EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC 4.0

AQ-FUSIONNET: Dual-Decomposition Attention-GRU Framework with Adaptive Optimization for Air Quality Forecasting

Jibin Joy, A Somasundaran, Meena Jose Komban, Raji N, Keerthika M

Environment Conservation Journal · 21.05.2026 · 27 (2) · 656-670

DOI: 10.36953/ecj.38793294 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Environment Conservation Journal, online erschienen am 21.05.2026. Verfasst von Jibin Joy und 4 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Air quality prediction has grown to be an essential component of sustainable urban management due to augmented health and environmental impacts of pollution. Nonlinear, non- stationary, and multi-temporal dependencies are however not in most cases capturable in the dynamics of air pollutants by currently existing predictive models. To overcome these shortcomings, in this work it has been suggested AQ-FusionNet, a new dual-decomposition attention-based hybrid intelligence model of precise and explainable air quality prediction. The model has employed the combination of Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) and Variational Mode Decomposition (VMD) in order to perform a finer separation of signals and reduction of noise obtained to get cleaner feature input. These sophisticated components are estimated by an Attention-Gated Recurrent Unit (A-GRU) network with an extension of a Memory-Gate Fusion (MGF) mechanism to pick out the complicated time dependences and point out interactions of pollutants. Moreover, Adaptive Multi Objective Optimization (AMO) strategy - Improved Whale Optimization Algorithm (IWOA) + Particle Swarm Optimization (PSO) is presented that will allow dynamically adjusting hyperparameters that will trade-off prediction accuracy and model generalization. Experiments on benchmark data of air quality data demonstrate that AQ-FusionNet can achieve a significantly higher degree of success than the existing AQ models like LSTM, GRU, CEEMDAN-GRU, and SARIMA-BiLSTM with an average coefficient of determination (R2) of 0.996, and with a decrease of approximately 23% of RMSE than the finest base model. The findings indicate AQ-FusionNet a scalable and flexible platform to the real-time monitoring of the environment air quality control.

Zitieren

APA

Joy, J., Somasundaran, A., Komban, M. J., N, R., & M, K. (2026). AQ-FUSIONNET: Dual-Decomposition Attention-GRU Framework with Adaptive Optimization for Air Quality Forecasting. Environment Conservation Journal, 27(2), 656-670. https://doi.org/10.36953/ecj.38793294

Vancouver

Joy J, Somasundaran A, Komban MJ, N R, M K. AQ-FUSIONNET: Dual-Decomposition Attention-GRU Framework with Adaptive Optimization for Air Quality Forecasting. Environment Conservation Journal. 2026;27(2):656-670. doi:10.36953/ecj.38793294

BibTeX
@article{Joy2026,
  author = {Joy, Jibin and Somasundaran, A and Komban, Meena Jose and N, Raji and M, Keerthika},
  title = {AQ-FUSIONNET: Dual-Decomposition Attention-GRU Framework with Adaptive Optimization for Air Quality Forecasting},
  journal = {Environment Conservation Journal},
  year = {2026},
  volume = {27},
  number = {2},
  pages = {656--670},
  doi = {10.36953/ecj.38793294}
}
Kostenlos starten →