Context-aware next location prediction using data mining and metaheuristics

Abstract

Due to the heavy use of smartphones and other GPS enabled devices, researchers have easy access to substantial mobility data. Many existing techniques predict the next location of the users based on their mobility traces which includes only geographical coordinates in the form of spatio-temporal data. These raw mobility traces possess hidden information known as location context. Contextual information of any location means its name, time spent there, associated activity, preferred visit time and many such parameters. Enriching raw mobility traces with such contextual information adds more value to it and more sense to it's applications. The proposed model performs geographical, contextual and behavioural enrichment of raw trajectories. It also assigns relevant tag to each identified location automatically using metaheuristic approach. This paper proposes a model CANLoc to perform data collection, trajectory enrichment and location prediction. Performance of the proposed model is verified using two datasets: GeoLife and Mobi-India, which shows significant improvement in the location prediction accuracy.