The Urbanity Global Network Dataset

The Global Urban Network (GUN) dataset provides pre-computed node and edge attribute features for various cities. Each layer is available in .geojson format and can easily be converted into NetworkX, igraph, PyG, and DGL graph formats. <br> For node attributes, we adopt a uniform Euclidean approach, as it provides a consistent, straightforward, and extensible basis for integrating heterogeneous data sources across different network locations. Accordingly, we construct 100 metres euclidean buffers for each network node and compute the spatial intersection with spatial targets (e.g., street view imagery points, points of interest, and building footprints). To ensure spatial consistency and accurate distance computation, we project spatial entities into local coordinate reference systems (CRS). Users can employ the Urbanity package to generate Euclidean buffers of arbitrary distance. <br> For edge attributes, we adopt a two-step approach: 1) compute the distance between each spatial point of interest and its proximate edges in the network, and 2) assign entities to the corresponding edge with lowest distance. To account for remote edges (e.g., peripheral routes that are not located close to any amenities), we specify a distance threshold of 50 metres. For buildings, we compute the distance between building centroids and their respective network edge. Accordingly, we compute spatial indicators based on the set of elements assigned to each network edge. <br> We also release aggregated subzone statistics for each city. Similarly, users can employ the Urbanity package to generate aggregate statistics for any arbitrary geographic boundary. <br> Urbanity Python package: https://github.com/winstonyym/urbanity. <br> <br> <br> <br> <br>全球城市网络（Global Urban Network, GUN）数据集为各类城市提供预计算的节点与边属性特征。所有图层均采用.geojson格式存储，可便捷转换为NetworkX、igraph、PyG及DGL图格式。 在节点属性构建环节，本数据集采用统一欧几里得方法，该方案可为整合不同网络节点处的异构数据源提供一致、简洁且可扩展的基础框架。据此，我们为每个网络节点构建100米欧几里得缓冲区，并计算其与各类空间目标的空间交集，例如街景影像点位、兴趣点（points of interest）及建筑基底（building footprints）。为确保空间一致性与精准的距离计算，我们将空间实体投影至局部坐标参考系统（coordinate reference systems, CRS）。用户可借助Urbanity工具包生成任意距离的欧几里得缓冲区。 在边属性构建环节，本数据集采用两步处理流程：其一，计算各空间目标点位与网络中邻近边的距离；其二，将实体分配至距离最小的对应网络边。针对偏远边（如未邻近任何配套设施的外围道路），我们设定了50米的距离阈值。对于建筑实体，我们计算其质心与对应网络边的距离。据此，我们基于分配至各网络边的元素集合计算各类空间指标。 本数据集还发布了各城市的聚合分区统计数据。同理，用户可使用Urbanity工具包为任意自定义地理边界生成聚合统计值。 Urbanity Python工具包：https://github.com/winstonyym/urbanity.