Advancing a Cascaded Machine Learning Approach for the Accurate Estimation of Antenna Parameters

This paper proposes the use of decision trees and linear regression to accurately predict the parameters of an antenna. Antennas remain a crucial component within RF systems. Computing pivotal antenna parameters such as the operational frequency, reflection coefficient (S11), antenna bandwidth, and radiation characteristics can prove to be a laborious endeavor. Obtaining these parameters necessitates meticulous optimizations in antenna design, consuming both time and human resources. In this study, we have undertaken an initial investigation into the utilization of machine learning regression techniques to predict antenna parameters, including the operating frequency, minimum value of S11 in dB, and bandwidth. A classical printed monopole antenna design is used in a case study. A three-step cascaded machine learning model trained on a four-dimensional initial feature set is used. Notably, Fine Gaussian support vector machines (SVM) provided the best result for S11 estimation, yielding a root mean square error (RMSE) of around 0.12, while fine decision trees provided the best results in the field of frequency optimization, resulting in an average RMSE of approximately 0.3. Finally, cubic SVMs resulted in an optimal RMSE of 0.009 for bandwidth approximation.