Water Quality Management Guidelines to Reduce Mortality Rate of Red Tilapia (Oreochromis niloticus x Oreochromis mossambicus) Fingerlings Raised in Outdoor Earthen Ponds with a Recirculating Aquaculture System Using Machine Learning Techniques

Abstract

Machine learning techniques have been widely adopted over the last few decades, especially in fisheries. This study aimed to determine the best practice of machine learning techniques with a decision tree algorithm in reducing the mortality rate of red tilapia (Oreochromis niloticus x Oreochromis mossambicus) fingerlings raised in outdoor earthen ponds with a recirculating aquaculture system. The study phase begins with collecting water quality parameters. The parameters were measured in the form of dissolved oxygen (mg L^-1), pH, temperature (^°C), total ammonia nitrogen (mg L^-1), nitrite-nitrogen (mg L^-1), alkalinity (mg L^-1), transparency (cm), and mortality rate (fish day^-1).  Data Modelling was carried out using 10-fold cross-validation. The results of the performance measurement obtained an accuracy of 89.67% with ± 5.11% (micro average: 89.60%), a precision of 86.71% ± 18.02% (micro average: 80.00%), and recall of 72.50% ± 24.86% (micro average: 71.79%), with the most influential water quality parameter being nitrite-nitrogen (mg L^-1). Based on the results of this study show that data classification using a decision tree algorithm can be used as a reference to determine the decisions or actions of fish farmers in reducing the mortality rate of red tilapia fingerlings raised in outdoor earthen ponds with a recirculating aquaculture system.