Predicting personal mobility with individual and group travel histories

Understanding and predicting human mobility is a crucial component of a range of administrative activities, from transportation planning to tourism and travel management. In this paper we propose a new approach that predicts the location of a person over time based on both individual and collective behaviors. The system draws on both previous trajectory histories and the features of the region—in terms of geography, land use, and points of interest—which might be ‘of interest’ to travellers. We test the effectiveness of our approach using a massive dataset of mobile phone location events compiled for the Boston metropolitan region, and experimental results suggest that the predictions are accurate to within 1.35 km and demonstrate the significant advantages of incorporating collective behavior into individual trip predictions. <br> <b>Keywords:</b> urban dynamics, human mobility, mobility prediction, mobile-phone data