Machine-Learning-Assisted Antenna Optimization With Data Augmentation
Jiapeng Zhang, Jiawen Xu, Qiang Chen, Hui Li
Abstract
In this letter, a machine-learning-assisted antenna optimization method is proposed based on the random forest (RF) algorithm with data augmentation (DA). Using only a small number of samples, the prediction and optimization accuracy of the RF algorithm is ensured with repeated DA, which balances different types of samples during the training. With the proposed DA-RF method, the AR bandwidth of a circularly polarized omnidirectional base station antenna is optimized. By learning the relationship between the loop orientations and the AR bandwidth efficiently, the AR bandwidth is improved by 41% compared with the best one in the samples. The estimation accuracy of the proposed method outperforms other similar methods, with fewer iterations as well. The method is also successfully applied to multi-objective optimizations.