Date of Publication :7th February 2025

Abstract: Air pollution poses a significant threat to global health and the environment. This study investigated air quality in Tirumala, a major pilgrimage destination in India, emphasizing the importance of clean air for the millions of visitors it receives annually and the preservation of its cultural and ecological integrity. Advanced machine learning techniques, Artificial Neural Networks (ANN) and Adaptive Neuro Fuzzy Inference System (ANFIS) were employed, to predict concentrations of key air pollutants: Nitrogen Dioxide (NO2), Ozone (O3), Particulate Matter (PM10) and (PM2.5). The models were developed and implemented using MATLAB 2023a, with a focus on optimizing model architecture by varying the number of neurons and hidden layers in the ANN and exploring different membership functions within the ANFIS framework. Model performance was rigorously evaluated using statistical metrics, specifically the coefficient of determination (R2) and the Root Mean Square Error (RMSE), to ensure accuracy and reliability of the predictions. The study compared the predictive capabilities of both ANN and ANFIS, identifying the best-performing model for each pollutant. Results indicated that both ANN and ANFIS successfully predicted pollutant concentrations, with ANFIS demonstrating a slight advantage in predictive accuracy. Crucially, the predicted levels of all measured pollutants remained within the permissible limits set by India's National Ambient Air Quality Standards (NAAQS). ANN achieved R2 values of 0.9390, 0.9387, 0.9234, 0.9114 and RMSE values of 6.601, 6.741, 5.119,4.724 between observed and predicted values of NO2, O3, PM10 and PM2.5 respectively. ANFIS attained R2 values of 0.9557, 0.9435, 0.9358, 0.9269 and RMSE values of 4.439, 6.024, 4.598, 4.214 between observed and predicted values of NO2, O3, PM10 and PM2.5 respectively.

Reference :