To investigate the influence of mix proportion parameters on the dilatancy behavior of hydraulic asphalt concrete, the Box–Behnken design in response surface methodology was adopted to establish regression models for failure strain, phase transformation stress ratio and maximum deviatoric stress. The effects of asphalt content, filler content and gradation index on dilatancy behavior were analyzed. Furthermore, NSGA-II combined with entropy-weighted TOPSIS was used for multi-objective optimization of the mix proportion. The results show that the correlation coefficients of the regression models for failure strain, phase transformation stress ratio and maximum deviatoric stress are 0.9627, 0.9596 and 0.9640, respectively, indicating good fitting accuracy. The interaction between asphalt content and filler content significantly affects the maximum deviatoric stress and phase transformation stress ratio, while the interaction between asphalt content and gradation index significantly affects the failure strain. The optimal mix proportion is determined as 6.50% asphalt content, 12.37% filler content and 0.44 gradation index. The verification results show small errors between predicted and measured values, indicating that the proposed method is accurate and applicable for mix design and performance optimization of hydraulic asphalt concrete.