Using machine learning for air quality prediction and sustainable urban planning
Authors: M.A. Mujtaba, A. Munir, Sheeraz Ali, Jana Petrů, Talha Ansar, Waiz Akhlaq, Muneeb Ahmad, Haseeb Iqbal, Faisal Ali, Muhammad Nasir Bashir, T. Jerry Alexander
Overview
PM2.5 and other pollutants are a growing health risk in Lahore, and reliable forecasts are what regulators need to plan against them. This study tests how well time-series and machine learning models can predict the city's air quality.
Using data from January 2003 to December 2022 — eight pollutants and four weather factors — it compares SARIMA, SARIMAX, LSTM and nonlinear autoregressive (NAR) models, scored with root mean squared error and dynamic time warping.
The best model is used to project air quality forward, and the paper closes with mitigation strategies for regulators, framed around the Sustainable Development Goal on good health and well-being.
Key Findings
- Twenty years of data for Lahore, Pakistan: eight pollutants and four weather-related factors.
- SARIMA, SARIMAX, LSTM and NAR models compared on RMSE and dynamic time warping (DTW).
- NAR performed best, with an RMSE of 23.52 and a DTW of 5023.
- The air quality index is projected to rise by 13% by 2030 against the 2022 baseline.
- Mitigation strategies for regulators, aligned with the SDGs.
This is our summary of the paper. The full text, with its abstract, figures and data, is on the publisher’s site.
More Papers
Solar Irradiance Forecasting Meta-review: An In-Depth Systematic Meta-review on Solar Irradiance Forecasting Datasets, Predictive Pipelines and Performance
A systematic meta-review of solar irradiance forecasting datasets, modelling pipelines, and predictive performance.
Progress in Additive Manufacturing · 13 Mar 2026Machine learning and physics based empirical equations for fatigue life prediction of AM Ti-6Al-4 V
Fatigue-life prediction combining data augmentation, ANN modelling, SHAP analysis, and physics-guided empirical equations.