Prediction of Air Pollutant Exposure in Traffic Environment using Machine Learning Techniques : A case study of Ibadan, Nigeria.
Ojuolape, Olaide Elijah (2024)
2024
All rights reserved. This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-202604146477
https://urn.fi/URN:NBN:fi:amk-202604146477
Tiivistelmä
Having access to air purification is essential for human survival, and air pollution is a key danger to both. Air pollution, mostly from industrial activity and car emissions, has grown due to the expansion of industry, migration, and vehicles. The primary source of air pollution is human activity, particularly the combustion of fossil fuels in homes and businesses, as well as from vehicles and natural disasters. Toxic chemicals and particle matter from pollution pose serious threats to human health. The target of this research is to predict traffic-related air pollution exposure using ML methods and to determine if there is a correlation between the study area's traffic volume and air pollution levels.
Exhaust, crankcase, and evaporative emissions are some of the many ways in which a car produces pollution while it's running. Between 65 and 70 percent of all air pollution comes from vehicle emissions. The health hazards associated with air pollution are especially substantial for youngsters, especially those who are exposed to greater quantities of contaminants. Solvents and motor fuels release volatile organic compounds (VOCs) into the environment in the form of combustion byproducts or unburned fuels in the form of harmful exhaust fumes. Acid rain and plant tissue death are two of the many worldwide consequences of air pollution. For financial reasons, it encompasses, among other things, metal corrosion.
Using three different ML models, this research attempts to forecast CO and O3 concentrations in the air. Because it includes both strong and weak leaners, Random Forest, an ensemble model, outperformed its competitors. One of the most important variables in determining the amount of air pollution is the volume of traffic. When compared to other methods, support vector regression performed the worst, with the greatest error rates and R-squared values. The models increased in performance after optimization and hyper-parameter adjustment. The PM2.5 model, which included several machine learning algorithms, provided a reliable and accurate assessment of Ibadan's pollution levels. The research shows that air quality projections may be made using nonlinear machine learning methods.
Exhaust, crankcase, and evaporative emissions are some of the many ways in which a car produces pollution while it's running. Between 65 and 70 percent of all air pollution comes from vehicle emissions. The health hazards associated with air pollution are especially substantial for youngsters, especially those who are exposed to greater quantities of contaminants. Solvents and motor fuels release volatile organic compounds (VOCs) into the environment in the form of combustion byproducts or unburned fuels in the form of harmful exhaust fumes. Acid rain and plant tissue death are two of the many worldwide consequences of air pollution. For financial reasons, it encompasses, among other things, metal corrosion.
Using three different ML models, this research attempts to forecast CO and O3 concentrations in the air. Because it includes both strong and weak leaners, Random Forest, an ensemble model, outperformed its competitors. One of the most important variables in determining the amount of air pollution is the volume of traffic. When compared to other methods, support vector regression performed the worst, with the greatest error rates and R-squared values. The models increased in performance after optimization and hyper-parameter adjustment. The PM2.5 model, which included several machine learning algorithms, provided a reliable and accurate assessment of Ibadan's pollution levels. The research shows that air quality projections may be made using nonlinear machine learning methods.