Predictive modeling of marine fish production in Brunei Darussalam's aquaculture sector: A comparative analysis of machine learning and statistical techniques
Haziq Nazmi,Nor Zainah Siau,Arif Bramantoro,Wida Susanty Suhaili
DOI: https://doi.org/10.21833/ijaas.2023.07.013
2023-07-01
International Journal of ADVANCED AND APPLIED SCIENCES
Abstract:The aquaculture industry has witnessed significant global growth, offering opportunities for sustainable fish production. This research delves into the application of data analytics to develop an appropriate predictive model, utilizing diverse machine learning and statistical techniques, to forecast marine fish production within Brunei Darussalam's aquaculture sector. Employing a machine learning-based algorithm, the study aims to achieve enhanced prediction accuracy, thereby providing novel insights into fish production dynamics. The primary objective of this research is to equip the industry with alternative decision-making tools, leveraging predictive modeling, to identify trends and bolster strategic planning in farm activities, ultimately optimizing marine fish aquaculture production in Brunei. The study employs various time series and machine learning techniques to generate a precise predictive model, effectively capturing the inherent seasonal and trend patterns within the time-series data. To construct the model, the research incorporates notable algorithms, including autoregressive integrated moving average (ARIMA), long short-term memory (LSTM), linear regression, random forest, multilayer perceptron (MLP), and Prophet, in conjunction with correlation analysis. Evaluation of the model's performance and selection of the optimal forecasting model are based on mean absolute percentage error (MAPE) and root mean squared error (RMSE) metrics, ensuring a robust analysis of time series data. Notably, this pioneering research stands as the first-ever attempt to forecast marine fish production in Brunei Darussalam, setting a benchmark unmatched by any existing baseline studies conducted in other countries. The experiment's results reveal that straightforward machine learning and statistical techniques, such as ARIMA, linear regression, and random forest, outperform deep learning methods like MLP and LSTM when forecasting univariate time series datasets.