Investigating spatial-temporal characteristics of joint activity/travel behaviour with smart card data

The process of urbanization and the rapid development of transit networks transform urban land use properties and individuals' travel demand, especially in densely populated areas. Traffic surveys reveal that joint activity/ travel behaviour constitutes an increasingly significant portion of overall travel behaviours. This paper aims to investigate the spatial-temporal characteristics of individuals' joint activity/travel patterns (JATPs) within the subway network and explores the factors that influence individuals' choices among different activity/travel patterns. First, we establish a social relationship classifier to infer acquaintances among individuals. Subsequently, we define four distinct types of JATPs and introduce an algorithm for recognizing these patterns. To implement the proposed framework, we conduct a real-world case study based on subway smart card data from Nanjing, China. Our findings reveal that approximately 19% of daily trips were identified as JATPs, while different JATPs exhibit diverse spatial and temporal characteristics. These differences can be attributed to various influencing factors, including travel/activity time, passenger flow, departure time, and passenger flow/ income around tap-in/out stations. In addition, we employ the machine learning classification method XGBoost to predict the discrete choice process of JATP. The classification results reveal the spatial-temporal characteristics of JATP and examine the potential influences of various factors on individuals' joint activity and travel behaviours. Ultimately, our research sheds light on the evolving landscape of urban travel behaviour in the context of joint activities and travel, with practical implications for transportation planning and policy.