In order to seek for the effect factors on turbocharger efficiency, and explore effective methods to improve turbocharger performance, the turbine of a turbocharger was taken as the study object. The improvement of isentropic efficiency, which is a key parameter to evaluate the performance of a turbine, is treated as the main optimization target based on the design of experiment (DoE) method. The characteristic blade curve of turbine, which is a mathematical method to describe the shape of blade and includes 50 parameters, was reasonably selected and adjusted to fit the turbine. According to the original turbocharger, a computational fluid dynamics (CFD) model was established, and it was verified through a turbine hot gas experiment. The calculation values were in good agreement with the experiment values, and the maximum errors were 512% in mass flow rate and 2.18% in isentropic efficiency. 14 parameters describing turbine blade shape were selected out as design variables from the 50 controlling parameters of characteristic blade curves, and then the orthogonal matrix method was carried out on the design space. According to the results of simulation calculation, five parameters of high sensitivity were selected. Combined with CFD simulation, the optimal Latin hypercube design method was employed to acquire the distribution scheme of characteristic sample points and the simulation data within the sample space, through which a response surface approximation model was established accordingly. Based on the response surface approximation model, a kind of combinatorial optimization method combined with multiisland genetic algorithm (MIGA) and sequential quadratic programming (SQP) was employed in the process of optimization simulation calculation. The results of optimization simulation showed that the decreasing of the optimized circumferential bending degree of the turbine blade and the inclination angle of turbine blade at the entrance leaded to the decreasing of vortex intensity in the flow tunnel, therefore flow field distribution was more uniform and the flow losses was reduced. The isentropic efficiency of the turbine was 74.04%, which enhanced by 2.16 percentage points compared with the original turbine. This study can improve the design efficiency of the turbocharger and could be a reference for the design and optimization method of turbocharger turbine blade shape to some extent.