Traffic Jam Detection Using Real-Time Bus Operation Data Considering Timetable Information in Various Conditions

Nozomi Hatanka, Hiroki Aoyagi, Tomoya Fujita, Hayato Yamana, Masato Oguchi

Abstract: Because of the significant losses caused by traffic congestion, the detection of traffic congestion is an urgent issue for reducing such losses. In this study, we propose a model that performs a binary classification of two consecutive bus stops as one section, using the bus speed calculated from the bus departure time, the time of day, and the difference between the time the bus actually leaves the bus stop and the time it arrives. Two types of learning were performed, one with a single learner in the system and one with a single learner each time period, with better results when one learner was placed in the system.