Abstract
In this paper, we discuss the development of a hybrid multi-strategy book recommendation system using Linked Open Data. Our approach builds on training individual base recommenders and using global popularity scores as generic recommenders. The results of the individual recommenders are combined using stacking regression and rank aggregation. We show that this approach delivers very good results in different recommendation settings and also allows for incorporating diversity of recommendations.
This is a preview of subscription content, log in via an institution.
Buying options
Tax calculation will be finalised at checkout
Purchases are for personal use only
Learn about institutional subscriptionsNotes
- 1.
75,559 numeric ratings on 6,166 books (from 0–5, Task 1) and 72,372 binary ratings on 6733 books (Tasks 2 and 3), resp., from 6,181 users for training, and evaluation on 65,560 and 67,990 unknown ratings, resp. See http://challenges.2014.eswc-conferences.org/index.php/RecSys for details.
- 2.
- 3.
- 4.
- 5.
This includes types in the YAGO ontology, which can be quite specific (e.g., American Thriller Novels).
- 6.
The reason for not including broader categories by default is that the category graph is not a cycle-free tree, with some subsumptions being rather questionable.
- 7.
- 8.
We used the implementation available at http://www.dice4dm.com/.
- 9.
In general, it holds that the higher \(k_1\) and \(k_2\) the better, since this increases the number of covered feature dimensions and the diversity of the ensemble. However, comparably small values of \(k_1\) and \(k_2\), around 10 or 20 and maximally 100, are sufficient according to experiments by Zhang et al. [11] and Kong and Yu [4]. In our experiments, we tried to find a good balance between computational costs and predictive quality, and we report the combination which we used for our final recommendations.
- 10.
The reason is that the challenge uses the average rank w.r.t. F1 and ILD as a scoring function, which makes the selection of an optimal parameter strongly depend on the other participants’ solutions. It turned out that \(m=4\) optimized our scoring.
References
Di Noia, T., Mirizzi, R., Ostuni, V.C., Romito, D.: Exploiting the web of data in model-based recommender systems. In: Proceedings of the Sixth ACM Conference on Recommender Systems, RecSys ’12, pp. 253–256, ACM. New York (2012)
Di Noia, T., Mirizzi, R., Ostuni, V.C., Romito, D., Zanker, M.: Linked open data to support content-based recommender systems. In: Proceedings of the 8th International Conference on Semantic Systems, I-SEMANTICS ’12, pp. 1–8. ACM, New York (2012)
Heitmann, B., Hayes, C.: Using linked data to build open, collaborative recommender systems. In: AAAI Spring Symposium: Linked Data Meets Artificial Intelligence (2010)
Kong, X., Yu, P.S.: An ensemble-based approach to fast classification of multi-label data streams. In: CollaborateCom, pp. 95–104 (2011)
Mihelčić, M., Antulov-Fantulin, N., Bošnjak, M., Šmuc, T.: Extending rapidminer with recommender systems algorithms. In: RapidMiner Community Meeting and Conference (RCOMM 2012) (2012)
Ostuni, V.C., Di Noia, T., Mirizzi, R., Di Sciascio, E.: Top-n recommendations from implicit feedback leveraging linked open data. In: IIR, pp. 20–27 (2014)
Passant, A.: dbrec — music recommendations using DBpedia. In: Patel-Schneider, P.F., Pan, Y., Hitzler, P., Mika, P., Zhang, L., Pan, J.Z., Horrocks, I., Glimm, B. (eds.) ISWC 2010, Part II. LNCS, vol. 6497, pp. 209–224. Springer, Heidelberg (2010)
Paulheim, H., Ristoski, P., Mitichkin, E., Christian, B.: Data mining with background knowledge from the web. In: RapidMiner World (2014)
Schmachtenberg, M., Strufe, T., Paulheim, H.: Enhancing a location-based recommendation system by enrichment with structured data from the web. In: Web Intelligence, Mining and Semantics (2014)
Ting, K.M., Witten, I.H.: Issues in stacked generalization. J. Artif. Intell. Res. 10(1), 271–289 (1999)
Zhang, X., Yuan, Q., Zhao, S., Fan, W., Zheng, W., Wang, Z.: Multi-label classification without the multi-label cost. In: Proceedings of the 2010 SDM (2010)
Acknowledgements
The work presented in this paper has been partly funded by the German Research Foundation (DFG) under grant number PA 2373/1-1 (Mine@LOD).
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2014 Springer International Publishing Switzerland
About this paper
Cite this paper
Ristoski, P., Loza Mencía, E., Paulheim, H. (2014). A Hybrid Multi-strategy Recommender System Using Linked Open Data. In: Presutti, V., et al. Semantic Web Evaluation Challenge. SemWebEval 2014. Communications in Computer and Information Science, vol 475. Springer, Cham. https://doi.org/10.1007/978-3-319-12024-9_19
Download citation
DOI: https://doi.org/10.1007/978-3-319-12024-9_19
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-319-12023-2
Online ISBN: 978-3-319-12024-9
eBook Packages: Computer ScienceComputer Science (R0)