomings still exist in practical applications: First, the evaluation unit is still the plot of land, and the differences in walking access costs caused by the location differences of different buildings within the same plot are ignored, failing to meet the needs of refined building-level diagnosis. Second, walking access time is only considered as part of the total travel cost, failing to achieve the native coupling between the walking micro-network and the macro-public transport network, making it difficult to independently examine the impact of the walking environment. Third, the use of a summation operator to overlay the service potential of multiple stations does not match the actual behavior of residents choosing the optimal station, easily overestimating convenience. Fourth, the calculation of walking costs based on an idealized road network does not consider actual physical barriers such as waterways and enclosed community walls, resulting in a mismatch between the spatial coverage and the actual walking reach. Summary of the Invention

[0004] Purpose of the invention: The purpose of this invention is to solve the problems existing in the prior art and provide a building-level public transportation convenience assessment method based on heterogeneous graphs, thereby improving the accuracy of urban public transportation convenience assessment and enhancing residents' actual travel experience.

[0005] Technical solution: The present invention provides a method for assessing building-level public transportation convenience based on heterogeneous graphs, comprising the following steps:

[0006] S1. Obtain building data, street block data, public transportation station and route data, gated residential area boundary data, water system data, and population distribution data for the assessment area;

[0007] S2. Preprocess the acquired data;

[0008] S3. Construct a heterogeneous graph model, including:

[0009] S31. Based on building data and street block data, obtain building nodes and street nodes;

[0010] S32. Assign weights to building nodes and street nodes, wherein the weights of building nodes are determined based on the population distribution data, and the weights of street nodes are determined based on the integrated public transportation service capacity associated with them.

[0011] S33. Based on the spatial topological relationships between building nodes, street nodes, and between building nodes and street nodes, construct the edges connecting each node;

[0012] S34. Define the walking cost for the constructed edges;

[0013] S4. Calculate the public transportation convenience index based on the constructed heterogeneous graph model.

[0014] Furthermore, in step S1, the building data includes building outline features, geographical location information, and number of building floors; the street block data includes street centerline, road grade, and planned width; the public transportation station and line data includes the names, geographical location information, service line numbers, and departure frequency of bus stops and subway stations; the gated residential area boundary data and water system data are used to identify physical barriers to pedestrian paths; and the population distribution data is raster-formatted population density data.

[0015] Furthermore, in step S2, the preprocessing includes:

[0016] The acquired data is checked, corrected, missing data is added, and duplicates are removed.

[0017] Street centerline data is generated based on street block data, and each street and block is assigned a unique number; buildings located in gated residential areas are identified through spatial overlay analysis, and residential area numbers are added;

[0018] Calculate the building's total floor area, the length and width of the street's centerline.

[0019] Furthermore, in step S32, the weight of the street node is determined based on its associated comprehensive public transportation service capacity, wherein the comprehensive public transportation service capacity is calculated by aggregating multiple public transportation service capacity indicators; the public transportation service capacity indicators include at least one of subway service capacity indicators and bus service capacity indicators; the subway service capacity indicators include indicators reflecting station service level and network accessibility, and the bus service capacity indicators include indicators reflecting bus service intensity and network connectivity; a multi-indicator comprehensive evaluation method is used to aggregate the public transportation service capacity indicators.

[0020] Furthermore, step S33 specifically includes:

[0021] S331. Perform Voronoi tessellation on each building and establish edges between building nodes for the building nodes corresponding to adjacent Voronoi polygons.

[0022] S332. Based on the road topology connection relationship, establish the edges between street nodes corresponding to intersecting street segments;

[0023] S333. Based on the adjacency relationship between the building Voronoi polygon and the street centerline, establish the edges between the building node and the street node.

[0024] Further, in step S34, defining the walking cost for the constructed edge specifically involves:

[0025] The edge cost between street nodes is the Euclidean distance between the building centroids;

[0026] The edge cost between street nodes is calculated using the length and width of the street segment.

[0027] The edge cost between building nodes and street nodes is the shortest Euclidean distance from the building boundary to the street centerline.

[0028] Further, in step S4, the calculation of the public transportation convenience index based on the constructed heterogeneous graph model specifically involves:

[0029] S41. Use Dijkstra's algorithm to calculate the minimum cost for each building node to reach each street node at the boundary of its block.

[0030] S42. Use Dijkstra's algorithm to calculate the minimum cost between any two street nodes in the evaluation area;

[0031] S43. Calculate the minimum total walking cost from each building node to any public transport service street node; set a maximum walking distance threshold to filter the public transport service streets accessible to each building;

[0032] S44. Using an improved potential model, the traditional summation operator is replaced with a maximization operator to calculate the public transportation convenience index for each building node, as shown in the following formula:

[0033] in, As street node weights, The minimum total walking cost to reach any public transport service street node from each building node. The maximum walking distance threshold, The threshold for comfortable walking distance;

[0034] S45. The public transport accessibility index for each block is calculated using a population-weighted average method.

[0035] Furthermore, the building nodes are based on the centroid of the building, and the street nodes are based on the centroid of the street.

[0036] The computer device of the present invention includes one or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the programs, when executed by the processors, implement the steps of the building-level public transportation convenience assessment method based on heterogeneous graphs.

[0037] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of a heterogeneous graph-based building-level public transport convenience assessment method as described above.

[0038] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages:

[0039] (1) This invention extracts the centroid of building elements directly as building nodes, with a single building as the smallest evaluation unit. Compared with the existing evaluation methods that use the centroid of a plot, grid, or administrative district as units, this invention can reveal the micro-differences in convenience between different buildings in the same block due to location differences, providing direct technical support for refined diagnosis and precise policy implementation in urban renewal.

[0040] (2) By constructing a heterogeneous graph model containing building nodes and street nodes, and defining the building-street edge cost as the shortest Euclidean distance from the building boundary, this invention unifies walking cost and the service capacity of public transportation network within the same calculation framework, which can more realistically restore the residents' travel experience and avoid the evaluation bias caused by the separate analysis of the two in traditional methods.

[0041] (3) This invention improves the summation operator in the traditional potential model into a maximization operator, which simulates the real decision-making behavior of residents rationally choosing the optimal site among multiple alternative sites. It avoids the overestimation problem caused by the traditional summation model simply superimposing the service capacity of multiple ordinary sites, making the evaluation results closer to reality in terms of behavioral logic.

[0042] (4) When constructing the edges of the heterogeneous graph, this invention introduces water system surface elements and boundary data of enclosed residential areas for spatial constraint correction. By deleting edges that cross insurmountable obstacles such as water systems and retaining only the connections corresponding to a limited number of entrances and exits for enclosed residential areas, it can accurately identify the phenomenon of "spatial misalignment between spatial coverage and actual walkable range" caused by physical barriers, significantly improving the simulation accuracy of walking costs and enabling the evaluation results to truly reflect the convenience shortcomings of areas blocked by rivers and walls. Attached Figure Description

[0043] Figure 1 This is a flowchart of the evaluation method of the present invention.

[0044] Figure 2 This is a schematic diagram of the intermediate results of each step in the public transportation convenience assessment method according to an embodiment of the present invention.

[0045] Figure 3 This is a graph showing the distribution results of the building-level public transportation convenience index. Detailed Implementation

[0046] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0047] like Figure 1 As shown, a building-level public transport convenience assessment method based on heterogeneous graphs includes the following steps:

[0048] S1. Obtain building data, street block data, public transportation station and route data, gated residential area boundary data, water system data, and population distribution data for the assessment area;

[0049] The building data includes building outline features, geographic location information, and number of building floors; street block data includes street centerline, road grade, and planned width; public transportation station and line data includes bus and subway station names, geographic location information, service line numbers, and line departure frequencies; gated residential area boundary data and water system data are used to identify physical barriers to pedestrian paths; and population distribution data is raster-formatted population density data.

[0050] S2. Preprocess the acquired data;

[0051] Specifically, all acquired data undergoes topological checks, error correction, missing data addition, and duplicate removal; street centerline data is generated based on street block data, and each street is assigned a unique number; blocks are divided based on street centerline data, and each block is assigned a unique number; each building is assigned to its block through spatial overlay analysis; buildings located within enclosed residential areas are identified through spatial overlay analysis, and residential area numbers are added; the total building area, street centerline length, and width are calculated.

[0052] S3. Construct a heterogeneous graph model, specifically:

[0053] Constructing heterogeneous graph models ,in For building node set, For street node set, For the edge set between buildings, For buildings - streetside collection, For street-side markets;

[0054] S31. Based on building data and street block data, obtain building nodes and street nodes;

[0055] Among them, the centroid of the building is extracted as the building node. ∈ Extract the centroid of the street as the street node. ∈ ;

[0056] S32. Assign weights to building nodes and street nodes respectively;

[0057] S321, Building Node Weights Based on population raster data, the traffic demand intensity of a node is estimated by allocating it to buildings.

[0058] S322, Street Node Weights This indicates the overall public transportation service capacity of the street location;

[0059] The comprehensive public transportation service capacity is calculated by aggregating multiple public transportation service capacity indicators. These indicators include at least one of subway service capacity indicators and bus service capacity indicators. The subway service capacity indicators include at least one of the following: indicators reflecting station service levels and network accessibility, such as departure frequency, network centrality, number of reachable stations, and number of entrances and exits. The bus service capacity indicators include at least one of the following: indicators reflecting bus service intensity and network connectivity, such as number of routes, directional diversity, and ease of connection with the subway. A multi-indicator comprehensive evaluation method is used to aggregate the public transportation service capacity indicators, such as at least one of the following: entropy weight method, analytic hierarchy process, factor analysis method, or equal weight method.

[0060] S33. Based on the spatial topological relationships between building nodes, street nodes, and between building nodes and street nodes, construct the edges connecting each node;

[0061] S331. Perform Voronoi tessellation on the buildings within each block, and establish edges between building nodes for the building nodes corresponding to adjacent Voronoi polygons. ;

[0062] S332. Based on the road topology connection relationship, establish edges between street nodes corresponding to intersecting street segments. ;

[0063] S333. Based on the adjacency relationship between building Voronoi polygons and street centerlines, establish edges between building nodes and street nodes. ;

[0064] S334. Based on the water system and the boundary data of the closed residential area, the edges that cross insurmountable obstacles are deleted and corrected, and only the connections corresponding to the limited entrances and exits of the closed residential area are retained.

[0065] S34. Define the walking cost for the constructed edges;

[0066] S341, Edge Costs Between Building Nodes The Euclidean distance between the centroids of the buildings;

[0067] S342, Edge Costs Between Street Nodes Calculated based on street segment length and width; the calculation formula is as follows:

[0068]

[0069] in, and They represent the lengths of street segments m and n, respectively. and This refers to the width of the street corresponding to the street.

[0070] S343, Edge Costs Between Building Nodes and Street Nodes It is the shortest Euclidean distance from the building boundary to the center line of the street.

[0071] S4. Calculate the public transportation convenience index based on the constructed heterogeneous graph model.

[0072] S41. Use Dijkstra's algorithm to calculate the minimum cost for each building node to reach all street nodes at the boundary of its block. ;

[0073] S42. Use Dijkstra's algorithm to calculate the minimum cost between any two street nodes in the regional street network. ;

[0074] S43. Calculate each building node. Arrive at any public transport service street node Minimum total walking cost Set the maximum walking distance threshold. Filter the streets accessible by public transport for each building;

[0075] S44. Using an improved potential model, the traditional summation operator is replaced with a maximization operator to calculate the public transportation convenience index for each building node. The formula is as follows:

[0076] in, As street node weights, The minimum total walking cost to reach any public transport service street node from each building node. The maximum walking distance threshold, The threshold for comfortable walking distance;

[0077] S45. Calculate the public transport accessibility index for each block using a population-weighted average method. .

[0078] The following describes the building-level public transportation convenience assessment method based on heterogeneous graphs in this embodiment, using specific application scenarios:

[0079] (1) Data acquisition

[0080] This embodiment acquires data for a certain area of ​​a certain city. All vector data is stored in Shapefile format (.shp), and raster data is stored in GeoTIFF format (.tif).

[0081] The data obtained includes:

[0082] (11) Building data

[0083] Data is derived from the 2020 3D-GloBFP global building height dataset. Attribute fields include: building ID (unique identifier) ​​and number of floors. The coordinate system uses the WGS-84 geographic coordinate system or a projected coordinate system (such as UTM 50N).

[0084] (12) Street block data

[0085] The data is derived from the 2022 City Block Division Vector Surface Dataset (MSDCW). A field "Block_ID" is added to the attribute table, and a unique number (incrementing from 1) is assigned using the field calculator.

[0086] (13) Public transportation data

[0087] The subway data is sourced from a city's subway map in 2025, including 13 subway lines, 223 subway stations, and the locations of each station's entrances and exits. The bus data is sourced from a 2024 vector dataset of bus stops and routes, containing 1561 bus routes and 1424 bus stops. Station data is stored as point features, with attribute fields including: station name, station type (subway / bus), and service line number (stored as a string). Subway entrance / exit data is stored separately as point features, with attribute fields including the station ID.

[0088] (14) Data on the boundaries and water systems of enclosed residential areas

[0089] The boundaries of gated residential areas are derived from data from Amap in 2024, containing 1,440 boundary features, each assigned a unique number (Gated_ID). The water system data is from OpenStreetMap, including isometric water bodies such as rivers and lakes.

[0090] (15) Population distribution data

[0091] The data is derived from the 2020 WorldPop dataset, consisting of raster data with a spatial resolution of 100 meters × 100 meters. The value of each raster is the population within that grid.

[0092] (2) Data processing

[0093] Spatial data processing is performed using ArcMap 10.8, and some batch operations can be completed with the help of Python 3.8 and libraries such as geopandas and momomepy.

[0094] (21) Data cleaning

[0095] Load all vector data in ArcMap, use the "Repair Geometry" tool to correct topological errors, and use the "Remove Duplicates" tool to remove duplicate features. For missing features, supplement them based on adjacent features or existing data. Check the coordinate system consistency of all vector data, and if necessary, use the "Project" tool to uniformly convert them to the same projected coordinate system (e.g., WGS_1984_UTM_Zone_50N).

[0096] (22) Generate street centerline data based on street block data

[0097] The streetscape data was mosaicked using Python's momomepy library to obtain Voronoi cells. Then, ArcMap's "Feature to Line" tool was used to convert the Voronoi cells into a series of line segments. Next, Python was used to identify line segments shorter than 15 meters, create their midpoints, and move the endpoints of other line segments connected to these midpoints to them. Finally, all identified short line segments were deleted to obtain the street centerline data. A field "Street_ID" was added to the attribute table of the street data, and a field calculator was used to assign this field a unique, incrementally increasing number starting from 1.

[0098] (23) Constructing “street blocks”

[0099] Using ArcMap's "Feature to Polygon" tool, with street centerline data as input, polygon features with streets as boundaries are generated, which are called "street blocks". Figure 2 As shown in b. Using ArcMap's "Spatial Join" tool, use "Street Blocks" as the target layer and the block polygons as the join layer, selecting "JOIN_ONE_TO_ONE" as the join operation. Add a field "Block_ID" to the "Street Blocks" features.

[0100] (24) Constructing the connection between buildings and "street blocks"

[0101] Use ArcMap's Spatial Connections tool to add a "Block_ID" field to the building feature to record the block number to which it belongs.

[0102] (25) Identify buildings within gated residential areas

[0103] Using ArcMap's "Spatial Connections" tool, add a "Gated_ID" field to the building feature and assign it a number corresponding to the residential area. If the number is 0, it indicates that the building is not located within a gated residential area.

[0104] (26) Preprocessing of building and street centerline data

[0105] For building data: Open the attribute table of the building features in ArcMap and add a floating-point field "TA" to store the total building area. The total building area is calculated as the product of the building's footprint area and the number of floors, i.e., TA = Building Footprint Area (Shape_Area) × Number of Floors. Figure 2 As shown in c.

[0106] For the street centerline data: First, add a floating-point field "Length" and use ArcMap's "Computational Geometry" function to calculate the length L of each street segment. Then, use a Python script to generate the midpoint of each street segment; using the midpoint as the center, generate a maximum circle that does not intersect any streetscape surfaces (i.e., the circle is completely within the street space), record the diameter of the circle, and assign this diameter value to the width field D of the street centerline.

[0107] (3) Constructing a heterogeneous graph model

[0108] This step uses the Python programming language and calls third-party libraries such as geopandas, networkx, numpy, momoepy, rasterio, and scipy to complete the model construction. All operations are based on the pre-processed Shapefile and raster data.

[0109] (31) Constructing nodes

[0110] Read the building surface features and street centerline features processed in step (2). Use ArcMap's "Feature to Point" tool to obtain building nodes. and street nodes Each building node records its block ID, residential area ID, and total building area TA. Each street node records its street length L and width D.

[0111] (32) Assign weights to nodes

[0112] (321) Assigning authority to building nodes

[0113] Building Node Weights This represents the traffic demand intensity (population size) generated by the node. Population raster data is used to allocate population to buildings: First, the population raster data is converted into point features using the "Raster to Point" tool in ArcMap. Second, based on a Python script, the total population within each block (i.e., the sum of the population values ​​of all raster points within that block) is calculated, along with the sum of the total building area (TA) of all buildings within the same block. Figure 2 As shown in e. Finally, based on a Python script, the total population of the block is allocated to each building node according to the proportion of each building's area to the total building area of ​​its block, thus obtaining the population of each building node, which is the weight value of that building node. ,like Figure 2 As shown in f.

[0114] (322) Empowering Street Nodes

[0115] Street node weights This indicates the overall public transportation service capacity of the street location, calculated by aggregating multiple public transportation service capacity indicators.

[0116] (3221) Station allocation: Based on Python, each bus stop and subway entrance / exit is allocated to the corresponding street segment, and the station information on each street segment is recorded, including the number of subway entrances / exits E.

[0117] (3222) Indicator Calculation:

[0118] Subway service frequency F: Identify the subway stations associated with each street segment and obtain all subway lines served by each station. Based on line operation data, count the number of departures per hour for each line during weekday morning peak hours (e.g., 7:00-9:00), and take the average of these departures as the subway service frequency F for that street segment.

[0119] Metro Network Betweenness Centrality (BC): This method uses Python's NetworkX library to construct a metro network topology graph, with metro stations as nodes and line connections as edges. It calculates the betweenness centrality of each node, which is the proportion of paths passing through that node in all shortest paths. The betweenness centrality value of each station is then assigned to its corresponding street segment.

[0120] Metro network proximity centrality (CC): In the metro network topology diagram above, NetworkX is used to calculate the proximity centrality of each station, which is the reciprocal of the sum of the shortest path distances from that node to all other nodes in the network. Similarly, the proximity centrality value of a station is assigned to the corresponding street segment.

[0121] Number of stations reachable within 20 minutes (N): Based on the average subway operating speed (40 km / h) and transfer time (5 minutes), calculate the number of other subway stations reachable within 20 minutes from the current station. Specific method: Taking the current station as the starting point, considering path time and transfer time, perform a time-constrained accessibility analysis on the NetworkX graph, and count the total number of stations reachable within 20 minutes (including transfer time).

[0122] Number of bus routes L: Identify all bus routes assigned to this street segment, remove duplicate routes, and count the outbound and return directions as independent routes to obtain the total number of bus routes L for this street segment.

[0123] Public transport directional diversity (D): For each street segment bus stop, based on the direction of each bus route it serves, calculate the average directional vector of the downstream stops of that route. Map the directional vectors to 8 directional intervals (North, Northeast, East, Southeast, South, Southwest, West, Northwest), and calculate the proportion of routes in each directional interval. The directional diversity index is calculated using the information entropy formula: .

[0124] Bus-subway transfer convenience MC: Starting from all bus stops on the street segment, considering all bus lines that stop at the stop, count the number of different subway lines that can be transferred to within 2 kilometers along the downhill direction of each line.

[0125] All the above indicators are calculated in batches using Python scripts (combined with libraries such as ArcPy, NetworkX, GeoPandas, Shapely, and NumPy), and the results are stored in the attribute table of the street node elements corresponding to the street centerline.

[0126] (3223) Entropy Weight Aggregation: Five indicators for the metro (metro service frequency F, metro network betweenness centrality BC, metro network proximity centrality CC, number of stations reachable within 20 minutes N, and number of metro entrances / exits E) are normalized, and the information entropy and weight of each indicator are calculated. The weighted sum is then used to obtain the metro service capacity score. Three indicators for public transportation (number of bus routes L, bus direction diversity D, and bus-metro transfer convenience MC) are normalized, and the information entropy and weight of each indicator are calculated. The weighted sum is then used to obtain the public transportation service capacity score. The metro score and the public transportation score are used as input, and normalization is performed again. The information entropy and weight are calculated. The weighted sum is then used to obtain the comprehensive service capacity weight of the street node. ,like Figure 2 As shown in g. For street segments without any public transportation service, .

[0127] (33) Construct edges;

[0128] (331) Construct the edges between building nodes ;

[0129] Using Python's momepy library, a Voronoi diagram is created for all buildings within each "street block." Voronoi polygon pairs sharing edges are identified, and undirected edges are established between the corresponding two building nodes. The cost of each edge is calculated. Defined as the Euclidean distance between the mass centers of corresponding building surfaces of two building nodes, such as... Figure 2 As shown in h.

[0130] (332) Construct the edges between street nodes ;

[0131] Street-based network topology: For any two street segments that intersect in space (share endpoints), an undirected edge is established between their corresponding two street nodes, such as... Figure 2 As shown in i. The cost of the edge. The calculation formula is ,in, and They represent the lengths of street segments m and n, respectively. and The corresponding street width.

[0132] (333) The edge between building nodes and street nodes ;

[0133] Based on the adjacency relationship between a building's Voronoi polygon and the street centerline: if a building's Voronoi polygon is spatially adjacent to a street segment (sharing a boundary), then a directed edge (directing from the building to the street) is established between the building node and the street node, such as... Figure 2 As shown in j. The cost of the edge. Defined as the shortest Euclidean distance from the boundary of a building surface element to the centerline of the corresponding street segment.

[0134] (334) Correcting edges based on spatial constraints;

[0135] Water system constraints: Read water system surface features and check the edges between building nodes. and the edges between building nodes and street nodes Check whether an edge in the polygon crosses a water system (i.e., the straight line connecting the edge intersects the polygon of the water system). If it does, delete the edge.

[0136] Enclosed residential area constraints: for the edges between building nodes If an edge in the diagram has two endpoints belonging to different enclosed residential areas and both located within those areas, then that edge is deleted. For buildings within enclosed residential areas, simulate real-world entrance / exit restrictions: if a building in a residential area is adjacent to an external street, then delete the edge connecting all building nodes and street nodes within that residential area. In the edge selection, retain the edge with the lowest cost as the main entrance / exit on that side, and delete the rest; if no residential building is directly adjacent to an external street, then the edges between the initial building nodes are... The system retains a minimum-cost edge that connects the internal building complex to the nearest entrance / exit, with all internal buildings connected by edges between the retained building nodes. They are interconnected and connect to external streets through this entrance / exit, such as Figure 2 As shown in k.

[0137] (4) Calculate the building-level public transportation convenience;

[0138] (41) Calculate the access cost from the building to the adjacent street in the block;

[0139] For each building node, using its location as the boundary of the "street block", extract the set of building nodes and the edges between building nodes. and the edges between building nodes and street nodes This forms a local subgraph. Using Dijkstra's shortest path algorithm, the minimum cost to reach each street node constituting the boundary of the block is calculated, denoted as . ,in For building node indexing, For street node indexing, such as Figure 2 As shown in m.

[0140] (42) Calculate the walking cost between nodes in the city street network;

[0141] From all street nodes The edge between street nodes On the city's street network map, Dijkstra's algorithm is used to pre-calculate the shortest path cost between any two street nodes, resulting in a cost matrix. The elements in the matrix represent the cost of the shortest path from the street node. To street nodes The shortest path cost, denoted as .

[0142] (43) Calculate the total walking cost from the building to each public transport service street;

[0143] For each building node It can reach any street node of public transport service. The minimum total walking cost is:

[0144]

[0145] Here, k iterates through all adjacent street nodes in the block containing building i. Only retaining... Street nodes with a maximum walking distance threshold are designated as accessible bus service streets for the building.

[0146] (44) Calculate the building-level public transport convenience index ;

[0147] An improved potential model is used, replacing the traditional summation operator with a maximization operator to simulate residents' behavior in choosing the optimal site, such as... Figure 2 n、 Figure 2 As shown in o. For each building node The formula for calculating its convenience index is:

[0148] Among them, 400 meters is the threshold for comfortable walking distance, and the attenuation coefficient makes the utility decrease to 1 / e for every 400 meters increase in distance.

[0149] Final city-wide building-level public transport accessibility index Distribution, such as Figure 3 As shown.

[0150] (45) Calculate the street-level public transport convenience index

[0151] To obtain the convenience pattern of an overall urban-scale layout, the building-level convenience index can be aggregated to the block-level, and the public transport convenience index of each block can be calculated. Calculated using a population-weighted average method:

[0152]

[0153] in, Represents building nodes. This represents the weight value of the building node. For each building node Convenience index.

Claims

1. A building-level public transportation convenience assessment method based on a heterogeneous graph, characterized by, Including the following steps: S1. Obtain building data, street block data, public transportation station and route data, gated residential area boundary data, water system data, and population distribution data for the assessment area; S2. Preprocess the acquired data; S3. Construct a heterogeneous graph model, including: S31. Based on building data and street block data, obtain building nodes and street nodes; S32. Assign weights to building nodes and street nodes, wherein the weights of building nodes are determined based on the population distribution data, and the weights of street nodes are determined based on the integrated public transportation service capacity associated with them. S33. Based on the spatial topological relationships between building nodes, street nodes, and between building nodes and street nodes, construct the edges connecting each node; S34. Define the walking cost for the constructed edges; S4. Calculate the public transportation convenience index based on the constructed heterogeneous graph model.

2. The building-level public transit accessibility evaluation method based on heterogeneous graph according to claim 1, characterized in that, In step S1, the building data includes building outline features, geographical location information, and number of building floors; the street block data includes street centerline, road grade, and planned width; the public transportation station and route data includes the names, geographical location information, service line numbers, and departure frequency of bus and subway stations; the gated residential area boundary data and water system data are used to identify physical barriers to pedestrian paths; and the population distribution data is raster-formatted population density data.

3. The building-level public transit accessibility evaluation method based on heterogeneous graph according to claim 1, characterized in that, In step S2, the preprocessing includes: The acquired data is checked, corrected, missing data is added, and duplicates are removed. Street centerline data is generated based on street block data, and each street and block is assigned a unique number; buildings located in gated residential areas are identified through spatial overlay analysis, and residential area numbers are added; Calculate the building's total floor area, the length and width of the street's centerline.

4. The building-level public transit accessibility evaluation method based on heterogeneous graph according to claim 1, characterized in that, In step S32, the weight of the street node is determined based on its associated comprehensive public transportation service capacity, wherein the comprehensive public transportation service capacity is calculated by aggregating multiple public transportation service capacity indicators; the public transportation service capacity indicators include at least one of subway service capacity indicators and bus service capacity indicators; the subway service capacity indicators include indicators reflecting station service level and network accessibility, and the bus service capacity indicators include indicators reflecting bus service intensity and network connectivity; a multi-indicator comprehensive evaluation method is used to aggregate the public transportation service capacity indicators.

5. The building-level public transit accessibility evaluation method based on heterogeneous graph according to claim 1, characterized in that, Step S33 specifically includes: S331. Perform Voronoi tessellation on each building and establish edges between building nodes for the building nodes corresponding to adjacent Voronoi polygons. S332. Based on the road topology connection relationship, establish the edges between street nodes corresponding to intersecting street segments; S333. Based on the adjacency relationship between the building Voronoi polygon and the street centerline, establish the edges between the building node and the street node.

6. The method for assessing building-level public transportation convenience based on heterogeneous graphs according to claim 1, characterized in that, Step S34, which defines the walking cost for the constructed edge, specifically involves: The edge cost between street nodes is the Euclidean distance between the building centroids; The edge cost between street nodes is calculated using the length and width of the street segment. The edge cost between building nodes and street nodes is the shortest Euclidean distance from the building boundary to the street centerline.

7. The method for assessing building-level public transportation convenience based on heterogeneous graphs according to claim 1, characterized in that, Step S4, specifically, calculating the public transportation convenience index based on the constructed heterogeneous graph model, involves: S41. Use Dijkstra's algorithm to calculate the minimum cost for each building node to reach each street node at the boundary of its block. S42. Use Dijkstra's algorithm to calculate the minimum cost between any two street nodes in the evaluation area; S43. Calculate the minimum total walking cost from each building node to any public transport service street node; Set a maximum walking distance threshold to filter the streets accessible by public transport for each building; S44. Using an improved potential model, the traditional summation operator is replaced with a maximization operator to calculate the public transportation convenience index for each building node, as shown in the following formula: ； in, As street node weights, The minimum total walking cost to reach any public transport service street node from each building node. The maximum walking distance threshold, The threshold for comfortable walking distance; S45. The public transport accessibility index for each block is calculated using a population-weighted average method.

8. The method for assessing building-level public transportation convenience based on heterogeneous graphs according to claim 1, characterized in that, The building nodes are defined by the centroid of the building, and the street nodes are defined by the centroid of the street.

9. A computer device, characterized in that, It includes one or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, wherein when the programs are executed by the processors, they implement the steps of a heterogeneous graph-based building-level public transport convenience assessment method as claimed in any one of claims 1-8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the building-level public transport convenience assessment method based on heterogeneous graphs as described in any one of claims 1-8.

Images