search

firefly-cpp / NiaARM

A minimalistic framework for Numerical Association Rule Mining
MIT License
15 stars 6 forks source link

Enhancement Idea - Association Rule Mining on RDF Data (Semantics) #121

Open erkankarabulut opened 3 months ago

erkankarabulut commented 3 months ago

I am writing this to propose a possible extension to your library for mining association rules from Resource Description Framework (RDF) triples. So that association rule mining is possible with semantic datasets, such as knowledge graphs from DBpedia.

Semantic Association Rule Mining: This concept refers to mining semantic association rules on Semantic Web data, mostly in the form of RDF triples. The form of triples is (subject, predicate, object), for instance ("Joe Biden", isPresidentOf, "USA"). These triples come together and form a graph in the case of knowledge graphs, which is a hot topic in Semantic Web research.

There is a paper that uses ARM on RDF data: SWARM: An Approach for Mining Semantic Association Rules from Semantic Web Data. The goal is to detect structural and instance-level commonalities (formally, commonality in the ontological level, vs. data level).

I am not yet sure how would this be integrated into "optimization-based" NARM, but if you think this makes sense, then I can spend some time on this to find a way. What do you think? @firefly-cpp @zStupan

firefly-cpp commented 3 months ago

Thanks, @erkankarabulut. Definitely, we are always open to extending the framework with new enhancements and additions. Supporting different data has always been a primary goal of NiaARM, and thus, I totally support your idea. There is a chance to develop a new method for combining NARM and RDF data, which can result also in a research paper.

I have also been curious about rule mining in graphs using NARM. This "may be close" to your idea.

Altogether, I support your idea and ask you to find a way. :)

@zStupan?