Title:	HYBRID SORT – A PATTERN-FOCUSED MATRIX REORDERING APPROACH BASED ON CLASSIFICATION
Author(s):	Celmar Guimarães da Silva
ISBN:	978-989-8704-21-4
Editors:	Yingcai Xiao, Ajith P. Abraham and Jörg Roth
Year:	2020
Edition:	Single
Keywords:	Reorderable Matrix, Heatmaps, Pattern Detection
Type:	Full
First Page:	35
Last Page:	43
Language:	English
Cover:
Full Contents:	click to dowload
Paper Abstract:	Matrix reordering algorithms permute rows and columns of an input matrix for unveiling patterns that are hidden due to inappropriate permutations. Most reordering algorithms are not pattern-focused and provide low-quality results for revealing some particular patterns. On the other side, few algorithms focus on revealing specific patterns with high-quality results, but they must know a priori that one of these patterns is hidden in the matrix. In this work, we define a matrix reordering algorithm that focuses on revealing a pattern from a pattern set while providing high-quality results. Based on permutation-invariant features of matrices, the algorithm Hybrid Sort uses an empirically-created classifier to classify an input matrix into a canonical data pattern. After that, the algorithm uses this classification to select a pattern-focused reordering algorithm to reorder the input matrix. Therefore, Hybrid Sort inherits the good output quality of the selected reordering algorithms. We report that the classifier reached F2-measures greater or equal to 85% to all patterns in our tests. Besides, we present some examples of this new method applied to synthetic and real-world matrices.
	Go Back