AI-Powered Air Quality Forecasting for Sustainable Cities: An LSTM-Based Approach Aligned with WHO/EPA Standards | ||
| JES. Journal of Engineering Sciences | ||
| Articles in Press, Accepted Manuscript, Available Online from 01 November 2025 | ||
| Document Type: Research Paper | ||
| DOI: 10.21608/jesaun.2025.373722.1473 | ||
| Author | ||
| Mohamed Laissy* | ||
| Associate Professor, Head of the Civil Engineering Program | ||
| Abstract | ||
| Air quality prediction is essential for safeguarding public health and guiding urban environmental strategies. Among various pollutants, fine particulate matter (PM₂.₅) is particularly hazardous due to its ability to penetrate deep into the respiratory tract, causing or exacerbating cardiovascular and respiratory illnesses. This work presents a Long Short-Term Memory (LSTM) neural network model that predicts PM₂.₅ by using a comprehensive multi-year dataset from many air quality monitoring stations including sulfur dioxide SO₂, nitrogen dioxide NO₂, carbon monoxide CO, ozone O₃, temperature, pressure, humidity, wind speed, and wind direction. By investigating complex temporal dependencies in time-series data, the LSTM method outperforms traditional statistical models, achieving a Mean Squared Error (MSE) of 0.015 and a Mean Absolute Error (MAE) of 0.097 in predicting PM₂.₅ levels. Moreover, these predictions are mapped to health-based categories aligned with World Health Organization (WHO) and U.S. and the Environmental Protection Agency (EPA) policies, to guarantee that the outcomes are practically applicable for decision-makers. Supported by Saudi Vision 2030 framework and the Saudi Green Initiative (SGI), this study offers an AI-driven system able to improve proactive environmental management. Recent studies confirm the increasing effectiveness of deep learning methods including GAN-based models for robust air quality forecasting and hybrid CNN-LSTM networks, so stressing the ongoing creativity in this field. | ||
| Keywords | ||
| Deep learning; sustainability; air quality prediction; LSTM; WHO | ||
|
Statistics Article View: 5 |
||
