Optimization of thermal biofuel production from biomass using CaO-based catalyst through different algorithm-based machine learning approaches

Optimization of biofuel production from algal oil through utilizing a CaO-based catalyst was carried out in this study. The optimal point for the highest yield of the reactions was determined using machine learning. To implement the optimization task, and to make predictions, we used three different methods, including Quantile regression, Logistic regression, and Gradient Boosted Decision Trees. The regression problem includes the amount of Catalyst, Reaction time, and Methanol/oil as input features, and FAME (fatty acid methyl ester) yield is the single output. We tuned the boosted version of these models with their important hyper-parameters and selected their best combination. Then different standard metrics are employed to assess their performance of them. Considering R2 score, Quantile regression, Logistic regression, and Gradient Boosted Decision Trees have error rates of 0.934, 0.996, and 0.998, and with MAE, they have 1.94, 1.68, and 1.17 errors, respectively. Also, Considering MAPE 2.14x10-2, 1.89x10-2, and 1.29x10-2 values obtained. Gradient Boosting is selected as the most appropriate model finally. Furthermore, the optimal output value with the proposed approach is 97.50, with the input vector being (x1 = 153, x2 = 0.625, x3 = 20).