Road traffic predictions across major city intersections using multilayer perceptrons and data from multiple intersections located in various places

Authors:
Open Access
Publisher website
Google Scholar
Abstract
This study presents a novel approach to the road traffic prediction using single multilayer perceptrons and their ensemble. Networks were trained on the basis of real‐world data from the intelligent transportation system Wroclaw. This system is installed in one of the largest cities in Poland. First, a number of neural networks were created, each of which was concurrently able to predict the state of traffic on a number of major intersections located in different parts of the city. Then the multilayer perceptrons were made, which predict the numbers of vehicles passing through selected intersections using the information about previous situations at other intersections. Furthermore, an ensemble method, which combine output values of multiple neural networks, were applied. In the worst case, mean absolute percentage error did not exceed 12.6%, even in cases when traffic prediction was based only on information from other intersections.
Funding Information
  • Wuhan University of Technology (S50242)

This publication has 20 references indexed in Scilit: