Dissertation
Investigation and Elimination of Substructures in Formal Concept Analysis focusing on Boolean Suborders and Subcontexts
Abstract
In the field of Formal Concept Analysis, data is mainly presented in so-called formal contexts, which assign to a set of objects their respective attributes. From those concept lattices can be generated, where the objects are grouped with respect to their common attributes to represent the relationships in the data in a way that enhances the understandability for humans. However, since a concept lattice can be of exponential size compared to its associated formal context, the presented relationships often become hard to grasp, even for data sets of moderate size. Therefore, the question arises of finding ways to reduce the size of a dataset to make it more understandable to the user while retaining the original information and structures as best as possible.
Since Boolean substructures are significantly responsible for the exponential size of concept lattices, whereas the objects in these structures just slightly differ concerning their attribute set, we consider them first in the present work. We give the possibility to infer from a Boolean subcontext in the formal context directly to a corresponding Boolean suborder in the associated concept lattice and vice versa. Next, we deal with reducing the size of the concept lattice. To this end, we consider two different types of feature selection in the formal context. Finally, we consider changes directly in the lattice. First, we give a way to collapse intervals (and thus Boolean suborders) by factorization while preserving (as many as possible of) the remaining elements. Second, we investigate under which conditions an interval can be entirely removed without changing anything in the rest of the structure.
Since Boolean substructures are significantly responsible for the exponential size of concept lattices, whereas the objects in these structures just slightly differ concerning their attribute set, we consider them first in the present work. We give the possibility to infer from a Boolean subcontext in the formal context directly to a corresponding Boolean suborder in the associated concept lattice and vice versa. Next, we deal with reducing the size of the concept lattice. To this end, we consider two different types of feature selection in the formal context. Finally, we consider changes directly in the lattice. First, we give a way to collapse intervals (and thus Boolean suborders) by factorization while preserving (as many as possible of) the remaining elements. Second, we investigate under which conditions an interval can be entirely removed without changing anything in the rest of the structure.
Citation
@phdthesis{doi:10.17170/kobra-202307148371,
author={Koyda, Maren},
title={Investigation and Elimination of Substructures in Formal Concept Analysis focusing on Boolean Suborders and Subcontexts},
school={Kassel, Universität Kassel, Fachbereich Elektrotechnik / Informatik},
year={2023}
}
0500 Oax 0501 Text $btxt$2rdacontent 0502 Computermedien $bc$2rdacarrier 1100 2023$n2023 1500 1/eng 2050 ##0##http://hdl.handle.net/123456789/14959 3000 Koyda, Maren 4000 Investigation and Elimination of Substructures in Formal Concept Analysis focusing on Boolean Suborders and Subcontexts / Koyda, Maren 4030 4060 Online-Ressource 4085 ##0##=u http://nbn-resolving.de/http://hdl.handle.net/123456789/14959=x R 4204 \$dDissertation 4170 5550 {{Datenkompression}} 5550 {{Formale Begriffsanalyse}} 7136 ##0##http://hdl.handle.net/123456789/14959
2023-08-01T11:17:56Z 2023-08-01T11:17:56Z 2023 doi:10.17170/kobra-202307148371 http://hdl.handle.net/123456789/14959 eng Urheberrechtlich geschützt https://rightsstatements.org/page/InC/1.0/ Formale Begriffsanalyse Datenreduktion 004 Investigation and Elimination of Substructures in Formal Concept Analysis focusing on Boolean Suborders and Subcontexts Dissertation In the field of Formal Concept Analysis, data is mainly presented in so-called formal contexts, which assign to a set of objects their respective attributes. From those concept lattices can be generated, where the objects are grouped with respect to their common attributes to represent the relationships in the data in a way that enhances the understandability for humans. However, since a concept lattice can be of exponential size compared to its associated formal context, the presented relationships often become hard to grasp, even for data sets of moderate size. Therefore, the question arises of finding ways to reduce the size of a dataset to make it more understandable to the user while retaining the original information and structures as best as possible. Since Boolean substructures are significantly responsible for the exponential size of concept lattices, whereas the objects in these structures just slightly differ concerning their attribute set, we consider them first in the present work. We give the possibility to infer from a Boolean subcontext in the formal context directly to a corresponding Boolean suborder in the associated concept lattice and vice versa. Next, we deal with reducing the size of the concept lattice. To this end, we consider two different types of feature selection in the formal context. Finally, we consider changes directly in the lattice. First, we give a way to collapse intervals (and thus Boolean suborders) by factorization while preserving (as many as possible of) the remaining elements. Second, we investigate under which conditions an interval can be entirely removed without changing anything in the rest of the structure. open access Koyda, Maren 2023-07-12 xii, 140 Seiten Kassel, Universität Kassel, Fachbereich Elektrotechnik / Informatik Stumme, Gerd (Prof. Dr.) Ganter, Bernhard (Prof. Dr.) Datenkompression Formale Begriffsanalyse publishedVersion false true
The following license files are associated with this item:
:Urheberrechtlich geschützt