An Intelligent Catfish Farm Management System Using Random Forest and a Rule-Based Expert System

Fish farming has been a critical aspect in the area of aquaculture and food production in Nigeria. Despite this, farmers in Nigeria still struggle with water quality management, disease diagnostics, feeding process, data collection, and lack of expert advisory services among other issues. Conventional fish farm management approaches depend on manual observations and experiences, which usually lead to delayed decision making, high fish mortality rates, and low fish productivity. In this study, the development of the intelligent catfish management system is introduced as a solution that uses modern information technology, machine learning algorithms, and expertise to solve catfish farm management problems. The system was developed using Next.js and React for the frontend, Flask (Python) for the backend, and MySQL as the database management system hosted on Aiven. The prediction system implemented in this project uses the random forest classifier algorithm which is trained on 66,868 filtered rows in the Fishpond Health Monitoring Dataset. The predictor uses six water quality parameters including pH, temperature, turbidity, total dissolved solids, nitrate, and ultrasonic depth to predict the state of the ponds into five categories (Excellent, Good, Fair, Poor, and Critical). A rule-based expert system has been added for advisory purposes where needed. The accuracy of the Random Forest model was recorded at 99.60%, with a macro-average precision, recall, and F1 score of 0.99, 1.00, and 0.99, respectively, and confusion matrix indicating minimal error classification in terms of health class. Importance of features analysis indicated that pH level, temperature, and nitrate were the most significant features contributing to the prediction of pond health status. The proposed system can be used for the purpose of managing farm records digitally, AI-powered health prediction with confidence values, automated feeding recommendations, and live expert consultation via a socket.io-powered chat interface. Testing of the system established that all features of the system work as expected.