Abstract Accurate estimation of transesterification efficiency is needed for designing operational conditions and obtaining the maximum yield of biodiesel production. Accordingly, the objective of this study is to analyse the applicability of three different hybrid soft computing techniques for the prediction of transesterification yield under ultrasound irradiation as a novel and under mechanical stirring as a traditional method of biodiesel synthesis. The models include ANFIS-PSO (Adaptive Neuro-Fuzzy Inference System linked with Particle Swarm Optimization), ANFIS-GA (Adaptive Neuro-Fuzzy Inference System linked with Genetic Algorithm) and ANFIS-DE (Adaptive Neuro-Fuzzy Inference System linked with Differential Evolution). Independent variables including reaction temperature, reaction time, reactant concentrations, catalyst loading and power input were considered as the network inputs while the reaction yield was considered as the network output. The obtained simulation results were then analysed using Kolmogrov-Smirnov method as well as root mean-square error (RMSE) and coefficient of determination (R 2 ). The analyses confirmed the validity of the proposed models. It was found that although ANFIS-PSO had better performance in training phase, it generated the weakest results in testing phase. Meanwhile, ANFIS-DE provided the best statistical characteristics compared to the other methods for estimating the transesterification yield under either ultrasound irradiation or mechanically stirring.