Fig. 5
From: City-scale car traffic and parking density maps from Uber Movement travel time data
The core stages of the modeling method. (a) The city of interest (here e.g. Melbourne) is divided into zones for which we are given mean travel time data for a subset of these zones and day times. (b) A graph representation of the city zones. Each node stands for a zone and each value incorporated in the node stands for the probability that a car located in that zone drives. The probabilities of choosing a certain destination zone are assigned to each latency of the graph. The flow of cars between city zones is sampled from the joint probabilities of driving and choosing a destination zone. (c) The distribution of cars parked in each zone in each time step is calculated as a function of traffic flow between all zones. (d) A set of gradient descent algorithms performs model selection by tuning the open model parameters.
