Capacity-Driven Smart Skin Loads Selection Utilizing KNN and Gradient Boosting

In this paper, we present a method for engineering loads terminated to smart skins across multiple operating angles using k-nearest neighbors (KNN) and gradient boosting (GB) regression algorithms. Modelling the smart skin as a multiport scattering structure loaded by the passive loads, we formulate requirements for the structure following the capacity-driven approach to smart radio environment design. The load selection process is framed as a regression task, addressed through the proposed machine learning algorithms. To validate the practicality of the method, we evaluate the performance of various regressors applied to a sample scatterer, demonstrating their effectiveness under different operational constraints.