Optimization Based Framework for The Predication of Biomass Liquefaction using ML Methods Integrated with Metaheuristic Techniques

Abstract

The liquefaction of microalgae yields bio-oil, that is a potential alternative fuel. However, it is a complicated and challenging process to use the method of experimentation to investigate the relationship between liquefaction conditions of, ultimate, and proximal analysis with bio-oil production. so, an efficient and well-organized model must be created to reliably predict the effect of input parameters on the bio-oil yield. This study utilizes a novel PSO-based and genetic algorithm-based approach for feature selection and hyperparameter optimization. Four different machine learning models were constructed and compared using these optimization methods. It was found that Ensembled Tree model performed better and R2=0.6801, RMSE=1.7226 for GA and the values of R2 (Coefficient of determination) = 0.983 and RMSE (Root mean Square Error) = 1.4548 while using PSO performed better and highly recommend based features selection. The result of SVM was the worst one R2 = 0.7483, RMSE =10.0063 for PSO and R2 = 0.6068, RMSE = 35.3462 for GA. For the PSO-based study, the R2 values for DT and GPR were 0.8125 and 0.9309, respectively, while the R2 values for the GA-based methods were 0.8757 and 0.9270.