Geopolymer concrete (GPC) is regarded as a more environmentally friendly construction material compared to conventional cement concrete, and its exceptional environmental capabilities are highly favored by the contemporary construction sector. Studying the mechanical properties of GPC upon exposure to elevated temperatures is a crucial aspect of evaluating structural damage and enhancing fire safety measures. Nevertheless, properly predicting the compressive performance of GPC upon exposure to high temperatures remains a formidable task. This study employs various machine learning techniques, such as single models, integrated models, neural network models, and hybrid models, to predict the compressive strength of GPC from room temperature to 1000°C. The results of each model are summarized, and the significant factors influencing compressive strength are analyzed to evaluate the thermal behavior of GPC. These findings offer recommendations for future in‐depth machine learning applications in the GPC field. The K‐fold cross‐validation shows that the hybrid model genetic algorithm–random forest has the highest prediction accuracy, while the single model performs the worst. Other models also provide favorable prediction results. The feature importance analysis revealed that the compressive strength of GPC is primarily influenced by heating temperature (HT) and hydroxide ion concentration, with fly ash and ground granulated blast furnace slag content being secondary factors. The partial dependence plot‐2D analysis indicates that as HTs increase, the influence of other variables on GPC compressive strength decreases significantly. These findings can inform the design of GPC mixing ratios for high‐temperature exposure. The machine learning technique proposed in this study accurately predicts GPC compressive strength across various temperatures, reducing experimental time and costs while promoting the GPC sector.
