CDRoute: A Randomized Algorithm to Estimate Street-Level Commuting Routes Using Call Detail Records

In this work we propose a randomized algorithm to estimate the street-level commuting route using call detail records (CDRs). Our algorithm works in three steps. First, for each individual, we create perturbations of estimated home and work locations within the cell tower coverage. Second, we identify a route between each pair of both locations through a web mapping service. Finally, we choose one route that maximizes the normalized number of commuting hour calls within certain distance from the route. We evaluate our algorithm with large-scale CDRs collected in two big cities during three months. The result shows that our algorithm finds, on average, 45.28%p and 35.58%p better routes than a naive approach without the perturbation step.