Travel Time Prediction on Urban Roadways Using ANNs with Neighboring Sections Relationship

抄録

This article described the details of the travel time prediction model on urban roadways using the Artificial Neural Networks (ANNs) technique by integrating data from the neighboring road sections as the candidates for model inputs. The multilayer feedforward neural network model was selected as the main structure for the travel time prediction model. The candidate inputs for the travel time prediction model were historical travel times of the target section and its neighboring sections including; upstream, downstream and signal sharing sections. The real Bluetooth dataset obtained from BMS systems installed on urban roadway networks in Bangkok CBD was used in verifying the applicability of the proposed technique. Results indicated the proposed technique was superior to the baseline models in all test scenarios. Furthermore, in the case that data from target section was missing, the ANN(Miss) model that does not require historical data of target section could be a good solution for use as a travel time prediction model with acceptable results.