Comparative Evaluation of Machine Learning Models for Predicting Biomass Production in Thraustochytrid Cultivation

Abstract

Thraustochytrids are promising marine microorganisms for the production of high-value compounds such as docosahexaenoic acid (DHA). However, optimizing their cultivation conditions using conventional experimental approaches is labor-intensive and time-consuming. In this study, the performance of Support Vector Regression and Random Forest models was compared for predicting biomass production under different cultivation conditions. Model performance was evaluated using the coefficient of determination (R²), root mean square error (RMSE), and mean absolute error (MAE). Among the evaluated models, SVR demonstrated superior predictive performance, achieving higher prediction accuracy and lower error values than RF. These findings suggest that SVR is a reliable tool for biomass prediction and can facilitate the optimization of Thraustochytrid cultivation while reducing experimental effort and cost.