摘要
在变形回归分析处理前消除或减弱误差的影响,可以有效地改善回归分析的质量.通过一元回归试验对比可以看出,经过小波消噪以后,线性关系之外因素引起的数据波动性减小,拟合的效果得到了改善,线性关系显著增强.
The method of regression analysis is an important tool for deformation. The observed signal of deformation is often non-stationary. It possibly includes spike signal, mutation signal and non-stationary white noise. They often have high frequencies. The method of wavelet can detect error and remove it. It can eliminate or minimize the error's influence before deformation data analysis by regression. It can efficiently improve the quality of regression analysis. By comparison tests with monolinear regression, it can be found that the volatilities of data caused by factors without linear relation are reduced. The effect of fitting is improved. Linear relation is enhanced.
出处
《山东理工大学学报(自然科学版)》
CAS
2007年第3期40-42,共3页
Journal of Shandong University of Technology:Natural Science Edition
关键词
小波分析
回归分析
消噪
奇异性
wavelet analysis
regression analysis
de-noising
singularity