<p>The growing demand for intelligent system in soft robotics, biomedical devices, and aerospace systems has created an opportunity to build smart composite structures. However, addressing the shape memory properties remains a great challenge, due to its mutual dependency of material and processing parameters. The parameters considered includes composite type (unmodified, and modified nanoclay), number of shape memory alloy wire (1, 2, 3), glass fiber layers (2 and 4), training cycles (1, 2, 3, 4), and nanoclay loading (1wt%, 3wt%, 5wt%). A powerful optimization framework that blends genetic algorithms (GAs) and machine learning (ML) is presented in this study. Experimental dataset from design of experiments is used to train the ML models. Gaussian Process Regression performs exceptionally better with high prediction accuracy of R² = 94. 34%, 91.98% for shape recovery and fixity followed by other ML models. Combined contribution of parametric combination on shape memory properties were evaluated via surface response plots by highlighting that higher cycles improve fixity, while higher nanoclay and SMA levels enhance recovery. GA optimization revealed that the modified nanoclay with 5wt% loading, 3 SMA wire, 4 glass fiber layers, and 2 training cycles was the optimal configuration. This setup achieved superior performance with 86.7% and 88.4% for shape recovery and fixity respectively. Pareto front plot is used to visualize the trade-off relationship between the shape recovery and fixity presents the inverse relationship. The proposed ML-GA approach offers interpretable insights into structure–property links. This paradigm offers a broadly applicable method for data-driven functional material design.</p>