Air Quality Prediction System Using Machine Learning

This paper presented the design, implementation, and performance analysis of the Air Quality Prediction System—a Django-based full-stack web application integrating a Random Forest machine learning engine with a multi-pollutant monitoring and prediction platform. The ML pipeline achieves 99% training accuracy and 87% test accuracy with a weighted F1-score of 0.812. All fifteen functional test cases passed with 100% correctness across authentication, prediction validation, admin operations, and route protection. Three key design decisions —Random Forest selection, hybrid regressor-classifier design, and module-level model loading—were analytically validated through performance measurements and literature comparison. The system successfully bridges the gap between isolated ML research on AQI datasets and deployable, user-accessible environmental intelligence platforms. Future implementation directions include: (i) hyperparameter tuning of the Random Forest using GridSearchCV or RandomizedSearchCV to reduce the training-test accuracy gap from 12 to under 5 percentage points; (ii) integration of real-time IoT sensor APIs replacing static CSV-based data ingestion with live multi-station streams; (iii) implementation of LSTM-based time-series models for sequential AQI forecasting, providing 24–72 hour forward predictions beyond the current point-in-time estimate; (iv) Android and iOS mobile application consuming the Django REST API via DRF (Django REST Framework) for broader citizen accessibility; (v) geo-location-based push notifications alerting users when AQI at their nearest station crosses personalised threshold levels; and (vi) SHAP-based feature importance visualisation exposing which pollutant drove each individual prediction for greater transparency and user trust.

[1] Central Pollution Control Board (CPCB), India, "National Air Quality Index," [Online]. Available: https://cpcb.nic.in/. [Accessed: May 2025]. [2] World Health Organization (WHO), "Ambient Air Pollution: A Global Assessment of Exposure and Burden of Disease," WHO Press, Geneva, Switzerland, 2022. [3] A. Kumar, "Prediction of Air Quality Using Machine Learning Techniques: A Review," Int. J. Computer Science and Mobile Computing, vol. 13, no. 3, pp. 1–10, 2024. [4] H. Bai, G. Wang, and X. Li, "Air Quality Prediction Using Machine Learning Methods Based on Monitoring Data and Secondary Modelling," Atmosphere, vol. 13, no. 2, p. 212, 2022. [5] S. Rybarczyk and R. Zalakeviciute, "Machine Learning Approaches for Outdoor Air Quality Modelling: A Systematic Review," Applied Sciences, vol. 8, no. 12, p. 2570, 2018. [6] L. Zhang, P. Liu, L. Zhao, G. Wang, W. Zhang, and J. Liu, "Air Quality Predictions with a Semi-Supervised Bidirectional LSTM Neural Network," Atmospheric Pollution Research, vol. 12, no. 1, pp. 328–339, 2021. [7] F. Pedregosa et al., "Scikit-learn: Machine Learning in Python," J. Machine Learning Research, vol. 12, pp. 2825–2830, 2011. [8] Y. Song, J. Wang, Y. Wang, and P. Ding, "Air Quality Prediction Using Random Forest Regression Model with Air Quality Index," J. Big Data, vol. 12, no. 1, pp. 1–22, 2025. [9] Django Software Foundation, "Django 4.x Documentation," [Online]. Available: https://docs.djangoproject.com/. [Accessed: May 2025]. [10] Scikit-learn Developers, "RandomForestRegressor — Scikit-learn Documentation," [Online]. Available: https://scikit-learn.org/stable/. [Accessed: May 2025]. [11] Kaggle, "Indian Air Quality Dataset," [Online]. Available: https://www.kaggle.com/datasets/jayaprakashpondy/air-quality-dataset. [Accessed: May 2025].