In hot-arid climates, the control of shade represents an essential strategy for improving outdoor thermal comfort. Yet, empirical comparative analyses of natural versus artificial shade effectiveness, based on advanced machine learning techniques, are still scarce. This research suggests an innovative solution to the evaluation of thermal comfort in extreme climates, i.e., a hybrid GA-PSO-SVR optimization model with Monte Carlo enhanced methods. Field measurements were made in three stations (reference, building shade, tree shade) at the University of Zabol, Iran, in summer 2022. The assessment of thermal comfort was performed through the PET, UTCI, and WBGT indices with the use of extensive model validation. The innovative approach showed enhanced performance compared with traditional methods, which was expressed through the improvement in predictive accuracy by 22.4%. The model validation procedure provided R-2 values between 0.65 and 0.98, depending on the dominant data conditions. Methodology effectiveness was indicated by R-2 > 0.98 in training. The cross-validation results, characterized by an R-2 of 0.65 +/- 0.20, provided realistic operational forecasts. Comparing tree shade to building shade, tree shade showed a 1.6 degrees C PET superiority [95% CI: 0.8-2.4 degrees C] over building shade (p = 0.003). In addition, both types of shade were also found to provide a significant enhancement in thermal comfort, with a PET reduction between 6 and 8 degrees C. The hybrid optimization method with extensive uncertainty quantification pushes the frontiers of thermal comfort modeling while being scientifically transparent regarding performance variation. The outcomes of this research offer a solid basis for evidence-driven prioritization of natural shade features in hot-arid urban design, with a qualified consideration of methodological limitations.