Thermal sensation votes (TSVs) are often modeled as real-valued regression targets or as unordered classes, despite the seven-point ASHRAE scale being inherently ordered. We reframe TSV prediction as an ordinal learning problem and train cumulative-link models that produce class probabilities suitable for risk-aware decisions. Using the ASHRAE Global Thermal Comfort Database II and a large Chinese dataset, we harmonize features ( ) with a fold-safe, simple imputation policy (primarily with an indicator) and evaluate 5-fold out-of-fold performance and cross-corpus swaps. Across metrics that capture accuracy and safety (MAE/Within-1, quadratic-weighted , FAR@2, CRPS, ECE), the ordinal booster achieves the highest quadratic-weighted (fewer long-distance errors). Overall FAR@2 is comparable across learning models on the aggregate, with nominal/neural baselines slightly lower and the ordinal model showing lower warm-tail but higher cold-tail FAR. Nominal baselines are best calibrated; post-hoc temperature scaling improves probability quality without changing rankings. Calibration and confusion diagnostics together with permutation-importance analyses indicate physically sensible effects. Cross- corpus tests show stable generalization across the two corpora. Ordinal learning with calibrated probabilities provides a transparent, reproducible path to risk-aware comfort prediction using existing field datasets. • Ordinal learning reduces long-distance TSV errors vs PMV and ML. • Fold-safe preprocessing avoids leakage; simple imputation is robust. • Calibrated probabilities enable simple risk-aware thresholds. • Cross-corpus validation confirms generalization across studies. • Transparent methods: splits, hyperparams, calibration and stats.
Guo et al. (2026) studied this question.