Analysis of Robust Principal Components Depends on the some methods of Projection-Pursuit


The analysis of the classic principal components are sensitive to the outliers where they are calculated from the characteristic values and characteristic vectors of correlation matrix or variance Non-Robust, which yields an incorrect results in the case of these data contains the outliers values. In order to treat this problem, we resort to use the robust methods where there are many robust methods Will be touched to some of them. The robust measurement estimators include the measurement of direct robust estimators for characteristic values by using characteristic vectors without relying on robust estimators for the variance and covariance matrices. Also the analysis of the principal components search for the trends of the highest scattered data projected on these vectors, but instead of using the variance as a measure for scattering, we will use robust measurement estimators as indicator for Projection-Pursuit. In this paper, we used Croux and Ruiz-Gazen algorithm, where the principal components are recognize by projection data on the highest vector for robust measurement estimators, focusing on the robust measurement to Qn and MAD.