Building workers' travel demand models based on mobile phone data

Abstract

Daily activity-travel sequences of individuals have been estimated by activity-based transportation models. The sequences serve as a key input for travel demand analysis and forecasting in the region. However, the high cost along with other limitations inherent to traditional travel data collecting methods has hampered the models’ further advancement and application, particularly in developing countries. With the wide deployment of mobile phone devices today, we explore the possibility of using mobile phone data to build such a travel demand model. Our exploration consists of four major steps. First, home, work and other stop locations for each user are identified, based on their mobile phone records. All the obtained locations along with their particular orders on a day are then formed into stop-location-trajectories and classified into clusters. In each cluster, a Hidden Markov Model (HMM) is subsequently constructed, which characterizes the probabilistic distribution of activities and their related travel of the sequences. Finally, the derived models are used to simulate travel sequences across the entire employed population. Using data collected from natural mobile phone usage of around 9 million users in Senegal over a period of one year, we evaluated our approach via a set of experiments. The average length of daily sequences drawn from the stop-location-trajectories and the simulated results is 4.55 and 4.72, respectively. Among all the 677 types of the stop-location-trajectories, 520 (e.g. 76.8%) are observed from the simulated sequences, and the correlation of sequence frequency distribution over all the types between these two sequence sets is 0.93. The experimental results demonstrate the potential and effectiveness of the proposed method in capturing the probabilistic distribution of activity locations and their sequential orders revealed by the mobile phone data, contributing towards the development of new, up-to-date and cost-effective travel demand modelling approaches.