Machine Learning-Based Algorithm for Fish Kill Evaluation using Differential Changes of Input Parameters via Wireless Communication

The existing methods of monitoring and management of freshwater ponds to prevent fish kills cannot adapt fast enough to the frequently changing environment. Recent inventions monitor water quality and water level in fishponds in response to fish mortality related to the dissolved oxygen level, water salinity, pH levels, temperature, and many more. With the additional use of machine learning algorithms, these methods can be further improved for more energy-efficient water quality detectors. This project is a buoy-type device that is designed to transmit data through an RF transceiver module outside the set threshold embedded in the microcontroller. The buoy transmits data when there are great differential changes in the measured parameters to save energy. Then the receiver hub is equipped with a microprocessor that utilizes K-means Clustering Machine Learning Algorithm that can reconstruct as much as possible the received data. The buoys were deployed in a fishpond and tested their functionality. The sensors were calibrated and the analysis of the gathered results for the deployment showed a significant amount of power reduction when compared with and without data transmission.