Determining Bus Stop Locations using Deep Learning and Time Filtering
Keywords:bidirectional LSTM, bus stop determination, convolutional neural network, deep learning, global positioning system
This paper presents an intelligent bus stop determination from bus Global Positioning System (GPS) trajectories. A mixture of deep neural networks and a time filtering algorithm is used in the proposed algorithm. A deep neural network uses the speed histogram and azimuth angle at each location as input features. A deep neural networks consists of the convolutional neural networks (CNN), fully connected networks, and bidirectional Long-Short Term Memory (LSTM) networks. It predicts the soft decisions of bus stops at all locations along the route. The time filtering technique was adopted to refine the results obtained from the LSTM network. The time histograms of all locations was built where the high potential timestamps are extracted. Then, a linear regression is used to produce an approximate reliable timestamp. Each time distribution can be derived using data updated at that time slot and compared to a reference distribution. Locations are predicted as bus stop locations when timestamp distributions close to the reference distributions. Our technique was tested on real bus service GPS data from National Science and Technology Development Agency (NATDA, Thailand). The proposed method can outperform other existing bus stop detection systems.
Authors who publish with Engineering Journal agree to transfer all copyright rights in and to the above work to the Engineering Journal (EJ)'s Editorial Board so that EJ's Editorial Board shall have the right to publish the work for nonprofit use in any media or form. In return, authors retain: (1) all proprietary rights other than copyright; (2) re-use of all or part of the above paper in their other work; (3) right to reproduce or authorize others to reproduce the above paper for authors' personal use or for company use if the source and EJ's copyright notice is indicated, and if the reproduction is not made for the purpose of sale.