13. COMPARATIVE PERFORMANCE OF LINEAR REGRESSION AND NEURAL NETWORK IN FORECASTING AIR QUALITY INDEX (AQI) IN HANOI

Abstract

This study compares the forecasting performance of two machine learning models, namely linear regression and artificial neural networks, in predicting the Air Quality Index in Hanoi based on key pollution indicators, including PM_2.5, NO₂, SO₂, CO, and O₃. The dataset was expanded to improve model training and evaluation. Updated results show that neural networks, when properly optimized, achieve higher accuracy and greater stability compared to linear regression. Model performance was assessed using RMSE, MAE, and R². These findings suggest that nonlinear modeling approaches hold significant potential for environmental forecasting while maintaining a balanced comparison with traditional methods.