From Ranking Instability to Score Robustness: Heuristic Optimization Methods as Applied to Constructing of a Composite Index

This study comparatively investigates the efficiency and robustness of heuristic optimization methods applied to the construction of Composite Indicators (CIs) under uncertainty, with a focus on the stability of score values. The determination of weights of separate components of the indicators weights is formulated as an optimization problem aimed at minimizing the variability of scores under stochastic perturbations in the input data. Four heuristic strategies are analyzed: clocktype local search, clock with an annealing mechanism, Particle Swarm Optimization (PSO), and the Ray method. Performance is evaluated using metrics such as the mean objective function value, score dispersion, sensitivity to perturbations, and computational time. Experiments conducted under different levels of uncertainty reveal structural differences among the methods, highlighting the role of global exploration in reducing variability. The results show that the choice of optimization
method directly affects score stability, contributing to greater reliability of the resulting CIs.