Enhancing Classification Accuracy in Cooperative Spectrum Sensing Using Machine Learning

In this article, we evaluate the ability of three algorithms (k-Nearest Neighbors (kNN), Partial Least Squares Discriminant Analysis (PLSDA), and K-means Clustering) to detect spectrum holes. The research results indicate that kNN clustering and PLSDA outperform K-means clustering when considering the different distances between Secondary Users (SU) and Primary Users (PU). In scenarios with high energy transmission, PLSDA and kNN exhibit excellent performance, with PLSDA showing balanced performance indicators, making it an ideal choice for high-energy environments. Future research will focus on other Machine Learning (ML) techniques to further improve the accuracy of spectrum sensing in various scenarios.