Distinguishing different types of city tourists through clustering and recursive logit models applied to Wi-Fi data

Abstract

We discuss the possibilities to distinguish different types of tourists based on Wi-Fi sensor data. The data are obtained from 20 sensors employed in Higashiyama, Kyoto, which is an area highly frequented by tourists. We describe tourist-tours as a sequence of sensors at which they are observed. Based on these records a clustering approach is chosen where we select as clustering variables, among others, the degree of detours and the length of time they are observed. We find that we can distinguish groups of tourists that are visiting a number of sightseeing spots in a short time from others who walk through the area more leisurely and are likely enjoying souvenir shops and restaurants. For the main tourist clusters than a Recursive Logit approach is applied to model their route-choice based on path length and attractions en-route. We find that the estimated parameters reflect these group characteristics.