Development of Machine Learning Model for Climatic Impact Prediction on Health Using AQI Dataset in Africa

Abstract

Air pollution remains one of the major concerns of health crisis in most developing countries with Africa facing a silent health crisis as air pollution worsens, yet predictive tools remain scarce. Pollutants such as PM2.5, NO₂, CO, and O₃ increase the risk of respiratory and cardiovascular diseases. This study develops a Machine Learning (ML) model to predict the Air Quality Index (AQI) and assess health risks across urbanized, industrialized, and rural regions using climatic parameters. A quantitative approach was applied to 23,463 AQI datasets obtained from Kaggle World AQI database. The data was pre-processed and feature engineering was used to remove null values and outliners then splitted into ratio 70:30 for training and testing. Four algorithms namely; Linear Regression, k-Nearest Neighbours, Decision Tree and Random Forest were evaluated using metrics such as R-Squared (R2), Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE). The result shows that Random forest model (R2 = 0.997316, RMSE = 2.865823, MAE = 0.2955499) demonstrated superior predictive performance followed by KNN (R2 = 0.996820, RMSE = 3.119500, MAE = 0.588252) while Decision Trees (R2 = 0.995046, RMSE = 3.893819, MAE = 0.302845) produced high accuracy with slightly higher error. SVR (R2 = 0.980160, RMSE = 7.792268, MAE = 1.301302) and Linear Regression (R2 = 0.975279, RMSE = 8.968122, MAE = 4.831951) showed moderate accuracy. This research confirms that Machine Learning models are valuable tools for predicting Air quality thus offering a powerful tool for mitigating the impact of deteriorating Air Quality in Africa.