Datasets and Scripts for Reproducing Results in "Algorithmic Optimization of Facility Location and Demand Allocation in Shared Mobility Systems with Asymmetric Distance Measures”

The dataset combines real-world travel distance data with simulated and empirical demand locations from three urban scenarios: Bangkok, Guangzhou, and the National Taiwan University (NTU) U-Bike shared bicycle system. Demand locations for Bangkok and Guangzhou are generated within the corresponding planning regions to approximate realistic spatial distributions of users, while the NTU U-Bike dataset is constructed from real-world shared mobility infrastructure. The provided map images illustrate these planning regions and serve as the spatial reference for the experiments.

Travel distances are obtained via the Google Maps Distance Matrix API and exhibit inherent asymmetry due to traffic conditions, one-way road structures, and urban network characteristics.

The dataset includes:
- Demand point data (data_*.csv): geographic coordinates of demand locations used in the experiments. For Bangkok and Guangzhou, the datasets contain simulated geographic coordinates of 100 demand locations within the corresponding planning regions. For the NTU U-Bike case, the dataset contains 60 real-world station locations collected from the shared bicycle system.
- Regional information (data_Bangkok_regions.csv): spatial partitioning used for the land cost analysis.
- Asymmetric dissimilarity matrices (dissimilarity_matrix_*.csv): pairwise travel distances between demand points, where d(i,j) ≠ d(j,i). Each matrix is aligned with the ordering of demand points in the corresponding data file.
- K-means evaluation data (KMeans-by-LatLng_Distances-To-Centers_*.json): precomputed travel distances between demand points and cluster centers. These are provided because K-means centers are not restricted to observed demand locations, and additional distance queries are required.
- Map images (Bangkok.png, Guangzhou.png, ntu_ubike.png): background maps representing the planning regions and serving as the spatial reference for visualization.
- Implementation script (main.ipynb): code for reproducing the experiments, including clustering procedures, synthetic asymmetric distance matrix generation, and objective value evaluation.

Travel distance data are obtained via the Google Maps Distance Matrix API. Due to usage restrictions associated with third-party map services, raw API responses are not included. However, all processed data, coordinates, and necessary procedures are provided to enable reproducibility.

The dataset supports research on asymmetric distance modeling, data-driven location–allocation, and clustering-based optimization in real-world urban environments.