Efficient Spectrum Utilization Through Optimized Learning Based Channel Selection Model

Channel selection based on learning plays a significant role in predicting channel quality for cognitive radio networks. However, due to the presence of various unwanted factors, reliable channel prediction accurately can be challenging. Our research paper includes a learning-based support-vector channel prediction model that performs reliable channel selection with minimal prediction errors. Moreover, we employ the particle swarm optimization technique for cost function optimization. The proposed optimization scheme in this paper helps to improve the resource selection accuracy. The training and testing of our proposed model have been performed by using influencing factors such as received signal strength indicators, packet loss ratios, and end-to-end delay to classify channels as unoccupied, partially occupied, or occupied. The proposed SVM-CP model has shown 87.5% accuracy in channel prediction. The learning model also demonstrates an improved prediction accuracy of 10–15% compared to well-known built-in classification models such as decision trees and random forests. As a result, our model significantly enhances spectrum allocation.