Prediction model for fish growth rate using machine learning in aquaculture technology

Meor Khazamuddin, Nurul Alyaa Izzati (2024) Prediction model for fish growth rate using machine learning in aquaculture technology. Project Report. Universiti Teknikal Malaysia Melaka, Melaka, Malaysia. (Submitted)

Text (Full text) Prediction model for fish growth rate using machine learning in aquaculture technology.pdf - Submitted Version Download (1MB)

Abstract

Grouper fish farming is gaining popularity among entrepreneurs in Malaysia due to high demand and income return. However, in general this sector requires constant effort and commitment as well as knowledge complemented by skills. Current practice in fish farming still rely on traditional or manual methods to estimate the physical dimensional measurement and the weight of fish to forecast those fish growth rates. However, the concern arises here where manual methods could affect the health of the fish itself by applying unnecessary pressure and injury threat every single time when manual measurement took place. With that, developing new technology-based methods to improve fish performance and growth rate is a priority. Based on the problems stated above, the main objective of this project is to develop a prediction model for fish growth rate realizing with machine learning regression method, thus enabling fish farming to be carried out more accurately and efficiently. This project also determines the prediction accuracy across several potential machine learning algorithms, which are selected from conducted literature on past research studies. The proposed algorithms for prediction model will be trained and tested using selected datasets. In general, realization of entire project uses PyCharm as an IDE platform with Python programming language, including Graphical User Interface (GUI) design implementation. Random Forest Regression model outperformed the Linear Regression and Polynomial Regression models in terms of accuracy in predicting fish weight, according to comprehensive analysis of the project. The Random Forest Regression was determined to be the most predictive capability that achieve a high accuracy. This model shown the lowest value of MSE, MAE, and RMSE which has been determined to be 6540986.24, 1468.15 and 2557.54 respectively. In conclusion, machine learning is very important to predict accuracy of fish weight which are cost effective, precise and less strain on the fish that could harm it.

Actions (login required)

View Item