Creation of a stress relief groove is a fairly simple yet high-performance method. During the application of this method, it is important to consider the location and size of the groove in order to achieve better performance. Consequently, this research proposes an approach for optimizing the application of the stress relief groove method to a press-fitted assembly. In a boss design, the position and diameter of the groove are configured as design variables and the design of experiments is applied. Based on this information, a 3D model is built and analyzed using the finite element analysis software ABAQUS. Meta-models are created using back-propagation neural networks. Then, deterministic optimization results obtained from a genetic algorithm are compared with the results of the finite element analysis. The temperature sensitivity of the optimized model is analyzed, and finally, reliability-based design optimization is conducted for enhancing the design quality.