脉冲响应不变解卷积Web数据奇异特征挖掘

Web Data Singular Feature Mining Based on Impulse Response Invariant Deconvolution

针对Web数据奇异特征挖掘时,信道对奇异特征具有卷积效应,影响特征挖掘精度。提出一种脉冲响应不变解卷积算法,实现对奇异特征对的盲解卷积,提高奇异特征挖掘性能。利用Web数据奇异特征的时间可预测性作为盲解卷积的解卷测度,采用脉冲响应不变算法对基于该测度的代价函数进行优化求解,从而成功得到解卷积滤波器系数,实现对Web数据奇异特征的盲解卷积。仿真实验表明,采用该算法挖掘Web数据的奇异特征,对于奇异特征信号具有很好的盲解卷积效果,所挖掘的奇异特征相关系数和重构信噪比均较高,特征挖掘聚类性好。

In Web data singular feature mining, the channel has the convolution effect on singular feature, feature mining ac-curacy is influenced. An improved pulse response invariant deconvolution algorithm was proposed, the singular feature s of blind deconvolution was improved, and the performance of singular feature mining was good. Using the Web data of singu-lar characteristics of the time predictable volume measurement solutions as blind deconvolution, and pulse response invari-ant algorithm was proposed to optimize the cost function based on the measure, thus deconvolution filter coefficients were calculated, the Web data of singular feature of blind deconvolution was realized. Simulation results show that it has the very good blind deconvolution effect for the singular signal, the evaluation index such as reconstruction SNR is high, and the fea-ture mining clustering performance is good.