MMTD: Madrid Fixed-Sensor Traffic Data (Processed Sample for Reproducible Research)

This dataset contains processed fixed-sensor traffic data from Madrid, Spain (August 2024), as part of the Metropolitan Multi-source Traffic Data (MMTD) project. It includes 15-min aggregated traffic states (flow, occupancy, speed/congestion index) from 2,476 loop detectors (2,371 urban + 105 highway), aligned to a unified OSM-based road network with 25,166 segments. Datasource Attribution: Original data source: Ayuntamiento de Madrid (datos.madrid.es) - Traffic historical data since 2013 Original data license: The underlying raw data is provided by the Madrid City Council under the terms of the Madrid Open Data initiative. Reuse requires proper attribution to the source and author, in accordance with Real Decreto Legislativo 1/1996. For full terms, please visit: https://datos.madrid.es/dataset/208627-0-transporte-ptomedida-historico. Key features: Spatio-temporally aligned with the MMTD pipeline Includes road static attributes (length, lanes, speed limit, centrality) Graph topology (adjacency matrix) for deep learning applications Cleaned (outlier removal) but no imputation – suitable for evaluating reconstruction methods This is a SAMPLE dataset containing only the fixed-sensor component. To obtain the full MMTD (including TomTom floating car data): Acquire TomTom Traffic Stats data for Madrid (August 2024) under TomTom's terms Use the binding scripts in the accompanying GitHub repository to align and merge with this fixed-sensor data Companion code: All processing scripts, pipeline documentation, and PyG-style data loaders are available on GitHub: https://github.com/qishenzhou/MMTD. Data format: distance_matrix.h5 `README`: JSON-encoded metadata and usage example. Always read this first to understand the file structure. `data/values`: 2D float array, shape (n_nodes, n_nodes), shortest path distances (km) `metadata/nodes`: 1D array of node IDs (for both rows and columns) road_info.h5 `README`: JSON-encoded metadata and usage example. Always read this first to understand the file structure. `data/values`: 2D float array, shape (n_nodes, n_features), road static features `metadata/nodes`: 1D array of node IDs `metadata/features`: 1D array of feature names fixed_sensor_data.h5 `README`: JSON-encoded metadata and usage example. Always read this first to understand the file structure. `data/values`: 3D float array, shape (n_timestamps, n_nodes, n_features), traffic sensor data `metadata/times`: 1D array of time indices `metadata/nodes`: 1D array of node IDs `metadata/features`: 1D array of feature names Example to load the data ```python import h5py import pandas as pd import json with h5py.File('distance_matrix.h5', 'r') as f: readme_str = f['README'].asstr()[()] readme = json.loads(readme_str) for key, value in readme.items(): print(key, value) # You will see the description of the file, data structure, features description, data format, and usage example ```