Predicting Health Impact Based on Air Quality Using Artificial Intelligence

Abstract

Air pollution is continuously increasing it is becoming a major issue that impacts on human health and environment. Air quality is degrading and is becoming worse day by day due to air pollutants such as particulate matter (PM2.5, PM10), nitrogen oxides (NO2), sulfur dioxide (SO2), ozone (O3), weather factors (temperature, humidity, windspeed), health features (hospital admissions, respiratory cases, cardiovascular cases) which are affecting human health. This study proposes an artificial intelligence (AI) framework which applies machine learning (ML) techniques for health impact class classification. Feature engineering played a very important role; the original dataset was integrated with weather parameters to capture air pollutant effect on weather factors and analyze overall health impact. Binning method and one-hot encoding was applied to address class imbalance problems. Multiple AI-based machine learning algorithms like KNN, decision tree, random forest, XGBoost, AdaBoost are implemented with hyperparameter tuning and a 5-fold GridSearchCV. Among all models XGBoost achieved higher accuracy of 93% with a macro F1-score of 0.71. The proposed model efficiently classifies health impact class based on severity of air pollution.