Youtube 3D traffic map in 20 France cities. The map is interactive, you can zoom in/out and rotate the map.

The access to the data is restricted to the participants of the Netmob 2023 Data Challenge and to the terms and conditions.

The NetMob 2023 Data Challenge is a unique opportunity to access data that is typically very difficult to come by, and employ it to generate new knowledge, support innovation, prove theories at scale, or validate solutions in real-world settings. Prospective participants shall apply to the challenge by registering at https://netmob2023challenge.networks.imdea.org/. Complete instructions are provided at the same webpage.

The dataset is composed by:

• 20 urban areas in France (Paris, Lyon, Marseille, Toulouse, Nice, etc.)
• 68 mobile services (including YouTube, Netflix, Facebook, Instagram, Gmail, etc.)
• 77 days continuous days
• 100 x 100 m2 spatial resolution
• 15 minute temporal resolution
• 400+ billion data points
• 2.3+ TB of data

The full description of the dataset and the methodology can be found in our pre-print available here: The NetMob23 Dataset: A High-resolution Multi-region Service-level Mobile Data Traffic Cartography.

Please cite our work, when using the dataset:

@misc{netmob23, 
  title={The NetMob23 dataset: A high-resolution multi-region service-level mobile data traffic cartography}, 
  author={Martínez-Durive, Orlando E and Mishra, Sachit and Ziemlicki, Cezary and Rubrichi, Stefania and Smoreda, Zbigniew and Fiore, Marco}, 
  year={2023}, 
  eprint={2305.06933}, 
  archivePrefix={arXiv}, 
  primaryClass={cs.NI} 
}

The spatial dataset is composed of 20 France cities. However, the formal definition in most of the cases is Metropole, an administrative entity in France, in which several communes cooperate, and which has the right to levy local tax, an établissement public de coopération intercommunale à fiscalité propre.

The list of communnes that compose these cities are given in the cities_communes_code.json file, while the list of IRIS (smaller administrative units) are given in the cities_iris_code.json file. GeoJson files for cities, using communes are in the cities_communes folder, while the ones using IRIS are in the cities_iris folder.

Each one is represented by a geojson file, that contains the grid of the city composed of 100 x 100 m² tiles. The tile (feature) is represented by a polygon using WGS84.

{"type": "Feature",
 "geometry": {"type": "Polygon",
  "coordinates": [[
    [4.7662070878542515, 45.55631465259445],
    [4.766246657177647, 45.55721386239888],
    [4.767526651889026, 45.5571860567685],
    [4.767487061740877, 45.55628684742171],
    [4.7662070878542515, 45.55631465259445]
    ]]},
 "properties": {"tile_id": 66}}

By using the tile_id we can also represent the spatial information in a matrix form; where each tile represented a matrix cell.

$row_{index} = floor(\frac{tile_{id}}{cols}) $

$col_{index} = tile_{id} \mod cols$

The matrix dimension (rows, cols) are given in the cities_dims dictionary.

cities_dims = {
    "Bordeaux": (334, 342), 
    "Clermont-Ferrand": (208, 268),
    "Dijon": (195, 234),
    "Grenoble": (409, 251),
    "Lille": (330, 342),
    "Lyon": (426, 287),
    "Mans": (228, 246),
    "Marseille": (211, 210),
    "Metz": (226, 269),
    "Montpellier": (334, 327),
    "Nancy": (151, 165),
    "Nantes": (277, 425),
    "Nice": (150, 214),
    "Orleans": (282, 256),
    "Paris": (409, 346),
    "Rennes": (423, 370),
    "Saint-Etienne": (305, 501),
    "Strasbourg": (296, 258),
    "Toulouse": (280, 347),
    "Tours": (251, 270)
    }

The notebook Regions.ipynb contains code snippets of how to load the geojson files and plot the regions. For a simple visualization, you can just load the geojson file in the geojson.io website.

The traffic dataset is composed of 209440 plain text files. Each one for each combination of city, service, day and direction of the traffic. A given file, contains the traffic record for all the tiles of the city every 15 minutes.

Due to daylight saving time in France on March 31, 2019, the traffic records for that day have 92-time columns instead of 96. These missing timestamps are between 02:00 and 03:00 French local time. Therefore, the first 8-time columns cover the period from 00:00 to 01:45, and the 9-time column corresponds to 03:00.

The notebook Traffic.ipynb contains code snippet on how to load the traffic records from the txt files and plot traffic maps.

Also, how is possible to aggregate the traffic records over space to obtain traffic time series.