Trip mode inference from mobile phone signaling data using Logarithm Gaussian Mixture Model

Abstract

Trip mode inference plays an important role in transportation planning and management. Most studies in the field have focused on the methods based on GPS data collected from mobile devices. While these methods can achieve relatively high accuracy, they also have drawbacks in data quantity, coverage, and computational complexity. This paper develops a trip mode inference method based on mobile phone signaling data. The method mainly consists of three parts: activity-nodes recognition, travel-time computation, and clustering using the Logarithm Gaussian Mixed Model. Moreover, we compare two other methods (i.e., Gaussian Mixed Model and K-Means) with the Logarithm Gaussian Mixed Model. We conduct experiments using real mobile phone signaling data in Shanghai and the results show that the proposed method can obtain acceptable accuracy overall. This study provides an important opportunity to infer trip mode from the aspect of probability using mobile phone signaling data.