Leveraging deep learning and ensemble learning for air quality forecasting in smart urban environment

Urban pollution has become a significant issue for the whole world, specifically for underdeveloped nations. This pollution poses significant challenges to public health, economic stability and environmental sustainability. The rapid growth of urbanization and industries, and inadequate regulatory frameworks has led to the deterioration of air, contamination of water and soil pollution. Major urban centers such as Lahore remain at the top among the most polluted cities, globally, with adverse effects such as rising respiratory diseases, contaminated water supplies and environmental degradation. The countries have proposed various policies and regulatory framework; however, these attempts do not reverse the trend of exacerbating urban pollution due to the lack of monitoring and measurable goals. This research proposes deep learning and ensemble learning approach to track pollution levels efficiently that could be utilized for policymaking and governance, supporting real time monitoring and data driven interventions. The findings indicate decision tree and random forest gave the most reliable and accurate air quality prediction, achieving an accuracy of 0.99 and 0.98, respectively, for particulate matter 2.5 (PM2.5) and particulate matter 10 (PM10), with high precision in classification across all categories. The smog-predict app has been made available via a user-friendly webserver at: https://smog-pred.streamlit.app .