Web Application Prototype on Air Quality Index Prediction, Monitoring, and Information Dissemination System: Machine Learning and Python-Streamlit-based Application Tailored for Kathmandu Metropolitan City

Abstract

Air pollution has become a critical issue in Kathmandu Valley, significantly impacting public health and daily life. Rising vehicular emissions, industrial activities, and insufficient environmental regulations have resulted in dangerously high Air Quality Index (AQI) levels. Residents frequently experience poor air quality, yet they lack easy and organized access to location-wise early AQI information and alerts, making it challenging for citizens and local governments to make informed decisions about their health and activities. Kathmandu Metropolitan lacks an integrated real-time Air Quality Index tracker and forecast-based IT application system, further highlighting the urgency of developing such an application. This paper proposes a web application prototype (Wolmann, 2023) tailored for Kathmandu Metropolitan City to predict, monitor, and disseminate information about the air quality index (AQI). The application prototype integrates three regression-based machine learning models (Random Forest Regressor, CatBoost Regressor, and Gradient Boosting Regressor) to provide accurate AQI predictions. It also includes an intuitive user interface (UI based on Python-Streamlit) that allows the admin side of KMC to fetch accurate AQI predictions/visualizations, disseminate personalized recommendations/ email alerts for residents and the federal government, and allow station-wise mapping on Kathmandu Metropolitan City’s map (using Python-Folium). Our ML Model achieved the R2 values of 0.86, 0.87, and 0.93 for the Gradient Boosting Regressor, CatBoosting Regressor, and Random Forest Regressor, respectively.