Measuring turbine blade temperature is critically important due to the material's thermal limits. In this study, machine learning (ML) method has been used to predict the heat flux of the outer surface of a turbine blade model. The heat flux is predicted from the outlet temperature and the inlet velocity and inlet temperature and blade model length, which is in inverse heat transfer problems (IHTP). The Reynolds-Averaged Navier-Stokes (RANS) equations were solved to analyze heat transfer, and the k–ω model was used to account for turbulence effects. Simulations were carried out in 648 different cases with varying inlet velocities (6 to 20 m/s), inlet temperatures (300 to 310 K), blade heat fluxes (400 to 2000 W/m²) and blade model lengths (180 to 220 mm). The channel surface temperature, measured 100 mm downstream of the blade model, was recorded as the output. The results of the numerical analysis showed that the surface temperature is directly proportional to the heat flux of the blade model and is inversely proportional to the flow rate. The simulation results were processed as an inverse problem using support vector regression (SVR), optimized support vector regression, Gussian Process Regression (GPR), and multi-layer Perceptron (MLP) algorithms. The Marine Predators Algorithm (MPA) was utilized to optimize the hyperparameters of the SVR model, resulting to a significant improvement in the results. The mean absolute error for support vector regression, optimized support vector regression, multi-layer Perceptron, and Gaussian process regression algorithms were 0.0564, 0.0094, 0.0037, and 0.0069, respectively. These results indicate that the multi-layer Perceptron algorithm provided the best prediction performance for this problem. Finally, the MLP method was also processed with 792 and 432 data sets, which shows that the results are independent of the number of data sets.