Abstract
Accurate characterization of microwave dielectric materials is essential for the effective design and optimization of high-frequency electronic systems, since material properties critically influence system performance. This research presents a novel Internet of Things (IoT)-integrated sensing architecture that synergistically combines unsupervised$\mathbf{k}$-means clustering with ensemble learning (EL) algorithms for robust extraction of complex permittivity from scattering parameter measurements. Unlike conventional cavity perturbation techniques limited by linear assumptions, the proposed hybrid framework effectively captures non-linear perturbation regimes through data-driven modeling. The system employs a cylindrical cavity resonator integrated with an IoT layer for automated data acquisition and remote analytics. The hybrid ML workflow enhances model robustness by restructuring the feature space via clustering prior to EL-based regression, improving predictive accuracy for both simulated and experimental datasets. Evaluation results confirm high accuracy and generalization capability even under extrapolated conditions, demonstrating the scalability of this smart sensing framework for industrial and scientific deployment.