Transformer oil-based nanofluids are known to have higher thermal conductivity and heat transfer performance compared to conventional transformer oils. In this study, four different types of transformer oil-based nanofluids are synthesized using the well-known two-step method. The first nanofluid contains pure multi-walled carbon nanotubes (MWCNTs), while other samples consist of 20 Vol% of MWCNTs and 80 Vol% of different oxide nanoparticles (i.e., Al2O3, TiO2, and SiO2). The dynamic viscosity and thermal conductivity of prepared samples are investigated in seven different volume fractions of 0.001, 0.0025, 0.005, 0.01, 0.025, 0.05, and 0.1%. Besides, the breakdown voltage of the pure transformer oil and nanofluids containing 0.05 and 0.1 Vol% of nanoparticles is investigated. The outcomes show that dielectric properties of hybrid carbon-based nanofluids are far better compared to those properties of the pure MWCNTs nanofluids. Finally, eight different soft computing approaches, including group method of data handling (GMDH), support vector machine (SVM), radial basis function (RBF) neural network, multilayer perceptron (MLP), and MLP and RBF models optimized with bat and grasshopper optimization algorithm (GOA), are used to model the viscosity and thermal conductivity of synthesized nanofluids. The outcomes show that the GMDH approach significantly outperforms all other models in terms of predicting the thermal conductivity and dynamic viscosity of transformer oil-based nanofluids.