Main Article Content
Spectrum sensing is of critical significance in cognitive radio (CR) systems. In this paper, a dependable spectrum sensing plan is proposed, which utilizes K-closest neighbor, a machine learning calculation. In the preparation stage, every CR client delivers a sensing report under changing conditions and, in view of a worldwide choice, either transmits or remains quiet. In the preparation stage the nearby choices of CR clients are joined through a larger part casting a ballot at the combination focus and a worldwide choice is come back to every CR client. A CR client transmits or remains quiet as per the worldwide choice and at every CR client the worldwide choice is contrasted with the genuine essential client activity, which is learned through an affirmation signal. In the preparation stage enough data about the encompassing condition, i.e., the activity of PU and the conduct of every CR to that activity, is accumulated and sensing classes framed. In the grouping stage, every CR client thinks about its present sensing report to existing sensing classes and separation vectors are determined.