Travel Time Prediction on Long-Distance Road Segments in Thailand

Applied Sciences

Surasee Intarawart¹  Rathachai Chawuthai¹  Nachaphat Ainthong¹  Niracha Boonyanaet¹  Agachai Sumalee² 

[1] School of Engineering, King Mongkut’s Institute of Technology Ladkrabang, Bangkok 10520, Thailand;School of Integrated Innovation, Chulalongkorn University, Bangkok 10330, Thailand;

关键词: GPS data analytics;    machine learning;    smart mobility;    travel time prediction;   

DOI  :  10.3390/app12115681

来源: DOAJ

【 摘 要 】

This study proposes a method by which to predict the travel time of vehicles on long-distance road segments in Thailand. We adopted the Self-Attention Long Short-Term Memory (SA-LSTM) model with a Butterworth low-pass filter to predict the travel time on each road segment using historical data from the Global Positioning System (GPS) tracking of trucks in Thailand. As a result, our prediction method gave a Mean Absolute Error (MAE) of 12.15 min per 100 km, whereas the MAE of the baseline was 27.12 min. As we can estimate the travel time of vehicles with a lower error, our method is an effective way to shape a data-driven smart city in terms of predictive mobility.

【 授权许可】

Unknown