Packet Scheduling Using SVM Models in Wireless Communication Networks

Abstract

This paper addresses a support vector machine (SVM) model-based packet scheduling in wireless communication networks. The linear SVM, as well as the polynomial SVM model-based proportional fair scheduling (SVM-PFS), are suggested. Moreover, their performances are analyzed for various system and machine learning parameters such as the average window, signal to interference ratio (SIR) and the degree of a polynomial. From the computer simulations, the performance measurements of PFS, such as, user selection fairness and user throughput are calculated, which proves that performance of SVM-PFS approaches to that of the conventional metric-based PFS in most cases, while the simulation results show that the polynomial SVM-PFS outperforms the linear SVM-PFS. It is also shown that the SVM-PFS performs better under higher SIR or larger average window.