<p>Injection molding of complex three-dimensional curved components is highly sensitive to processing conditions, particularly when dimensional accuracy and packing quality must be balanced simultaneously. This study proposes an integrated optimization and decision-making framework to improve the molding quality of a commercial curved shin guard plate (SGP) by combining Response Surface Methodology (RSM), adaptive non-dominated sorting genetic algorithm II (NSGA-II), and entropy-weighted Technique for Order Preference by Similarity to Ideal Solution (TOPSIS). A Box–Behnken design was employed to investigate the effects of packing pressure (<i>PP</i>), melt temperature (<i>MT</i>), and cooling time (<i>CT</i>) on absolute dimensional deviation (<i>ΔD</i>) and molded part weight (<i>PW</i>). The results showed that the developed quadratic response surface models exhibited good predictive capability, and analysis of variance identified <i>PP</i> as the most influential parameter. The surrogate models were integrated with standard and adaptive NSGA-II algorithms to perform multi-objective optimization and generate well-distributed Pareto-optimal solutions, with the adaptive NSGA-II showing improved Pareto-front exploration capability based on crowding-distance and hypervolume analyses. Entropy-weighted TOPSIS was subsequently applied to identify the most suitable compromise solution. Experimental validation was conducted at the TOPSIS-selected optimum together with two additional Pareto-optimal conditions. The validation results showed satisfactory agreement between predicted and experimental responses, with deviations remaining below 7.41% for <i>ΔD</i> and below 1.70% for <i>PW</i>. These results demonstrate the practical applicability of the proposed optimization framework for complex injection-molded components under industrial manufacturing conditions.</p>