Ensembled Machine Learning Models for Antenna Optimization in Wireless Communication/ Biomedical Applications

In this work, the antenna optimisation process is done by using ML algorithms such as KNN, random forest, Gradient Boost, XG Boost and MLP. Then the predicted results are ensembled using Voting Regressor. A Cylindrical Dielectric Resonator Antenna (CDRA) was designed and simulated in CST MWS’23 for this purpose. Dataset is created by varying the radius of the DR using parametric sweep function of the EM simulator. A close prediction between actual values and the predicted ones is obtained using these ML models. It is observed that by ensembling these varied approaches, a model that can handle different types of relationships in the data, can be created. This algorithm can be further utilized for optimisation of other antenna parameters and thus helping in reducing time in optimising complex antenna designs. The accuracy achieved by the ensemble model is 98.6% while maintaining a small MSE of 0.001. The optimized design shows stable characteristics within the ISM band and hence can be used for biomedical as well as wireless communication applications.