Predicting Water Quality Parameters in Mahseer Fish Farming Using Machine Learning Techniques

Abstract

Mahseer fish farming faces challenges in maintaining optimal water quality, essential for fish health and growth. Poor water quality can lead to stress, disease, and mortality, impacting productivity. This study compares Random forest Regression (RF) and Support Vector Regression (SVR) models in predicting water quality parameters, such as pH, dissolved oxygen (DO), and temperature. The RF model outperformed SVR, showing superior accuracy with lower Mean Squared Error (MSE) and Mean Absolute Error (MAE) and higher R-squared values (99% for DO, 98% for temperature, and 95% for pH). RF’s superior performance makes it a reliable tool for tracking water quality trends and fluctuations. Recommendations for enhanced monitoring include extending data turbidity and capturing seasonal and long-term trends, integrating sensors for additional parameters like ammonia and turbidity, and developing a user-friendly mobile app for real-time data and alerts. These improvements aim to support the sustainability and productivity of Mahseer fish farming.