Modeling Traffic Congestion with Spatiotemporal Big Data for An Intelligent Freeway Monitoring System

TL;DR

This paper presents a spatiotemporal big data model for predicting and analyzing traffic congestion on Taiwan's freeways, utilizing large-scale data processing and revealing patterns influenced by time, direction, and holidays.

Contribution

It introduces a novel spatiotemporal modeling approach using big data and MapReduce for intelligent freeway traffic monitoring in Taiwan.

Findings

Traffic flow is strongly affected by direction, time of day, and holidays.

Recurring weekly traffic patterns are identified.

Spatiotemporal data can inform AI-driven traffic management systems.

Abstract

Traffic congestion is a complex, nonlinear spatiotemporal modeling problem. By collecting and analyzing a vast quantity and different categories of information, traffic flow, and road congestion can be predicted and controlled on an intelligent transportation system. This report provides an analysis of traveling time across Taiwan from North to South, vice versa. We analyze traffic in a national freeway between Tainan and Kaohsiung section, which represents the common trip of the population in Southern Taiwan. The data is recorded using the Electronic Toll Collection System (ETC) provided by Ministry of Transportation in Taiwan. We use MapReduce framework to process data into a smaller task which can be distributed on several computer clusters to speed up the process. The results show that the spatiotemporal model of traffic flow is strongly influenced by direction, working hour, and holidays with a recurring pattern for each week. The distinctive pattern inside the spatiotemporal dataset can be used on an AI-powered decision-making system for future development.