MULTI-BRANCH DEEP LEARNING BASED TRANSPORT MODE DETECTION USING WEAKLY SUPERVISED LABELS

Abstract. Mobility data, based on global positioning system (GPS) tracking, have been widely used in many areas. These include, but not limited to: user direction guidance, analyzing travel patterns, and evaluating impacts. Transport Mode Detection (TMD) is an essential factor understanding mobility within the transport system. A TMD model assigns a GPS point or trajectory to particular mode user’s current activity. However, complexity of prediction procedure increases with number modes that need be predicted given increasing overlaps feature values between multiple transportation modes. Hence, this study proposes two-branch deep learning-based predicts multi-class improve accuracy. In addition, it proposed weakly supervised labelling using snorkel volume labelled data resulting We considered publicly available road networks, railway bus routes, etc., for creating road, bus, train labels by overlaying points these networks. introduced boolean (true/false) soft-labelling function, where same overlaid network. The raw were generate point-level features such as speed, speed difference, acceleration, acceleration initial bearing all derived model. To construct we opted use two branches latitude longitude one are other.