Predicting complex trajectories in large-scale moving object database
Today, moving object databases can answer near future queries by using mobility patterns. Distant future queries can also be an- swered by using pattern mining, statistical models or machine learning techniques. However, since the latter models relies on frequently visited places to make predictions, they are unable to answer queries related to the trajectories taken by moving objects falling in-between these locations. Such queries could be expressed as: “Find all the users that will travel in the vicinity of a given location during their next move with a probability higher than a given threshold”.
A key motivation to answer queries of such nature may be to lter customer notications based on their movements. This would eliminate mass mailing all the customers of a restaurant or a shop and target those who will be around a specic location with a high probability at some point in the future. At this connected age, where the services are spamming users with a large volume of no- tications, it would be benecial for all parties to lter out selected users so as to improve the general experience.
This project aims at improving an existing model we recently proposed to solve this issue. The goal will then consists in demonstrating that the proposed solution can be scaled to large scale databases.