Predicting from GPS and Accelerometer Data When and Where Tourists Have Viewed Exhibitions

Research has been conducted to understand tourists’ spatio-temporal behaviours. However, it is very costly to investigate what the tourist was actually doing at each location and moment and what he/she was interested in Kawase et al. (When and where tourists are viewing exhibitions: Toward sophistication of GPS-assisted tourist activity surveys. Springer, Vienna, pp. 415–425, 2012) demonstrated the possibility that we can predict only from a tourist’s GPS log whether he/she is viewing an exhibition or not, which is one of the most basic activities in tourism. Following their work, we conduct an additional experiment two types of subjects, students and kindergarteners with parents, and refine the prediction model with additional explaining parameters. We found that the model for students could be successfully improved, while that for kindergarteners has a problem due to the inconsistency of their behaviour. In addition, we experimentally investigated the combined use of a GPS sensor and an accelerometer, both usually equipped in smartphones, for predicting tourists’ viewing state. The result shows that the combined use of these sensors seems promising to infer tourists’ activities.