Latent Class Analysis for Driving Behavior on Merging Section

This study proposes a modeling framework, and the complicated phenomena in merging sections are analysed with the model. The proposed modeling framework is based on the latent class model, and contains a simplified estimation method for all parameters. In the model, it is assumed that driving behaviors consist of multiple driving-modes, and the latent driving-modes are based on drivers’ intensions or surrounding circumstances.The modeling structure enables us to analysing complicated phenomena in merging sections which often becomes a bottleneck or an accident hotspot because of drivers’ inevitable lane-changing and vehicles’ complicated movements. Driving behaviors of vehicles on main-lane of a merging section are analysed and effects of information for safety driving are verified.