COBRA-ML: HYBRID MULTI-CRITERIA DECISION SUPPORT METHOD WITH OBJECTIVE WEIGHTING AND MACHINE LEARNING

Distance-based multicriteria methods such as TOPSIS and VIKOR face limitations from subjective weighting and outlier sensitivity. This paper proposes COBRA-ML (Comprehensive Distance-Based Ranking with Machine Learning), a hybrid extension incorporating: (i) Euclidean vector normalization; (ii) objective weighting via Preference Selection Index (PSI); (iii) Gaussian structural balancing to correct PSI's tendency to underweight dispersed criteria; and (iv) automatic parameter calibration via Random Forest Regressor (cross-validated R² = 0.880 ± 0.015). Applied to renewable energy source selection in Brazil, the method identifies Onshore Wind as optimal (score = 0.6843), with mean Kendall τ = 0.933 and perfect agreement with PROMETHEE II (τ = 1.000). Software registered at INPI (blind review) and available as open source.