This paper presents a novel approach to calibrate the Universal Thermal Climate Index (UTCI) stress categories. Subjective thermal sensation votes (TSVs) are determined using questionnaire survey of 1030 individuals across 15 monitoring campaigns. Objective thermal stress (UTCI) is determined from on-site meteorological measurements, corresponding to the survey time, spanning various seasonal environmental conditions, in the hot-arid climate of Cairo, Egypt, and is then compared against subjective thermal sensation votes. Boundary thresholds are first derived using three common calibration techniques, namely, Linear Regression (LR), Ordinal Logistic Regression (OLR) and the Percentage Dissatisfied (PD) fitted curve, and their abilities to predict subjective votes are evaluated. Leveraging the parametric capabilities of Grasshopper for Rhino3D, a new methodological framework is then presented through multi-objective optimisation of per-class recalls. The Hypervolume Estimation (HypE) algorithm is employed, with the boundary thresholds as design parameters, along with the geometric-mean recall, minimum per-class recall and Quadratic Weighted Kappa (QWK) as objective functions. The Pareto front comprises solutions with up to 0.2 lower Mean Absolute Error (MAE) and up to 15% higher average recall, compared to LR. The final optimised scale achieves 7% higher average recall, 12% higher precision, and 0.1 higher QWK, relative to OLR, providing a balanced prediction across all stress categories. The optimised neutral range indicates higher tolerance to heat stress in hot-arid climates than that of the original no thermal stress range. The framework's generalisation to other thermal indices across different climatic regions is discussed, highlighting its applicability for broader thermal comfort assessment.