New paper: Quantifying the uncertainty of mobility predictions

July 17, 2023

Our latest work on the potential of deep learning for predicting people flows in cities is out:

Steentoft, A., Lee, BS. & Schläpfer, M. Quantifying the uncertainty of mobility flow predictions using Gaussian processes. Transportation (2023).

We propose a Bayesian approach that formulates deep neural networks as Gaussian processes and integrates automatic variable selection. This provides uncertainty estimates for predicted origin-destination flows and allows us to identify those geographic features that drive the mobility flows.

The basis for this work has been carried out by Aike Steentoft while doing his Ph.D. at Nanyang Technological University Singapore under the supervision of Asst. Prof. Markus Schläpfer.

Full text available here: