A Trajectory Cleaning Framework for Trajectory Clustering
Agzam Y. Idrissov, Mário A. Nascimento · 2012
clustering is the process of grouping similar tra- jectories according to a similarity distance. Several methods for trajectory clustering have been proposed. However, most of these methods require initial data preprocessing which are seldom explained. To that end, this paper proposes a framework for cleaning trajectory data, which consists of three steps: stop detection, missing segment interpolation and inaccuracy removal. Using Nokia's Mobile Data Chal- lenge dataset as a sample application for our framework and objective metric, we show that our data preprocessing ap- proach improves the quality of trajectory clustering.