An Enhanced Transportation Mode Detection Method Based on GPS Data

Abstract

Understanding meaningful information such as transportation mode (e.g., walking, bus) from Global Positioning System (GPS) data has great advantages in urban management and environmental protection. However, the urban traffic environment has evolved from “data poor” to “data rich”, resulting in the decline in the accuracy of transportation mode detection results. In this paper, an enhanced approach for effectively detecting transportation mode with a detection model and correction method from GPS data is proposed. Specifically, we make the following contributions. First, a trajectory segmentation method is proposed to detect single-mode segments. Secondly, a Random Forest (RF)-based detection model containing several new features is introduced to enhance discrimination. Finally, a correction method is designed to improve the detection performance, which is based on the mode probability of the current segment and its adjacent segments. The results of our detection model and correction method outperform state-of-the-art research in transportation mode detection.