Spectrum Scheduling Classification Using Conditional Probability and a Decision Tree Supervised Learning Approach
AbstractSpectrum Scheduling is an efficient scheme of improving spectrum utilization for faster communications, higher definition media (HDM) and data transmission. Radio spectrum is very limited in supply resulting in enormous problems related to scarcity. It owes the physical support for wireless communication, both fixed applications and mobile broadband. Basically, effective use of the spectrum depends on the channel settings, sensing performance, detection of spectrum prospect as well as effective transmission of both Primary Users (PUs) and Secondary Users (SUs) packets at a specific time slot. In order to improve spectrum utilization this paper adopted quantitative method which employs Probability Theorem to identify the probabilities of both primary Users (PUs) and secondary users (SUs) in the spectrum datasets allocation and further used conditional probability to compare two Frequency Bands i.e., High Frequency (HF) and Very High Frequency (VHF). The result indicates available spectrum holes (SH) left unutilized in the Secondary User (SU) resulting in the need for spectrum scheduling for the SU. The procedure makes the secondary users occupy a probability of 0.002mhz compared to the primary users on 0.00004mhz utilization. This further indicates that some spectrum holes were left unutilized by the license users (Primary Users). However, spectrum allocation is one of the major issues of improving spectrum efficiency and has become a considerable tool in cognitive wireless networks (CWN). Consequently, the goal of spectrum allocation is to assign leisure spectrum resources efficiently to achieve the optimal Quality of Service (QOS and cognitive user requirements of wireless network. Again, classification of spectrum allocation was carried out through difference methods. Firstly, we employ a probability theorem to identify the probability of both Primary Users (PUs) and Secondary Users (SUs) in the allocated spectrum data sets. Secondly, conditional probability was used to compare two frequency band based on primary and secondary allocation policies designed to identify the specific allocation of each band. Thirdly, Machine Learning (ML) Algorithm based on Decision Tree-Supervised Learning (DTSL) approach was adopted to classified our data sets. The result yielded 68% which correctly classified instances based on the total records of sixty-nine (69) data sets. Research findings demonstrate a highly optimized spectrum scheduling for efficient networks service provisions.