BIPLOT DENGAN DEKOMPOSISI NILAI SINGULAR BIASA DAN KEKAR UNTUK PEMETAAN PROVINSI BERDASARKAN PRESTASI MAHASISWA

Abstract

Biplot analysis can be constructed from  an ordinary and robust Singular Value Decomposition (SVD) approach.  A SVD approach ordinary need data matrix that no outliers or extreme data.  If a outliers data is found in this research, ordinary SVD biplot will not able to assure the mapping illustration between observation object and variable. To solve this problem, approaching method to robust SVD is applied.  The couple of  ordinary and robust SVD’ eigenvalue couple and eigenvector have distance minimizing feature among data matrix and estimation matrix with  Euclid norm or L2 norm and city block distance or L1 norm.  Biplot analysis which is based on  robust SVD method can give illustrations to a outliers resist observation object and variable.  The boxplot of the achievement of TPB IPB show that the data has outliers.  Robust and ordinary SVD to data matrix based biplot analysis shows the data graph and structure is almost identic.  It implies that outliers has no effect to data structure alteration.  A couple of extreme datas are given to the data matrix and then anlysised using ordinary and robust SVD. Ordinary SVD biplot result is far more different from data matrix biplot without extreme data, in the other hand robust SVD biplot shows small data structure which means it is not influenced by extreme data..