This work presents a metamaterial absorber designed for terahertz applications, featuring a dual-elliptical split-ring resonator geometry enclosed within a square frame. The structure is made on a polyimide substrate with metallic layers and exhibits six distinct absorption peaks at 3.45, 5.15, 7.01, 8.97, 10.50, and 10.89 terahertz, with corresponding absorptivity values of 94.2%, 94.2%, 98.06%, 99.45%, 85.87%, and 97.7%, respectively. At its lowest operating frequency, the proposed absorber is ultra-thin (0.068λ) and electrically compact (0.285λ), making it suitable for potential applications in stealth and sensing technologies. To improve multi-band absorption, machine learning models including Extra Tree, XGBoost, and an ensemble approach were employed for optimization. The ensemble model, is evaluated for enhancing absorber efficiency, with model performance validated through 10-fold cross-validation. Model accuracy was assessed using R-squared, Mean Absolute Error, and Root Mean Squared Error metrics. Ensemble approach was applied, with the ensemble achieving the highest R-squared value of 0.9944 for predicting the effects of geometric parameters.
P. Bhatt, P. Jain, A. Joshi, T. Razuvaeva and O. Kuznetsova, "Machine Learning-Driven Compact Hexa-Band THz Metamaterial Absorber," 2025 IEEE XVII International Scientific and Technical Conference on Actual Problems of Electronic Instrument Engineering (APEIE), Novosibirsk, Russian Federation, 2025, pp. 1-4, doi: 10.1109/APEIE66761.2025.11289231.
