This project is the implementation of my Bachelor Thesis and the poster contribution
"Auswertung der Performanz der Klassifizierung von Entscheidungsbäumen auf Grundlage von Linked Data und der Vergleich zu
einem aktuellen Ansatz im Bereich der Grapheinbettung"
"Binary Vector based Propositionalization Strategy for Multivalued Relations in Linked Data"
for the ISWC2017 conference.
In order to reproduce the results we like to share this code. The two main classes can be found under the following path
EdtEvaluation/src/test/java/de/uni/due/paluno/dt/linkeddata/expansion
For expanding the rdf graphs the Class
Expanding.java
is used. This class is supposed to be run first. It produces various features. We pay special attention to multi-valued attribues. Those are then further processed into a binary vector reprenstaion.
The main step of transforming the multivalued attributes into a binary vector is handled by
MuliValueFormatter.java
For loading the DBPedia entries Ristoski's and Paulheim's code was used. It is referenced in their paper
"RDF2Vec: RDF Graph Embeddings for Data Mining"
and can be found here
http://data.dws.informatik.uni-mannheim.de/rdf2vec/