Proceedings of the 13th ACM SIGKDD international conference on Knowledge discovery and data mining
WhereNext: a location predictor on trajectory pattern mining
Proceedings of the 15th ACM SIGKDD international conference on Knowledge discovery and data mining
Prediction of indoor movements using bayesian networks
LoCA'05 Proceedings of the First international conference on Location- and Context-Awareness
Hi-index | 0.00 |
Existing prediction methods in moving objects databases cannot work well on fragmental trajectory such as those generated by mobile data. Besides, most techniques only consider objects' individual history or crowd movement alone. In practice, either individual history or crowd movement is not enough to predict trajectory with high accuracy. In this paper, we focus on how to predict fragmental trajectory. Based on the discrete trajectory obtained from mobile billing data with location information, we proposed two prediction methods: Crowd Trajectory based Predictor which makes use of crowd movement and Individual Trajectory based Predictor uses self-habit to meet the challenge. A hybrid prediction model is presented which estimates the regularity of user's movements and find the suitable predictor to gain result. Our extensive experiments demonstrate that proposed techniques are more accurate than existing forecasting schemes and suggest the proper time interval when processing mobile data.