RANDOM-KEY OPTIMIZER FOR THE UNIVARIATE STRATIFICATION PROBLEM

The problem of univariate stratification is recurrent in the sample planning of institutes
national statistical authorities. Once a level of precision has been established, the aim is to partition the population into L
strata, in order to minimize the total sample size n. In this work, an approach is proposed
hybrid with metaheuristic-inspired heuristics from the Random-Key Optimizer framework
(RKO) and an exact sample allocation method, ensuring optimal allocation in the final step. Performance
It was evaluated in an extensive study with 20 populations, 10 scenarios and 24,000 executions,
including statistical analyses based on the Friedman and Wilcoxon tests. The results obtained
were competitive with the algorithms in the literature, standing out for explicitly incorporating
optimization criteria and ensure sample completeness and feasibility, overcoming limitations of
classical approaches based on rounding.