Research on a Distribution-Outlier Detection Algorithm Based on Logistics Distribution Data

Abstract

In this paper, we propose an outlier detection algorithm, which improves the existing algorithm from three aspects: the time outlier, the outlier detection of logistics distribution path, the time complexity. Aiming at the time outlier, this paper proposes an abnormal objective function based on the average speed of the logistics distribution process and realizes the detection of the outlier time of logistics distribution. In the aspect of the outlier detection of logistics distribution path, this paper combines the distance- and the angle-based outlier detection algorithms to realize the detection of the local abnormal situation of the logistics distribution trajectory. At the same time, the high time complexity of the algorithm is reduced by only retrieving the path sub-sections in the neighborhood of the proper radius. Experimental results indicate that the efficiency and speed of the proposed algorithm are better than those of the trajectory outlier detection (TRAOD) algorithm, and the proposed algorithm is suitable for outlier detection of logistics distribution.