DMRI-Net: A Federated Multimodal Explainable Intelligence Framework for Dynamic Molting Risk Assessment and Predictive Water Quality Analytics in Smart Vertical Crab House Aquaculture
Journal of Intelligent Decision Making and Information Science · 21.08.2026 · 3 (2) · 1475-1500
DOI: 10.59543/jidmis.v3.2214 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Intelligent Decision Making and Information Science, online erschienen am 21.08.2026. Verfasst von Doan Perdana. Die Arbeit wurde bisher 1-mal zitiert (Crossref). Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Existing smart aquaculture systems are mostly based on reactive monitoring and unimodal analytics, which are inadequate for predicting environmental deterioration and biological responses. Based on these issues, this study proposes a federated multimodal explainable intelligence framework for predictive wwater-quality analytics and dynamic molting-risk assessment in a Smart vertical Crab House aquaculture system, named ater quality analytics and dynamic molting-risk assessment in Smart vertical Crab House aquaculture system, named as DMRI-Net. It combines a Temporal Transformer module for environmental encoding, a Vision Transformer module for behavioral encoding, cross-modal fusion, explainable AI (SHAP and Grad-CAM), and federated edge learning to model the environmental dynamics and the crab behavioral patterns. Also, a new Dynamic Molting Risk Index (DMRI) is provided to represent the probability of imminent molting as a continuous risk value, and a Predictive Water Quality Intelligence (PWQI) module is added to provide the multi-horizon forecasting of critical water-quality parameters. Experiments with the Scylla serrata IoT Aquaculture Dataset (18,144 environmental observations and 12,480 annotated behavioral images) show that DMRI-Net significantly outperforms classical machine learning, deep learning, and state-of-the-art multimodal baselines. The proposed framework has achieved the accuracy of 96.0%, F1 score of 95.0%, AUC of 97.0%, the R² of 0.94, the MAE of 0.12 and the RMSE of 0.18, all of which are good. The findings provide strong evidence of the potential benefits of integrating multimodal learning, explainable intelligence, and federated analytics for intelligent, proactive, interpretable, and scalable aquaculture management, setting a new benchmark for intelligent crustacean farming systems.
Zitieren
APA
Perdana, D. (2026). DMRI-Net: A Federated Multimodal Explainable Intelligence Framework for Dynamic Molting Risk Assessment and Predictive Water Quality Analytics in Smart Vertical Crab House Aquaculture. Journal of Intelligent Decision Making and Information Science, 3(2), 1475-1500. https://doi.org/10.59543/jidmis.v3.2214
Vancouver
Perdana D. DMRI-Net: A Federated Multimodal Explainable Intelligence Framework for Dynamic Molting Risk Assessment and Predictive Water Quality Analytics in Smart Vertical Crab House Aquaculture. Journal of Intelligent Decision Making and Information Science. 2026;3(2):1475-1500. doi:10.59543/jidmis.v3.2214
BibTeX
@article{Perdana2026,
author = {Perdana, Doan},
title = {DMRI-Net: A Federated Multimodal Explainable Intelligence Framework for Dynamic Molting Risk Assessment and Predictive Water Quality Analytics in Smart Vertical Crab House Aquaculture},
journal = {Journal of Intelligent Decision Making and Information Science},
year = {2026},
volume = {3},
number = {2},
pages = {1475--1500},
doi = {10.59543/jidmis.v3.2214}
}
EUVIMED