In this work we presented an implementation that uses decision trees to learn highly accurate link specifications. We compared our approach with three state-of-the-art classifiers on nine datasets and showed, that our approach gives comparable results in a reasonable amount of time. It was also shown, that we outperform the state-of-the-art on four datasets by up to 30%, but are still behind slightly on average. The effect of user feedback on the active learning variant was inspected pertaining to the number of iterations needed to deliver good results. It was shown that we can get FScores
above 0.8 with most datasets after 14 iterations.
| Identifer | oai:union.ndltd.org:DRESDEN/oai:qucosa:de:qucosa:17168 |
| Date | 13 February 2018 |
| Creators | Obraczka, Daniel |
| Contributors | NGONGA, Axel-Cyrille, Universität Leipzig |
| Source Sets | Hochschulschriftenserver (HSSS) der SLUB Dresden |
| Language | English |
| Detected Language | English |
| Type | info:eu-repo/semantics/publishedVersion, doc-type:bachelorThesis, info:eu-repo/semantics/bachelorThesis, doc-type:Text |
| Rights | info:eu-repo/semantics/openAccess |
| Relation | urn:nbn:de:bsz:15-qucosa2-163403, qucosa:16340 |
Page generated in 0.002 seconds