<p>The prediction of damping ratio for sandwich composite structures is very important in terms of architectural and structural vibration performance enhancement. The solution to this task for real and practically useful core topologies is difficult to be achieved by currently available analytical or numerical tools because of geometrical and material nonlinearities. This study introduces a novel machine learning framework for the comprehensive prediction of damping properties in structured bio-based composites. Conventional analytical or numerical methods often struggle with the geometrical and material nonlinearities of complex designs. This approach systematically analyzes a wide array of core topologies, including honeycomb, auxetic, and tetrachiral geometries. It also evaluates various bio-based materials. By leveraging Random Forest, SVR, XGBoost, and Decision Tree regression algorithms, the significant potential of machine learning methods to accurately predict the damping ratio was demonstrated. Specifically, the XGBoost model achieved an impressive R² of 0.9826 in predicting the damping ratio across various core geometries. This demonstrates its superior accuracy in predicting energy dissipation performance. The integrated methodology offers an efficient way to address design complexities. It helps accelerate the development of high-performance, sustainable composite structures for architectural and engineering applications.</p>