Abstract: Burnup credit accounts for changes in nuclide composition during irradiation and thus reduces the estimated effective multiplication factor (k_eff) in criticality-safety assessments of spent nuclear fuel. However, uncertainties arising from sample-to-calculation deviations in nuclide inventories degrade the accuracy of criticality calculations. To quantify those uncertainties across different burnup ranges and to improve the precision of burnup-credit analyses, this work develops a nuclide-bias partitioning strategy based on linear-regression fitting and mean-squared-error minimization to determine optimal burnup interval boundaries for target nuclides. Using an AFA-3G 17×17 PWR fuel assembly as a benchmark, depletion and criticality calculations were carried out on SCALE 6.1 with ORIGEN and KENO modules, and Monte Carlo sampling was employed to propagate nuclide bias uncertainty into k_eff. Results indicate that the proposed partitioning strategy reduces criticality uncertainty by approximately 40% under low-burnup conditions, offering an optimized approach for burnup-credit evaluations in spent-fuel criticality safety assessments.