With the intensification of climate change, the energy consumption pattern in buildings has undergone changes, especially in cold and dry climates. Tabriz, as one of the cold cities in Iran, will face an increase in cooling energy demand in the coming decades. The present study aimed to simultaneously evaluate the energy consumption and thermal comfort of office buildings under future climate scenarios and identify physical parameters affecting the thermal resilience of the building. In this study, an office building was modeled using DesignBuilder software and future climate data for the horizons of 2050 and 2080 based on the RCP 8.5 scenario and generated through Meteonorm 8 software and applied in simulations. For extensive parametric analysis, more than 10,000 scenarios were generated using the jEPlus tool and the LHS sampling method. Machine learning algorithms were used to reduce the computational load and extract nonlinear relationships between variables. Performance indicators including total annual energy consumption and hours of thermal discomfort were considered. The results showed that the XGBoost algorithm with determination coefficients of 0.998 for energy consumption and 0.995 for thermal discomfort hours provides the highest prediction accuracy. The analysis of the importance of features indicates that the orientation of the building and the thickness of the thermal insulation of the shell (among the variables examined) have the greatest impact. In addition, the results of supplementary analyses indicate that minor variations in certain design parameters can lead to significant changes in the thermal performance of buildings. This finding highlights the necessity of adopting multi‑variable design approaches and employing advanced analytical tools in the design process of future buildings. Overall, the findings of this study can serve as a scientific basis for the development of climate‑responsive design strategies in cold climatic regions.