28 November 2024 in IEEE Journal of Biomedical and Health Informatics
Countries all around the world recognise the impact of air quality on public health, advocating for city centre decarbonisation and pollutant monitoring via Internet of Things technologies. Using data collected from these systems, it is possible to generate models that predict pollution based on regular patterns where mobility data can enhance the accuracy and robustness of these advanced machine learning models. This paper follows this approach, utilising vehicle traffic data from image recognition, on-site vehicle detectors, and synthetic data to maximise prediction accuracy in various urban environments. The results reveal that this proposal improves prediction for traffic-related pollutants, such as $SO_{2}$ and $PM_{2.5}$ , which are linked to severe respiratory diseases. These results also highlight the role of synthetic data in enhancing prediction performance under limited datasets.
3 April 2025 in Applied Soft Computing
4 March 2026 in IEEE Access
In air-pollution studies, accurate estimation of pollutant levels at unmonitored locations is essential due to the spatial mismatch between fixed monitoring stations and experimentation. When dealing with multi pollutant datasets, Principal Component Analysis (PCA) is commonly applied to extract a low rank structure, which serves as a prior reducing dimensionality and improving the accuracy of subsequent spatial prediction models. However, both classical interpolation techniques and recent deep learning approaches fail to provide reliable estimates under conditions of data sparsity and low spatial sampling rates. To address this, we present a fast and accurate infinitely wide convolutional neural network approach to perform matrix imputation for air pollutant estimation in an urban environments. Speed comes from the convolutional neural tangent kernel computed from the infinite width network whose training dynamics can be fully described by a closed form formula. Experiments on two real world air pollution datasets demonstrate that our strategy accurately estimates from uneven data and significantly outperforms traditional interpolation methods.