Optimizable KNN and ANFIS Algorithms Development for Accurate Antenna Parameter Estimation

Abstract

The process of smart antenna synthesis involves the automatic selection of the optimal antenna type and geometry in order to enhance antenna performance. A model for intelligent antenna selection employs an optimizable K-nearest neighbors (KNN) classifier to determine the optimal antenna choice. To optimize the utilization of different learner types, the geometric parameters of the antenna are presented as the final step prior to the construction of the Adaptive Neuro-Fuzzy Inference System (ANFIS) model, which involves the integration of five distinct primary learners. The classification of three distinct types of antennas, namely helical antenna, pyramidal horn antenna, and rectangular patch antenna, is performed using an optimizable K-nearest neighbors (KNN) classifier. Additionally, an ANFIS approach is employed to determine the optimal size parameters for each antenna. The accuracy is used to evaluate the performance of an optimizable KNN classifier, whereas Mean Squared Error and Mean Absolute Percentage Error are used to evaluate the performance of an ANFIS. The proposed technique demonstrates high performance in parameter prediction and antenna categorization, achieving a Mean Absolute Percentage Error of less than 3% and an accuracy exceeding 99.16%. The recommended methodology holds significant potential for widespread application in the development of practical smart antennas.