Deep Feedforward Air Pollution Classification Empowered by Sophia-G Optimization

Air pollution is a global top-burning issue that carries significant health consequences as well as environmental deterioration. The stratification of air pollution according to its levels is crucial to public health policy and administration and sound environmental policymaking. This paper introduces a Deep Feedforward Neural Network (DFFN) architecture to classify air pollution into various levels based on the concentration of major atmospheric pollutants such as carbon monoxide (CO), nitrogen dioxide (NO2), ozone (O3), and fine particulate matter (PM2.5). The model preprocesses the pollutant data using a strong preprocessing pipeline of missing value imputation, normalization, and label encoding so that the data is consistent and of high quality. The input pollution was then categorized into fixed stages: Good (1), Satisfactory (2), Moderate (3), Poor (4), and Severe (5). For training and generalization of the model, it was fine-tuned by using a new second-order optimizer called Sophia-G, which unites curvature approximation and adaptive gradient methods. Sophia-G has faster convergence and better performance than traditional optimizers. Experimental results validate that the new DFFN model attains high classification precision and robustness and provides an intelligent and scalable solution for real-time air pollution monitoring and smart environmental applications.