Friction stir welding is of the most applicable methods to join dissimilar metals. In this study, the thermal distribution during the joining of 304 stainless steel and 5083 aluminum alloy by friction stir welding method was simulated by the finite element method. Both, transient and stationary thermal solutions were used in the simulations and the two methods were compared correspondingly. To verify the model, two sheets of stainless steel and aluminum were prepared and the friction stir welding was applied. Additionally, by using thermocouples temperature, the history of points on the sheets was obtained during welding. Then, the simulation and the experimental results were compared to validate the model. Finally, an artificial neural network model was created and the effect of different input parameters on the maximum temperature under the tool was investigated.