Abstract

Predicting the next location of a user from their movement history is useful in building intelligent applications that can continuously assist users without explicit user-input. Data collected by applications on consumer-grade mobile devices, such as GPS data, can have missing records (eg, due to the application crashing) and the sensor sampling frequency needs to be kept low so that it does not drain out the mobile battery. Thus, there can be a significant time gap between each pair of recordings. Our work in this paper focuses on predicting the next location of a mobile user using such sparse GPS data, collected at a very low frequency of once in every 10min. To give an example of dense data, Krumm and Horvitz (2005, 2006) use data collected once in every six seconds.

Date

October 14, 2025

Authors

Ayush Jaiswal, Y Chiang, Craig A Knoblock, Liang Lan

Journal

Proceedings of the 8th International Conference on Geographic Information Science

Pages

315-219