Spectrum Occupancy Classification Using SVM-Radial Basis Function

Abstract

With recent development in wireless communication, efficient spectrum utilization is major area of concern. Spectrum measurement studies conducted by wireless communication researchers reveals that the utilization of spectrum is relatively low. In this context, we analyzed big spectrum data for actual spectrum occupancy in spectrum band using different machine learning techniques. Both supervised [Naive Bayes classifier (NBC), K-NN, Decision Tree (DT), Support Vector Machine with Radial Basis Function (SVM-RBF)] and unsupervised algorithms [Neural Network] are applied to find the best classification algorithm for spectrum data. Obtained results shows that combination of SVM-RBF is the best classifier for spectrum database with highest classification accuracy appropriately for distinguishing the class vector in the busy and idle state. We made analysis-based on empirical SVM-RBF model to identify actual duty cycle on the particular band across four mid-size location at Ahmedabad Gujarat.