Automated Siphoning and Monitoring System Prototype for Water Quality Parameters in a Controlled Small-Scale Litopenaeus Vannamei Environment

In intensive Litopenaeus vannamei farming, residual waste influences the water quality, which directly affects the growth rate of shrimp. Most shrimp farms utilize manual methods for waste removal and test kits for water quality monitoring causing inefficiency and inconvenience for the farmers. The purpose of this study is to automate water quality monitoring in terms of pH, dissolved oxygen, salinity, and temperature and to investigate the behavior of these parameters to trigger the automated siphoning system. A web interface displayed the system's operations, and a small-scale replica of a Litopenaeus vannamei farm was used. A dataset with 2880 training data points and 288 test data points was gathered. Principal Component Analysis (PCA) was performed, and two machine learning algorithms - Logistic regression and Decision tree - were developed and compared. Cross-validation and verification using the test data was conducted to evaluate the models' performance. The Decision tree showed outstanding performance due to its higher test accuracy of 95.14% and relative simplicity compared to Logisitic regression with a test accuracy of 94.79%. Moreover, the machine learning methods and PCA identified dissolved oxygen as the most influential parameter on the water quality of shrimp farm, serving as triggering factor for the operation of the automated siphoning system.