Traffic Speed Prediction Using Machine Learning: in Case of Addis Ababa City
No Thumbnail Available
Date
2023-10
Authors
Teka Mosisa
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
The demand for mobility and transportation in Addis Ababa is growing along with the city's population expansion. Even though Addis Ababa City is having low motorization rates by global standards, the challenge of traffic congestion is get worse by the city's growing urban and industrial populations, which demand more vehicles. Studies shows that Ethiopia has an unreliable mobility traffic control and management system. Nevertheless, building infrastructure in accordance with international standards is a challenging and long-term solution due to the state of the economy and organizational difficulties.
Analysis of traffic speed and its variables is crucial given the effects of traffic congestion on the economy and other aspects. In order to effectively utilize the limited resources in Addis Ababa City, traffic speed prediction, one of the elements of an intelligent transportation system, will be a solution. The Addis Ababa Traffic Management Agency has only performed manual traffic count flow study of traffic congestion at intersections but actual traffic movement patterns are dynamic and route-based. Various sectors employ machine learning techniques, including online transportation networks and traffic prediction. This study aimed to develop machine learning models that predict the road traffic speed of Addis Ababa City. ML methods use less time and computer power and are considered to be a viable model for the presented dataset.
Sheger Public Bus Transportation traffic dataset which compiled and published by Addis Ababa Science and Technology University was used for this study. Five regression models were used to train and test the data. To find the most effective method of predicting the speed of traffic on a city-wide scale, three scenarios, including model parameters, feature importance identification, and hyper-parameters tuning, were tested. With consistent performance across all scenarios and scoring values Random Forest (RF) achieved best performance and after parameter tuning other ensemble algorithm Gradient Boosting achieved best performance.
Description
Keywords
Machine Learning, Traffic Speed Prediction, Regression Model, Intelligent Transport System, Random Forest