Comparative Machine Learning Analysis of PM2.5 and PM10 Forecasting in Albania

Air pollution continues to be a critical health risk factor for populations worldwide. Tirana, the capital city of Albania with a high population density, confronts significant environmental challenges, particularly concerning air quality, which are being increasingly assessed Internet of Things (IoT) technologies and sensors. Previous research has utilized IoT system to gather air quality time series data (particulate matters, temperature, humidity, air pressure) during 24 - 28 September 2019 and using only random forest regression model for predictions of each parameter. This paper extends the research by conducting a comprehensive comparative analysis of four machine learning models, Long Short-Term Memory (LSTM), Random Forest Regression (RFR), Gradient Boosting Regression (GBR), and Support Vector Regression (SVR), to predict particulate matter (PM2.5 and PM10) levels. The novelty of this study lies in its multi-model approach, using the same dataset with six attributes, to determine the most effective model for air quality forecasting. Evaluation metrics such as Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and R-squared score are utilized to assess the models’ performance. The results demonstrate that the Random Forest model outperforms others in terms of accuracy and efficiency, providing a robust solution for air quality prediction in urban settings.