Optimized Feature Selection for Air Quality Index Forecasting using GPR and SARIMA Models

Abstract

The importance of predicting and forecasting the Air Quality Index (AQI) using machine learning approaches is growing due to the negative health impacts of air pollution. This study proposes the use of Gaussian Process Regression (GPR) and Seasonal Autoregressive Integrated Moving Average (SARIMA) for AQI prediction and forecasting. GPR is a potent probabilistic model capable of representing complicated nonlinear interactions among input variables and output variables. SARIMA is a popular time series model that can capture seasonal and trend patterns in AQI data. The proposed approach was evaluated on real-world AQI data from a major city in India. The results showed that the GPR and SARIMA models outperformed traditional statistical models in predicting AQI levels. The study highlights the potential of machine learning techniques for AQI prediction and forecasting and demonstrates the importance of accurate air quality monitoring and management.