Hansen’s Accessibility Theory and Machine Learning: a Potential Merger

Accessibility is a central concept in transport geography, given its relationship with land development. It is defined as "the opportunity that an individual at a given location possesses to take part in a particular activity or set of activities" (Hansen, 1959). Hansen's Accessibility Model (HAM) can be computed using mobility flows between regions or employment data as a proxy for centres of attraction, coupled with an impedance function that incorporates travel costs. It has been the basis of multiple theoretical and empirical approaches over the years. In the last decades, advances in Machine Learning (ML) have also opened new possibilities for developing innovative approaches in the transport field. The objective of this work is primarily the study of the dynamics of urban accessibility, by considering two interrelated perspectives. Firstly, we aim to explore whether alternative data-based techniques, such as ML, can replicate the behaviour of HAM and thus capture, from data, the underlying theory linked to the spatial interaction model, by embedding the influence of geography, transport network, and socioeconomic factors on accessibility. Secondly, we investigate the feasibility of employing ML where flow data are unavailable, ensuring consistent measurements over time. A combined approach HAM-ML is developed and applied to this aim. As a case study, we examine inter-urban accessibility in two Italian regions, Lombardia and Emilia Romagna, based on socioeconomic and transport data from 2011. The results show the potential of this joint approach, opening new research prospects on accessibility from the theoretical, empirical, and policy viewpoints.