A Novel Traffic Prediction System based on Floating Car Data and Machine Learning

What is it about?

This paper presents a novel system capable of forecasting the traffic in urban road networks. This study aims to analyze traffic patterns based on Floating Car Data using a supervised learning approach. The prediction of the mean travel time and the mean waiting time is used to evaluate the proposed method with the R-Square evaluation metric for Lasso and residuals. We also evaluate the prediction performance by reducing the number of connected vehicles. Based on these evaluations, the proposed system can accurately forecast the traffic conditions in the urban road networks, only with 10% of connected vehicles. Such a system presents a technical means for road managers to understand travel patterns better and helps authorities to make strategic decisions.

Featured Image

Photo by Denys Nevozhai on Unsplash

Photo by Denys Nevozhai on Unsplash

Why is it important?

The successful deployment of an intelligent transportation system relies significantly on a good understanding of traffic patterns. Furthermore, urban traffic flow is a complex problem because of its large number of parameters. For example, traffic congestion occurs suddenly and may be caused by unexpected events (e.g. accidents, road works, bad weather conditions, etc.). Two major challenging problems should be evoked to design a traffic prediction system in urban road networks. The First one consists of the data collection technologies that will be used to gather accurate traffic data for a wide-area road network. The second one is the data analysis approach that will perform a better recognition of the traffic patterns and improve the traffic prediction.

Perspectives