Modeled probability of congestion for selected cities based on Twitter and OpenStreetMap data on a grid cell basis with a resolution of 100 meters. The data set includes the cities of Barcelona, Berlin, Cincinnati, Kiev, London, Madrid, Nairobi, New York City, San Francisco, Sao Paulo and Seattle. The range of values is from 0 (probably normal traffic flow) to 1 (high probability of traffic flow delay). Methodology: Based on Twitter and OpenStreetMap (OSM) data, a model was trained with the help of machine learning, which predicts the probability of traffic jams within the cities. Publicly provided data from UBER was used as reference data (https://movement.uber.com). The number of tweets and the number of points of interest from OSM near roads were used as indicators in the model. In addition, car journeys were simulated with the help of the openrouteservice based on the spatial distribution of the population and relevant POIs and taken into account in the model.
