AI-Driven Hybrid Deep Learning Approach for Predicting Air Quality in India Using AlexNet-TabNet Architecture

Air pollution is now a major environmental issue in India, and its impact on public health and the environment cannot be ignored. Timely intervention and policy making depend on the accurate forecasting of air quality. This paper proposes to classify and analyze air pollution levels based on past AQI data from several Indian cities and also build predictive models for forecasting future trends in air quality based on hourly and daily measurements of pollutants. This study aims to analyze air quality data, classify the degree of pollution, and predict future trends using machine learning techniques. The proposed work is to obtain air quality data from the Air Quality Data in India (2015–2020) dataset and then further preprocessing operations such as normalization, removal of noise, and filling missing values. Temporal features such as day, month, hour, season, and weekday are extracted from timestamps to improve model performance. Classification is carried out using an AlexNet-TabNet hybrid deep model that leverages AlexNet's feature extraction capability and gains access to TabNet attention mechanism for tabular data. Evaluation is performed with classification metrics (accuracy, precision, recall, F1-score) and prediction metrics (R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> score, MSE, RMSE, MAE). Experimental results establish that the proposed approach provides accurate air quality classification and prediction and is useful to provide meaningful information about air quality trends and aid in better control of pollution levels in Indian cities.