Shape Memory Alloy (SMA) dampers have gained significant attention for their potential to mitigate structural vibrations in buildings. Recent studies have high-lighted the effectiveness of SMA dampers in various building models. However, the performance of these dampers is also influenced by structural parameters such as the time period, mass, and stiffness of the building. This study develops different machine learning techniques to evaluate the effectiveness of SMA dampers in torsionally coupled buildings by varying the time period, frequency ratio, and eccentricity ratio of the structures. Initially, the theoretical background of SMA dampers and the structural system is presented. Artificial Neural Network (ANN) and Decision Tree (DT) learning process were employed to create predictive models. The structural response data served as the training dataset. Input parameters included the time period, frequency ratio of torsional to lateral frequencies, and eccentricity ratio of the structure, while output parameters comprised lateral and torsional displacements, as well as lateral and torsional accelerations. Additionally, this study compares these machine learning techniques to identify the most effective and robust methods for predicting the structural response of buildings equipped with SMA dampers. The results indicate that the SMA wire damper effectively reduces the peak lateral displacement of the structure by approximately 22%. Furthermore, the decision tree (DT) model demonstrates strong predictive performance, achieving an R-value of up to 0.98.