Data de-noise for discriminant analysis by using multivariate wavelets: (simulation with practical application)‎

Abstract EN

In this paper, we suggest using the multivariate wavelet analysis in higher dimensional space (Symlet, Daubechies' least-asymmetric wavelets) with soft thresholding to de-noise of the data (Shrinkage) before use it in the Discriminant Analysis to obtain more accurate and reliable results by comparing it with the Discriminant analysis used on data before de-noise.

And to know the effect of de-noise from data (proposed method) on Discriminant analysis results by simulating random data with normal distribution repeated 1000 times for different combinations of number of variables and sample sizes and real data represent leukemia patients taken from the Nanakele Hospital in Erbil.

We analyzed the data depending on MATLAB Language and statistics program (SPSS).

One of the most important conclusions reached by the researcher that use proposed method led to the separation between the two groups better than before de-noise (classical method) and this mean that data were classified for proposed method better than classical method.