Travel Time Prediction Using Multi-layer Feed Forward Artificial Neural Network

Traffic jam is a major problem in Bangkok and nearby provinces in Thailand. Currently, there have been several attempts to solve this elevating problem by using GPS together with GPRS technologies in tracking and collecting traffic data from vehicles. In this work, we obtained one-month records of GPS data from 297 volunteered vehicles. Using vehicles' velocity as input, we have developed a travel time prediction model using artificial neural network. However, due to the enormous amount of database, we focus on testing our model on a certain major road, inbound of Hwy35 or Thonburi-Paktor. We applied our ANN model in predicting travel time during rush-hour traffic in the morning/evening and non-rush hour traffic on the weekday and weekend. The predicted results from the proposed model can accurately approximate the actual travel time. Furthermore, the predicted travel time during non-rush hour data set are very close to the predicted travel time provided by GoogleMap.