Geometry Prediction of the Antenna Design Using Machine Learning Method

To design and optimise an antenna for a specific pur-pose, engineers combine engineering concepts, electromagnetic theory and practical factors that take much longer. However, in the midst of machine learning methods, this work can be done effectively and efficiently. Therefore, this paper is concerned with predicting the geometry of a microstrip patch antenna using machine learning. Parameter sweeps for the patch antenna geometry are proposed and obtained from a microwave simulator with varying width values to be used as a dataset. The data collected for the design was fed into a support vector machine algorithm to verify that the support vector regression modelling was correct. With an average prediction error of less than 0.0903 and a correlation coefficient of 0.98, the regression model predicted the results very accurately. The time taken by the machine learning method was shorter than that required by conventional methods.