The increasing availability of large-scale vehicle traffic data provides us great opportunity to explore them for knowledge discovery in intelligent transportation systems. Many mechanisms have been proposed to discover all outliers in a road network lately due to an increasing capability to track moving vehicles. In this paper, we propose a new problem called the road segment-based outliers detection problem, which is to find all road segments, called outliers, each of which “real” traffic deviates from its “expected” traffic. However, the recent state-of-the-art algorithms which was proposed for the region-based outlier detection problem is insufficient to solve our road segment-based outliers detection problem. Based on these insights, we propose a method find all outliers in the road segment-based road network. Finally, we conducted experiments on a large real dataset containing trajectories from 20,000 taxis. The results show that our proposed method outperforms the state-of-the-art method by 54%, 36% and 46% respectively in terms of precision, recall and F1-measure.
展开▼
