Research On Trajectory Dataset Simplification Based On Features

The pervasiveness of location-acquisition, mobile computing and sensor network techniques has generated massive spatial trajectory data, which has given rise to the locationbased services to make people’s lives more convenient. However, the large data scale has brought great challenges to the simplification and compression of such a big data. In this paper, we focus on the trajectory compression and trajectory dataset simplification respectively. In conclusion, this paper consists of the following achievements:(1) We have analyzed the exiting technology of trajectory preprocessing about the noise filtering, map-matching and trajectory segmentation. By comparing the different trajectory dataset simplification and compression technologies, we analyze the main problems of the exiting methods which have laid the groundwork for the next works.(2) We focus on the trajectory compression problem in road network, and propose a prediction-based trajectory compression method, which separate the spatial representation of a trajectory from the temporal representation. In spatial aspect, we use the prediction by partial matching model trained by the historical trajectories, which can predict the next location of a trajectory. In temporal aspect, we get the speed model of a trajectory to predict the timestamp when the trajectory enter the next road, which can ensure the temporal error is less than a given threshold. the experiments show that our method significantly outperforms existing approaches in terms of saving storage cost of trajectory data with bounded errors.(3) To tackle the problem of the trajectory dataset simplification, we propose a feature based sub-trajectory dataset simplification method, to extract the representative subtrajectories from the raw trajectories as a subset, which can best describe the whole dataset.First, we cut the raw trajectory by concerning their feature representativeness. Then, the extension method of the simulated annealing algorithm is performed to find the trajectories combination which minimize the feature error, and a local heuristic evolution strategy is proposed to speed-up the convergence. We evaluate our method based on extensive experiments by using two real-world trajectory datasets. The results demonstrate the efficiency and effectiveness of our methods in different applications.In conclusion, we have a research on the technology of trajectory dataset simplification and compression. It turns out that our methods not only reduce the data size but also reduce the information loss in the process, which can provide some reference for the related researches.