The 20th and 21st joint COTA International Conference of Transportation Professionals (CICTP 2020-21)

In recent years, many researches focus on inferring transportation modes from traveler's GPS trajectory data. Generally speaking, algorithms for solving this problem can be divided into two steps. The first step is to divide the GPS trajectory which may have multiple transportation modes into single transportation mode segments, and the second step is to infer the transportation mode of each segment. Most of the current researches focus on the second step, while ignoring the attempt to improve the effectiveness of the first step. In this paper, we focus on the first step of inferring transportation mode from GPS trajectory data, and try to improve the time accuracy, recall, and precision values of the transportation mode change point detection. Compared with previous solutions to this problem, our method adopts some new strategies to improve the effectiveness of detection. Applied to the Geolife dataset, our method achieves relatively good result. The recall of transportation mode change point detection is 100%, and 87.8% of them are detected within 20 seconds of their real moment. In the experiment analysis section of this paper, we describe the time intervals between each real transportation mode change point and the corresponding point we detected, which is not mentioned in previous researches.

CICTP