Genetic algorithm–based support vector regression modeling and optimization of neem oil transesterification using synthesized donkey bone–derived nanocatalyst

This research explores artificial techniques (support vector regression and genetic algorithm) to model, optimize and predict neem oil transesterification process via donkey bone nanocatalyst. The nanocatalyst was developed via the sol-gel method and extensively characterized using SEM-EDX, FT-IR, XRD, BET, and TGA techniques. The process parameters (time, temperature, agitation speed, and methanol/oil ratio and catalyst concentration) significantly impacted the biodiesel output. Analysis of Variance (ANOVA) confirmed the adequacy of a second-order polynomial model with an R² of 0.9704, adjusted R² of 0.9468, and predicted R² of 0.8817, reflecting a strong relationship between predicted and experimental results. The process was optimized using a Central Composite Design (CCD) implementing response surface methodology (RSM) and compared with support vector regression (SVR) data and optimizaed with Genetic algorithm (GA). RSM optimum conditions for biodiesel yield of 90.15%, GA(92.067%) and GA-SVR (93.29%) were obtained at a time of 1.0016 h, temperature(40 °C), methanol/oil ratio (10.138:1) catalyst concentration (2 wt%) and agitation speed (359.475 rpm). Also, the SVR model had a better prediction with an R2 (0.9918) and RMSE (1.01) than RSM model with an R2 (0.9704) and RMSE (1.562), however both models demonstrated a significant predictive behaviour. GA-SVR had a prediction than RSM by accurately modeling heterogeneous catalytic nonlinearity and achieving global optimization under experimental uncertainty. Characterizations result via FT-IR indicated a successful development of the biodiesel. Furthermore, the produced neem-based biodiesel complied with ASTM D6751 quality standards, underscoring its viability as a green fuel alternative.