Mining moving patterns for predicting next location

摘要: Next location prediction has been an essential task for many based applications such as targeted advertising. In this paper, we present three basic models to tackle the problem of predicting next locations: Global Markov Model that uses all available trajectories discover global behaviors, Personal focuses on mining individual patterns each moving object, and Regional clusters mine similar movement patterns. The are integrated with linear regression in different ways. We then seek further improve accuracy by considering time factor, a focus clustering periods, methods train time-aware periodic Therefore, our proposed have following advantages: (1) consider both collective making prediction, (2) similarity between trajectories, (3) factor build suited periods. conducted extensive experiments real dataset, results demonstrate superiority over existing methods. HighlightsWe propose (PMM, GMM RMM) combine them ways obtain new predict object.GMM patterns; PMM object using its own past trajectories; RMM patterns.Based observation often change time, can capture relationships use knowledge more refined models.To best knowledge, first ones take holistic approach individual, prediction.We conduct dataset effectiveness models.