An ensemble Method for Urban Travel Time Prediction

An accurate time prediction of a path is of great importance to route planning and navigations. With the widespread use and popularity of the Global Positioning System (GPS), it is possible to collect massive vehicle trajectories for travel time prediction. However, most existing travel time prediction approaches are segment-based methods, which are not accuracy as they are not able to capture many cross-segment factors. In this paper, we propose a unique feature extraction structure which takes multiple features into account. Besides, we also use ensemble method to predict travel time which can take advantage of every single model strength. A series of experiments have been conducted on real datasets, the result demonstrates that MFPEL outperforms over existing approaches.