Comparison of Effective Machine Learning Technique for Air Quality Forecast

Air pollution is a complex mixture of toxic components with a considerable impact on living beings. Many people around the world live in places where constant air pollution exposes them to higher risks. Among the particles present in the air, the particulate matter PM2.5 and PM10 are considered to be the most dangerous causing pulmonary cancer. Other pollutants include CO, NO2, SO2, NH3, lead, and ground-level O3 also show adverse effects on humans. It is possible to forecast air quality that helps people plan ahead, decreasing the effects on health and the costs associated. Air pollution monitoring is receiving increasing attention from researchers all over the world. It is the key to assuring public health. The air quality index (AQI) measures the quality of air under six levels. By calculating AQI, it is easy to interpret the quality of air and is easily understandable. Machine learning is able to efficiently train a model on big data by using effective algorithms. Out of various machine learning models, random forest of ensemble methods is best suitable for prediction due to its processing time with minimum error rate. The major work in this paper is to predict the AQI based on air pollutant levels and meteorological data by implementing random forest against SVM and decision tree machine learning models. The outcome of the project shows the classification of AQI that has numeric value, AQI levels of health concern, AQI color indicator, and corresponding level of health concern.