Trip-Chain-Based Travel-Mode-Shares-Driven Framework using Cellular Signaling Data and Web-Based Mapping Service Data

Abstract

The signaling data of cellular phones, as a passively generated, real-time, wide-coverage, and low-cost data source, have been widely used in recent studies to understand human activity and model urban travel demand. However, in contrast with the Global Positioning System (GPS) data, cellular phone signaling data are sparsely distributed in time and space, which makes travel-mode inference a challenge. Recent studies presented methods of deriving users’ home and work locations, origin-destination trips, and other activities. Very few provided a complete and feasible framework for travel-mode derivation with effective validation methods. This paper provides a real-time travel-mode derivation framework using signaling data and a web-based mapping service. A trip-chain model is proposed to detect individual activity patterns and derive the trips of mobile phone users. Then, the travel mode of each trip is identified by a Fuzzy K-Means model, which is trained and validated by the point-to-point travel time from a web-based mapping service. Finally, the travel-mode shares are aggregated and scaled to the whole population of the study area. The framework is demonstrated using cellular signaling data from 1.9 million users in Shanghai, China for seven days, and citywide point-to-point travel times from a web-based mapping service for three of those seven days. Comparing the modeled travel-mode shares with travel survey data and transportation hub statistics demonstrates the plausibility and efficiency of using a large data source (mobile-trace data and web-based mapping) to accurately assess the travel modes of people in a big city using the proposed framework.