A Matrix Completion Based Spectrum Prediction Scheme under Incomplete Data

Abstract

Spectrum prediction is a technology to estimate the channel status by mining and using the correlation of the spectrum data. Using a reliable prediction scheme, the performance of opportunity spectrum access can be more effective. However, the traditional study on the spectrum prediction methods only use the correlation of the historical data in time domain, and on the basis of the spectrum historical data is complete. In reality, the historical data are often incomplete. The reality calls for a deeper understanding of the correlation of current spectrum data. In this paper, we present a correlation study, with data collected by measure both on the time domain and spectrum domain. Main findings include that the channel status information has correlation not only in the time domain, but also in the spectrum domain. We then exploit such time and spectrum correlation to develop a matrix completion based spectrum prediction scheme under the incomplete data that can predict channel availability based on past observations with considerable accuracy.