Paper Info
Title
Decomposition of Boolean multi-relational data with graded relations
Abstract
Decomposition (or factorization) of Boolean multi-relation data, i.e. data in the form of Boolean matrices, together with relation between them, containing zeros and ones only, received a considerable attention in data mining research. The main aim is to find new variables-factors-hidden in data that explain data. The main advantage of Boolean data is interpretability. In this paper we argue that considering only Boolean data can be limiting. Especially the relation between input matrices is not necessarily of a Boolean nature. Usually this relation represents linkages to some degree, e.g. how much a user likes or dislikes a movie. Using Boolean method for such data-data must be somehow binarized first-leads to a loss of information. First, we reformulate decomposition problem for multi-relational data with ordinal relations. Then we propose a new algorithm for such data along with an experimental evaluation.
Year
DOI
Venue
2016
10.1109/IS.2016.7737425
2016 IEEE 8th International Conference on Intelligent Systems (IS)
Keywords
Field
DocType
matrix decomposition,factor analysis,formal concept analysis,multi-relational data
Maximum satisfiability problem,Boolean network,Boolean circuit,Computer science,Theoretical computer science,Product term,Standard Boolean model,Boolean data type,Boolean expression,Two-element Boolean algebra
Conference
ISBN
Citations 
PageRank 
978-1-5090-1355-5
0
0.34
References 
Authors
8
2
Name
Order
Citations
PageRank
Martin Trnecka132.10
Marketa Trneckova233.46