Authors: Bruno Figueiredo Medina, Celmar Guimarães da Silva
Abstract: Providing good permutations of rows and columns of a matrix is necessary in order to enable users to understand underlying patterns hidden on its data. A previous work provided a preliminary version of an algorithm based on matrix binarization and PQR trees for reordering quantitative data matrices. This work proposes an improved version of this algorithm. It uses smoothing as an intermediary step for enhancing the capability of finding good permutations, without modifying original matrix data. The work exemplifies the potential of this technique by reordering scrambled synthetic matrices with underlying patterns.