Effectiveness of Public Transport Networks in Motorized Mode Detection: A Case Study of a Planning Survey in Nanjing

As an integral part of smartphone-based travel behavior research, trip mode detection has attracted the attention of many scholars who have used various methods to classify trip modes automatically. In these studies, network data and geographic information system (GIS) information on, for instance, the public transport network, have been used to promote detection accuracy. However, few studies have focused on its utility pointedly. This research collected a series of GPS trajectory data using a planning survey method and developed two models consisted with the criteria-based random forest (RF) algorithm to explore the impact of the public transport network in the comparison of automobile travel and public transport. The results show that the utility of public transport network information depends on the traffic environment. During peak hours, the public transport network can help the RF algorithm improve the accuracy of motorized mode detection nearly 6% more than that during non-peak hours. Public transport network information is a useful predictor of travel mode identification in situations where the researchers consider the instability of smartphone-based data and the diversity of the data collection environment.