An integrated method for inferring multimodal travel mode choices using mobile network data

With the high coverage of mobile network data, the travel patterns of urban populations can be studied on a large scale at a relatively low cost. Existing research has primarily focused on inferring single modes of trips, ignoring the transitions between different transport modes within trips. This study integrates mobile signaling data with travel surveys, transport network data, and census data to infer multimodal travel choices. We first develop an adaptive distance-based clustering method to dynamically segment data into trips based on the surrounding built environment. Then, we utilize the Bayesian inference and hidden Markov models (HMM) with multiple observation sequences, effectively combining discrete and continuous observation states, to generate transport mode sequences throughout a day. We demonstrate the proposed integrated method through a case study in Nanjing, China for inferring trip chains of five transport modes. The inferred transport mode choices are extensively validated based on travel surveys, official statistical data, and smart card data at different spatial scales. From our results, we observe temporal and spatial patterns of travel for various transport modes. These findings confirm the performance of the integrated method in capturing multimodal travel patterns for an urban population. The inferred multimodal trip chains are useful for travel demand management and developing sustainable transport systems.