XGBGenA in Dual-Band Antenna Design Optimization

Antenna design optimization for modern wireless systems presents significant challenges due to the complexity of navigating vast design parameter spaces. Traditional approaches involving manual tuning and exhaustive simulations often falter in tackling these computational demands efficiently. This paper introduces XGBGenA, a novel antenna design optimization methodology leveraging machine learning to circumvent time-consuming simulations. XGBGenA synergistically combines the predictive capabilities of the XGBoost algorithm with the global search proficiency of genetic algorithms. Generative algorithm is used to find the best hyperparameters for XGBoost which will be later used for prediction of the operating frequencies. Comprehensive evaluations demonstrate XGBGenA's superiority over benchmarked techniques, achieving over 96% accuracy in predicting antenna metrics with Mean Square Error (MSE) of 0.1735 for first resonating frequency (Fr1) and 98.39% accuracy for second resonating frequency (Fr2) with an MSE value of 0.05356. The validated performance of XGBGenA on unseen antenna designs further demonstrates its potential for real-world applications in accelerating the development of optimized antennas for modern wireless systems.