Abstract: Optimization of stilling basin designs has attracted research efforts in hydraulic engineering. In the optimization, the widely used trial-and-error approach is not only inefficient, but difficult to take multi-objectives into account. Thus it is essential to design a systematic and goal-oriented optimization approach. This study adopts the support vector regression (SVR) to construct an approximate model for the relationship of optimized variables (height of falling step, expansion ratio, and end sill slope) versus optimization targets (energy dissipation ratio and near-bottom velocity) through a case study of designing a stilling basin for the Dazhai reservoir in Sichuan. And we solve it using a genetic algorithm (GA) and obtain an optimized scheme. Comparison with the pre-optimized design shows that the energy dissipation ratio of this optimized design is roughly the same, while its overall performance is significantly improved, particularly the near-bottom velocity that has a much lower peak and the time-average pressure that tends to be uniform along the bottom wall. This study indicates that our new optimization approach would be applicable to other similar hydraulic structures.