| Mobile communication has reduced the amount of spectrum available, but many of these resources are still underutilized. E2SS methodologies for cognitive radio (CR) systems based on machine learning are presented in this paper. Before it can classify, we have to teach the classifier. Unsupervised classification algorithms such as K-means segmentation and the Variational model are used in cooperative spectrum sensing. Mixture designs like Gaussian mixtures, on the other hand, are unable to understand the relationships between data when dealing with large dimensional data. This paper's responsive Gaussian Mixture Model utilizes a mixture of Gaussian distributions to model each data value. A Gaussian is a user feature. As the means and standard deviations of each distribution evolve, the algorithm gains a better understanding of the relationships between classes of channels, both available and unavailable. The weight and variance of each distribution are used to determine this. Using the expectation maximising algorithm for reducing dimension data features, this method renews the data group in the feature space. Each classification technique's precision, recollection, F1-measure, categorization error, ROC curve, likelihood of detection, and false detection probability are evaluated.
Keywords: Noice pollution, Cooperative spectrum sensing, K-means clustering. |