Smart aquaculture is one of the most important latest development trends in the aquaculture industry for farmers to monitor water quality, feed supply, temperature imbalance and water recycling.
Intelligent real-time water quality prediction using machine learning models and Internet of Things sensors lay the foundation for intelligent aquaculture system assessment, planning and regulation.
A team of researchers from Heritage Institute of Technology, and Jadavpur University proposed a biofloc system based on the Internet of Things and intelligent Machine Learning to improve efficiency, production, water recycling system and feeding system automatic.
Smart aquaculture
The development of technologies such as the Internet of Things, big data and artificial intelligence has allowed aquaculture to progressively become more intensive, precise and intelligent.
In intensive aquaculture systems, continuous monitoring and prediction of water quality in real-time is of great significance to preventing water deterioration and avoiding disease outbreaks.
A smart aquaculture system based on the Internet of Things can help farmers continuously measure water quality parameters using sensors.
Machine learning models
The study proposes a system that collects process data from sensors, stores data in the cloud and analyzes it using different machine learning models (Decision tree, Random Forest, Logistic Regression & Support Vector Machine), to predict water quality and provide real-time monitoring, through an Android application.
The database that was generated with the Internet of Things sensors, and that has been used for training the machine learning model and validation, includes the following parameters: pH, hardness, solids, chloramines, sulfates, conductivity, organic carbon, trihalomethanes, turbidity and potability.
According to the researchers, the Random Forest Classifier outperformed all other models in terms of accuracy, recall, and F1 score.
“Random Forest Classifier exhibited an accuracy of 73.76%, recall, precision, and F1 score of 90%, 75%, and 82%, respectively; while the Support Vector Machine exhibited an accuracy of 71.29%, recall, precision, and F1 score of 81%, 76%, and 79%, respectively,” the researchers report.
As for the prediction time, the Random Forest Classifier recorded a prediction time of 0.054 seconds.
Conclusion
“In this study different machine learning models such as Logistic Regression, Random Forest, Support Vector Machine, Decision Tree, K-nearest Neighbors, XGBoost, Gradient Boosting and Naive Bayes have been developed to predict the values of pH, hardness, solids, chloramines, sulfate, conductivity, organic carbon, trihalomethanes, turbidity and potability”, they highlighted.
“Random Forest Classifier outperformed all other models in terms of accuracy, recall, accuracy, and F1 scoring. Random Forest Classifier has exhibited an accuracy of 73.76%, recall, precision and F1 score of 90%, 75% and 82% respectively”, they conclude.
Reference (free access)
Sen, Sohom; Maiti, Samaresh; Manna, Sumanta; Roy, Bibaswan; GHOSH, ANKIT (2023): Smart Prediction of Water Quality System for Aquaculture using Machine Learning Algorithms. TechRxiv. Preprint. https://doi.org/10.36227/techrxiv.22300435.v1
