This study investigates a hybrid methodology that integrates stochastic optimization algorithms with multi-criteria decision analysis (MCDA) methods to identify decision-makers’ preferences for criteria weights. Specifically, Particle Swarm Optimization (PSO) and Cat Swarm Optimization (CSO) were combined with the TOPSIS (Technique for Order Preference by Similarity to Ideal Solution) method to approximate criteria weights when expert-derived reference values are unavailable. A two-dimensional decision matrix comprising 30 alternatives evaluated across nine criteria was generated, with reference weights computed using the CRITIC objective weighting method. The stochastic algorithms were then employed to iteratively determine weights that minimize the Euclidean distance between the generated TOPSIS rankings and the reference ranking. The results demonstrate that both PSO and CSO effectively identify decision-makers’ preferences, with CSO achieving superior accuracy. Convergence analyses and fitness function evaluations revealed that CSO maintains consistent progress throughout iterations, ultimately producing weights and rankings more closely aligned with reference values. Both algorithms successfully identified the top-ranked alternative, confirming their practical utility. This approach provides a significant advantage over traditional MCDA methods by capturing non-linear preferences and enabling automated preference identification without direct expert input. The study highlights the potential of stochastic optimization for enhancing MCDA frameworks.