Prediction of biodiesel production from microalgal oil using Bayesian optimization algorithm-based machine learning approaches

Biodiesel has appeared as a renewable and clean energy resource and a means of diminishing global warming. This study provides Bayesian optimization algorithm (BOA) based machine learning techniques such as artificial neural network (ANN) and Support vector regression (SVR) as the potential tool for modeling biodiesel production using microalgae oil as feedstock. Novelties of this study as in comparison with the existing Raj et al. model include (i) implementation of BOA to tune the model hyperparameters, (ii) hybridization of BOA with ANN, and SVR for modeling biodiesel production for the first time, (iii) the model performance was compared between the developed models and the existing model using several performance indicators (viz., R_pred², residual analysis, RE, MAE, RMSE), and (iv) validation of the model using extra experimental data published elsewhere. The developed hybrid BOA-ANN and BOA-SVR models show better performance in comparison with the existing Raj et al. model. Comparing BOA-ANN and BOA-SVR, the later model shows excellent performance. Based on root mean square error (RMSE), the developed hybrid BOA-SVR shows higher performance than Raj et al. model with a performance enhancement of 36.03%. The precision of the hybrid BOA-SVR model was further validated with extra literature data. Thus, the proposed model would certify rapid estimation of biodiesel yield from microalgal oil that may reduce laborious, expensive, and time-consuming laboratory trials.