Recent research has highlighted the potential of shape memory alloy (SMA) dampers in mitigating structural vibrations across various building models. However, the performance of these dampers is significantly influenced by structural parameters such as mass, stiffness, and the building’s time period. This study aims to study the potential of SMA dampers in torsionally coupled buildings by employing various machine learning techniques, specifically focusing on variations in the time period, frequency ratio, and eccentricity ratio of the structures. Initially, the theoretical framework of SMA dampers and the associated structural systems is discussed. Predictive models were developed using artificial neural network (ANN) and ensemble bagged tree (EBT) learning processes, with structural response data serving as the training dataset. The input variables included the time period, frequency ratio of torsional to lateral frequencies, and eccentricity ratio, while the output variables consisted of lateral and torsional displacements and accelerations. Moreover, this study compares these machine learning techniques to determine the most effective and reliable methods for predicting the structural response of buildings equipped with SMA dampers.