Abstract—Mobility prediction is an important problem having numerous applications in mobile computing and pervasive systems. However, many mobility prediction approaches are not noise tolerant, do not consider collective and individual behavior for making predictions, and provide a low accuracy. This paper addresses these issues by proposing a novel dependency-graph based predictor for real-time route prediction, named MyRoute. The proposed approach represents routes as a graph, which is then used to accurately match road network architecture with real-world vehicle movements. Unlike many prediction models, the designed model is noise tolerant, and can thus provide high accuracy even with data that contains noise and inaccuracies such as GPS mobility data. To cope with noise found in trajectory data, a lookahead window is used to build the prediction graph. Besides, the proposed approach integrates two mechanisms to consider both the collective and individual mobility behaviors of drivers. Experiments on real and synthetic datasets have shown that the performance of the designed model is excellent when compared to two state-of-the-art models.

Index Terms—Real-time, route prediction, dependency graph, mobility graph, noise tolerance.

Cite: Hanane Amirat, Nasreddine Lagraa, Philippe Fournier-Viger, and Youcef Ouint, "MyRoute: A Graph-Dependency Based Model for Real-Time Route Prediction," Journal of Communications, vol. 12, no. 12, pp.  668-676, 2017. Doi: 10.12720/jcm.12.12.668-676.