In general, our algorithm is a supervised learning mechanism with
two stages. In stage 1 (learning stage), the data from the GPS
sensor report is merged with the transportation network data and
labeled ground truth. This data is used to create a classification
feature set that we use to train our classification model. In this
stage, mobile devices submit GPS sensor reports every t seconds,
where t is a system parameter. These incoming sensor reports are
labeled with the corresponding transportation modes