Empirical Evaluation of Deep Learning-Based Travel Time Prediction

Abstract

Travel time prediction is critical in the urban traffic management system. Accurate travel time prediction can assist better city planning and reduce carbon footprints. In this paper, we conducted an empirical work on deep learning-based travel time prediction. The objective of this study is to compare the prediction performance of different machine learning methods. Meanwhile, through the comparison, a neural network module with high prediction accuracy can be offered for alleviating traffic congestion. In addition, to eliminate the influence of nonlinear external factors, a variety of extrinsic data with abrupt properties will be acquired in real time and become part of the research considerations.