认知无线电中基于支持向量机的频谱空闲度预测

SVM-based prediction of spectrum idle rate in cognitive radio systems

研究认知无线系统中的频谱小时空闲度预测问题,针对GSM系统的载频小时空闲度时间序列的非线性特点,提出一种基于支持向量机的预测模型构建方法。为提高模型的预测精度,在GSM系统小时空闲度时间序列特征分析的基础上,利用序列的节假日特性和日周期特性,对数据序列进行了重构。仿真结果表明,与采用基于神经网络的预测模型相比,该预测方法对工作日和周末均具有较高的预测精度,其预测绝对百分比误差在4以内。

A construction method of SVM-based prediction model for the spectrum idle rate in cognitive radio systems is proposed according to the non-linear characteristics of carrier frequency idle rate time sequence of GSM system. To enhance the prediction accuracy of the model, the data sequence is reconstructed according to the holiday and daily-cycle properties, and feature analysis of the idle rate time sequence of GSM system. The simulation results show that the proposed model has more accurate prediction in workday and weekend than the prediction model based on neural network. The absolute percentage error of its prediction is within 4.