Seriation and Matrix Reordering Methods for Asymmetric One-Mode Two-Way Datasets Innar Liiv and Leo Vohandu Abstract Analyzing asymmetric datasets and finding novel insights from traces of asymmetries is a non-trivial challenge in data analysis. We propose a novel use of seriation and matrix reordering methods to find insights from asymmetric one-mode two-way datasets. This article addresses the following research questions: How to use seriation and matrix methods with asymmetric one-mode two-way matrices? What insights and patterns can be found from asymmetric structure using such an approach? 1 Introduction Seriation is an exploratory data analysis technique to reorder objects into a sequence along a one-dimensional continuum so that it best reveals regularity and patterning among the whole series (Liiv [8]). Seriation if often called matrix reordering, when applied to two-way datasets. The scope of this paper is limited to asymmetric entity to entity data tables. Using Tucker’s terminology (Tucker [22]) and Carroll-Arabie taxonomy (Carroll & Arabie [2]), we focus on one-mode two-way (N × N) data tables. Definition Seriation can be defined as a combinatorial optimization problem for minimizing a loss function L on a matrix A using permutation matrices  and  for reordering the rows and columns in a way that maximizes the local and global patterns (Liiv [7, p. 51]): argmin , L (A) (1) Seriation methods can be applied to analyze asymmetric one-mode two-way datasets as presented on Fig. 1. Fionn Murtagh has very eloquently called such a data analysis I. Liiv (B ) · L. Vohandu Tallinn University of Technology, Tallinn, Estonia e-mail: innar.liiv@taltech.ee L. Vohandu e-mail: leo.vohandu@taltech.ee © Springer Nature Singapore Pte Ltd. 2020 T. Imaizumi et al. (eds.), Advanced Studies in Behaviormetrics and Data Science, Behaviormetrics: Quantitative Approaches to Human Behavior 5, https://doi.org/10.1007/978-981-15-2700-5_10 159