Explainable deep learning hybrid modeling framework for total suspended particles concentrations prediction
Article
Article Title | Explainable deep learning hybrid modeling framework for total suspended particles concentrations prediction |
---|---|
ERA Journal ID | 1955 |
Article Category | Article |
Authors | Ghimire, Sujan, Deo, Ravinesh C., Jiang, Ningbo, Ahmed, A. A. Masrur, Prasad, Salvin S., Casillas-Perez, David, Salcedo-Sanz, Sancho and Yaseen, Zaher Mundher |
Journal Title | Atmospheric Environment |
Journal Citation | 347 |
Article Number | 121079 |
Number of Pages | 17 |
Year | 2025 |
Publisher | Elsevier |
Place of Publication | United Kingdom |
ISSN | 1352-2310 |
1873-2844 | |
Digital Object Identifier (DOI) | https://doi.org/10.1016/j.atmosenv.2025.121079 |
Web Address (URL) | https://www.sciencedirect.com/science/article/pii/S1352231025000548 |
Abstract | Total Suspended Particles (𝑇𝑆𝑃 ) is an important indicator of air quality, yet traditional prediction models lack comprehensive consideration of spatio-temporal interactions of different meteorological and air pollution phenomena. To address these limitations, this study introduces an explainable (X) deep hybrid (H) network, integrating Convolutional Neural Networks (CNN) and Bidirectional Gated Recurrent Units (BGRU), for hourly 𝑇𝑆𝑃 concentration prediction. The model was trained and evaluated using meteorological and air quality data from Canon Hill, Australia. By combining CNN’s spatial feature extraction capabilities with BGRU’s temporal dependencies, the model effectively captures complex spatial–temporal patterns in the data. The X-HCBGRU model outperforms fifteen competing benchmark models such as deep neural network, extreme learning machine, multilayer perceptron, support vector regression, random forest regression, light gradient boosting, gradient boosting regression, long short-term memory network, as well as their hybrid CNN counterparts in terms of the accuracy evidenced by a lower Root Mean Square Error (𝑅𝑀𝑆𝐸 ≈ 6.302 μg∕m3 ) and higher Correlation Coefficient (𝑟 ≈ 0.91) compared to other models. Moreover, the model demonstrates strong probabilistic performance with a high Prediction Interval Coverage Probability (𝑃 𝐼𝐶𝑃 ≈ 0.98) and low Prediction Interval Normalized Average Width (𝑃 𝐼𝑁𝐴𝑊 ≈ 0.18), indicating its reliable prediction intervals. To enhance model interpretability, Shapley Additive Explanations (SHAP) and Local Interpretable ModelAgnostic Explanations (LIME) methods were employed, revealing 𝑃𝑀10 concentration, relative humidity, air temperature, and wind speed as key predictors of 𝑇𝑆𝑃 concentrations. The Diebold–Mariano statistical test further confirmed the model’s superior performance. This study contributes towards advancing 𝑇𝑆𝑃 prediction by providing a robust, accurate, and interpretable model which has particular importance in locations such as mining regions. The X-H-CBGRU model holds potential for improving public health protection and informing air pollution mitigation strategies. |
Keywords | Air pollutant prediction; Optuna optimization; Deep learning; Bidirectional GRU; ‘‘Black-box’’ model |
Contains Sensitive Content | Does not contain sensitive content |
ANZSRC Field of Research 2020 | 5101. Astronomical sciences |
Public Notes | Files associated with this item cannot be displayed due to copyright restrictions. |
Byline Affiliations | School of Mathematics, Physics and Computing |
Department of Environment, Climate Change and Water, New South Wales | |
Fiji National University, Fiji | |
Rey Juan Carlos University, Spain | |
University of Alcala, Spain | |
King Fahd University of Petroleum and Minerals, Saudi Arabia |
https://research.usq.edu.au/item/zx13x/explainable-deep-learning-hybrid-modeling-framework-for-total-suspended-particles-concentrations-prediction
33
total views0
total downloads9
views this month0
downloads this month