Predicting S-Parameters in Microstrip Trisection Passband Filter Using a S-KNN Algorithm

Abstract

In the field of microwave modelling and design the Supervised K-Nearest Neighbor (S-KNN) algorithm has emerged as a valuable tool. An S-KNN based approach to the modelling of passband trisection filter components with emphasis on the optimization of the hyperparameter K is presented in this paper. Our technique introduces a novel S-KNN topology specifically designed for parametric modeling of microwave components with S-parameters as outputs. Unlike previous methods which often rely on manual parameter tuning or lack robust hyperparameter optimization our optimized S-KNN model achieves high accuracy (0.9429) and low mean squared error (0.0109) by tuning K to the specific dataset. The study compares various distance metrics and employs Euclidean distance for predictions. Results demonstrate that the optimized S-KNN model achieves strong alignment with electromagnetic (EM) simulations with an average accuracy of 94.29% offering a faster and potentially more efficient alternative to traditional design techniques. © 2024 Elsevier B.V. All rights reserved.