Walkability is much studied, but the relative importance of perceptions and motivations is still
not consensual. This study took a holistic approach to examine the comparative importance
of a range of possible perceptions, motivations and individual characteristics on walking
levels. Data from Auckland Transport's Active Modes online survey (AT survey, N= 4,114)
captured environmental perceptions and travel behaviour. Machine learning (gradient
boosting) was used to predict walking levels from perceptual data and individual …