Traffic congestion prediction using machine learning: Amman City case study

Abstract

Traffic congestion leads to wasted time, pollution, and increased fuel usage. Traffic congestion prediction has become a developing research topic in recent years, particularly in the field of machine learning (ML). The evaluation of various traffic parameters is used to predict traffic congestion by relying on historical data. In this study, we will predict traffic congestion in Amman city specifically at Northbound Street at 8th Circle using different ML classifiers. The 8th Circle links four main streets: Westbound, Northbound, Eastbound, and Southbound. Because of the heavy traffic on Northbound, it was chosen to predict traffic congestion. Datasets were collected from the greater Amman municipality hourly based. Datasets were divided and per-process to be understandable. The Logistic Regression, Decision Tree (DT), Random Forest, and multilayer perceptron (MLP) classifiers have been chosen to predict traffic congestion at Northbound Street linked with the 8th Circle. Four experiments were run under The Waikato Environment for Knowledge Analysis (WEKA) tools to find the best classifiers to predict traffic congestion at Northbound. The chosen classifiers have been evaluated using F-Measure, sensitivity, precision, and accuracy evaluation metrics. The obtained results from all experiments have demonstrated the Logistic Regression is the best classifier to predict traffic congestion. The accuracy of Logistic Regression to predict traffic congestion at Northbound Street was 97.6%.