<p>This paper introduces the optimization of micromixers with modified Cantor fractal structures through CFD and machine learning. To create the desired contoured channel structure and obstacle configuration in the main channel, we developed micromixers with modified Cantor fractal structures. The effects of the ratio of the edge length of a large obstacle (R₁/R₂) and the ratio of the edge length of a small obstacle (r₁/r₁), respectively, on the percentage of mixing index (MI), pressure drop (∆p), and mixing energy cost (mec) were investigated. For the study of design parameters, the Latin hypercube sampling method is used to select design points in the design parameters. Then, the datasets consisting of inputs r₁/r₂ and R₁/R₂ and outputs mixing index, pressure drop, and mixing energy cost are trained using the gradient boosting regression model. The genetic algorithm obtains the geometric parameters which provide the maximum mixing index and minimum pressure drop. The results show that the value of mixing index increases with the increase in r₁/r₂ and R₁/R₂, and the pressure drop and mixing energy cost increase with the increase in r₁/r₂ and R₁/R₂ at all Reynolds numbers. Finally, compared with the reference design, the mixing index is increased by 81.19%, the pressure drop is increased by 30.23%, and the mixing energy cost is decreased by 37.74%.</p>