An expert system to discover key congestion points for urban traffic

Abstract

Discovering key congestion points periodically in traffic jams is a critical issue. It supports road managers to make sense of the situations, and rule out the congestion economically and efficiently. However, city-scale and synchronal traffic data bring hardships for such kind of analyses. With recent developments in data science, the availability of traffic conditions data generated by the rising digital map applications makes this issue feasible. Therefore, we firstly propose a digital map data-driven expert system to discover and measure the city-scale key congestion points. It is based on a state-of-the-art feature selection method, BSSReduce (Bijective soft set based feature selection). Data from Baidu Map for Chongqing and Beijing are collected as a case to conduct this study. The results indicate that our proposed method helps the road managers recognize 75 and 300 key congestion points from over 10,000 and 50,000 points of the urban roads each month. The visualized results, as well as the significance measurements, provide road managers an expert system to quickly rule out congestion and work out new solutions to future traffic management.