An Integrated IoT–AI Framework for Sustainable Aquaculture: Real-Time Water Quality Prediction and Vision Transformer–Based Fish Species Identification in Bangladesh

Aquaculture is central to Bangladesh’s economy, supplying vital protein and supporting the livelihoods of millions. Nevertheless, the industry faces several challenges, especially in water quality management and species identification, which threaten its future, particularly in the export market. This paper presents an integrated solution, including IoT-based water quality monitoring, AI-driven water quality classification, and fish species identification based on Vision Transformers (ViT), to address these problems in the aquaculture industry of Bangladesh. We designed an IoT-based system for real-time monitoring of important water parameters such as pH, dissolved oxygen (DO), turbidity, temperature, and total dissolved solids (TDS). The solution is based on a system for continuous data collection and cloud service integration, enabling farmers to remotely monitor their farm’s water quality. A machine learning model, namely XGBoost optimized with the Honey Badger Algorithm (HBA), classifies water quality levels into Excellent, Good, and Poor categories with an accuracy of 98.05%. Finally, we introduce a ViT-based model for precise fish classification and report 100% classification accuracy across 26 freshwater fish species. The system contributes to improved decision-making in aquaculture, facilitating a reduced incidence of misidentification during export, enhanced environmental management, and sustainable fish farming practices (SfAM). This study provides insight into how IoT, AI, and computer vision can revolutionize the management of this sector and lead to more efficient and resilient farming in Bangladesh.

• Supervision