A Comparative Study of API-Based and Edge AI Weather Prediction for Automated Fish Feeding

Aquaculture feeding efficiency is crucial since improper feeding schedules result in wastage, increased charges, and adverse impacts on the fish's health. Standard fixed schedules controlled conventional automatic fish feeders despite accounting for no dynamic real-time conditions of the environment. This paper introduces a weather-aware fish feeding system that dynamically adjusts feeding based on real-time weather conditions. The system integrates OpenWeatherMap API and a machine learning-based approach to optimize feeding decisions. The system classifies weather conditions and determines whether feeding should proceed or be delayed using real-time sensor data. The system was tested using an API-based method that retrieves external weather forecasts and an ML-based approach that processes local sensor data for real-time weather prediction. Experimental results show that the ML-based system outperforms the API-based approach in accuracy, response time, and reliability, achieving a 96.5% decision accuracy compared to 85.3% for the API-based system. These results suggest implementing edge AI for future work in weather-dependent fish feeding control, reducing dependence on external data sources and improving real-time responsiveness.