The predictability study of channel state duration based on Hurst index

Spectrum prediction is one key enabling method for cognitive radio to improve spectrum utilization. Different from the traditional methods which predict the spectrum state slot-by-slot, in this paper we investigate the issue of prediction analysis of channel state duration (CSD). Specifically, we first introduce the concept of Hurst index to characterize the predictability between different scales of historical data and use the method of R/S (rescaled range) analysis the Hurst index of three different kinds of Large-Scale data and validate if we can obtain best prediction result from high predictability. Then, we introduce a pattern matching approach to validate the practical prediction performance. Furthermore, we focus on the predictability in small-scale data and find sometimes it performs even better than using large-scale data. Moreover, real world spectrum measurements are used to show that selecting historical data based on the predictability theory in this paper, we can smartly improve the predictive efficiency and enhance the prediction accuracy.