Urban Traffic ST-KG: A Geographic-Scene-Based Spatiotemporal Knowledge Graph for Urban Traffic Using Trajectory Big Data
收藏资源简介：
Description This Zenodo repository contains the datasets and codes required to construct and apply the Urban Traffic Spatio-Temporal Knowledge Graph (ST-KG) and to reproduce the experimental results reported in the accompanying paper. Note on trajectory data:Due to data sharing restrictions, the released trajectory dataset includes trajectory data for two vehicles only, which are provided for methodological demonstration and reproducibility purposes. Researchers requiring access to larger-scale or complete trajectory datasets should contact the authors for further information. Code language Python (tested with Python ≥ 3.9) Main Requirements The code has been tested with the following dependencies: Fiona==1.8.22 folium==0.18.0 GDAL==3.6.2 geopandas==0.13.2 shapely==2.0.3 langchain==0.2.17 openai==1.65.2 py2neo==2021.2.4 pyarrow==17.0.0 pytorch-lightning==1.9.0 torch==1.13.1+cu117 Data Files and Descriptions All datasets are stored in the directory: code and data upload/Main data/ This directory includes the traffic, meteorological, spatial, and auxiliary datasets necessary for reproducing the experiments presented in the paper, subject to the trajectory data limitation described above. Detailed descriptions of: dataset contents, data preprocessing procedures, code execution steps, and the correspondence between datasets and experiments are provided in the file: readme.docx ⚠️ Users are strongly encouraged to consult readme.docx for a comprehensive explanation of the data structure and code execution before running the experiments. Please follow the instructions in readme.docx carefully and reproduce our experiments step by step. Naming Notes In the paper, the spatial unit is referred to as a “grid”. Since the grid cells are hexagonal in shape, they are referred to as “honeycomb” in the dataset and code. In the paper, the concept of “state” is derived from mapped trajectory points. In the dataset and code, this concept is directly referred to as “trajectory_point”, reflecting its origin from trajectory data after map matching.
