samson19977/Use-machine-learning-Early-Warning-System-EWS-to-predict-harmful-levels-of-air-pollutants-in-Rwanda

Prediction of Air Pollution Development and Validation of a Machine Learning-Based Early Warning System for Air Pollution in Rwanda," aims to create a predictive model for air quality levels in Rwanda using machine learning. Using data from the Rwanda Environment Management Authority (REMA), spanning 2020 to 2024, it analyzes pollutant data from two distinct areas: Kigali City and rural regions. The objective is to design an Early Warning System (EWS) capable of forecasting harmful air pollution levels, providing timely alerts to minimize health risks. This approach involves selecting suitable machine learning models like Random Forest, Gradient Boosting, and time-series models such as Seasonal ARIMA, Prophet, and LSTM, based on data characteristics. Cross-validation and hyperparameter tuning will enhance model accuracy. By assessing pollutants like PM10, SO2, and NO2, the EWS will predict high-risk pollution days and determine thresholds for critical alerts. The project evaluates models using metrics such as Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE) for regression, ensuring robust generalization across urban and rural datasets. Ultimately, this EWS will help Rwanda respond proactively to pollution events, enhancing public health protection, particularly in vulnerable areas.