The context of outdoor walking: A classification of user-generated routes

Leisure walking has known benefits to public health, from both physical and psychological viewpoints. Complementing traditional print information sources, dedicated online platforms and apps provide tools to search, discover, plan and share routes. While walking routes are highly heterogeneous in terms of their properties and geographical context, current platforms adopt simple representations of basic attributes. In this article, we report on a research project at the Ordnance Survey on the representation and recommendation of walking routes, which comprises the following contributions. Firstly, we outline a theoretical framework of leisure walking, intersecting its individual, social and environmental dimensions. Secondly, analysing about 4 million user-generated walking routes produced in Great Britain, we characterise routes combining primary attributes and contextual information, including land cover and points of interest. Thirdly, by applying unsupervised learning to this data model, we produce the Walking Route Classification. This classification is methodologically similar to geo-demographic classifications and identifies groups and supergroups of similar routes in a large multidimensional attribute space. This body of work is evaluated through a survey and a focus group with experts, showing encouraging results.