Predicting Parameters of the Metamaterial Antennas based on Machine Learning Model

This paper presents a metamaterial antenna design method using Split Ring Resonator (SRR) structures based on the Machine Learning (ML) model using the Random Forest (RF) algorithm combined with the Gaussian Process (GP). Thanks to this combination, the optimal hyperparameters are found to improve the model’s prediction accuracy. The training and evaluation model dataset is the antenna dimensions and structure with 572 samples for nine features. The model’s output predicts the results of the bandwidth parameters and $\left|S_{11}\right|$ values of the metamaterial antenna. The proposed ML model’s effectiveness and dataset suitability are evaluated by Root Mean Squared Error (RMSE) and $R^{2}$ criterion, which reach 0.11 and 0.99, respectively. The proposed ML model archives a high prediction accuracy thanks to the combination of RF and GP and a suitable data set. This model’s application in antenna design significantly saves time when optimizing the dimensions of metamaterial antennas.