Adaptive thermal comfort is essential for guaranteeing the well-being of building occupants. Therefore, this study emphasizes its importance and applies advanced computational techniques as a surrogate model and sensitivity analysis for an in-depth analysis of a social housing in the Dominican Republic evaluated under five micro-climates. The methodology includes a computational design with an energetic simulation. An adaptive model was developed following the ASHRAE-55 standard, and an artificial neural network was trained to predict comfort temperature. The results showed that Neiba city, corresponding to a hot semi-arid climate, presented the maximum hours of discomfort (8463), representing 96.6% of the total annual hours with a cooling degree day of 157.5, whereas Constanza city, corresponding to the oceanic climate, reported 2025 hours of discomfort (23.1%) and a cooling degree day of 48. The surrogate model based on artificial neural networks achieved a coefficient of determination of 0.9998, and the sensitivity analysis revealed a more significant influence of radiant temperature (55.96%) over comfort temperature. Finally, this study highlights the significance of adaptive thermal comfort in building design and the potential of surrogate models for energy analysis. Therefore, it is necessary to implement passive strategies to enhance indoor thermal comfort in a sustainable building sector.