Private Cars Trajectory Interpolation by Ensemble Learning Optimization in Vehicular Social Networks

Abstract

As an important part of Vehicular Social Network (VSN), the researches of private cars are neglected some factors due to the difficulty of collecting. For example, the social characteristics are ignored, which hinder the study of the social attributes of vehicles in VSN. However, trajectory data for large-scale private car researches often have the challenge of trajectory availability, mainly due to the missing of trajectories under GPS outages, which results in the collected trajectory data not being directly applicable to VSN applications. In this paper, a novel ensemble learning framework with Learn++ and Bayesian optimization is proposed, with ensemble learning as the major solution for trajectory missing during GPS outages. We utilize the real-world datasets to verify the accuracy of the comparison methods and the proposed framework in the prediction of trajectory interpolation. The experimental results show that our framework is significantly superior to the comparison methods. We utilize several common road trajectories to verify the accuracy of the proposed trajectory interpolation prediction framework. The experimental results show that our framework is superior to the baselines.