Increasing Prediction Performance Using Weighting Methods in Multi-Criteria Item-Based Collaborative Filtering

Abstract

Recommender systems have been popularly applied in various domains such as e-commerce, tourism, movie, music, and restaurants in recent years. Although researchers have introduced various algorithms, collaborative filtering is one of the most widely used algorithms in recommender systems. Collaborative filtering aims to recommend items that users might like in the future by taking into consideration the past preferences of users. In existing single criteria systems, users are expected to give a single rating about the items. However, a single criterion may not reflect the user's opinion. Thus, multi-criteria collaborative filtering systems have been introduced. Multiple criteria rating instead of a single criterion can increase the accuracy of future recommendations especially in domains such as movies, hotels, and restaurants. The first step in multi-criteria collaborative filtering systems is to find users or items similar to a user who asks for a recommendation. There are similarity-based methods commonly used in the literature to calculate similarity. While calculating the similarity in these methods, the ones that are co-rated between users / items are used. Although the number of co-rated items / users is very small, the correlation between them might be calculated very high. High correlation values may not always guarantee the best neighbors. Given these disadvantages, high correlation values might prevent reliable and accurate predictions. We propose to improve existing similarity calculations in order to increase performance in similarity-based approaches in multi-criteria item-based collaborative filtering. As the number of common users rating two items increases, the similarity calculated between them becomes more reliable. Therefore, it is recommended to use Jaccard and significance-weighting methods used in traditional collaborative filtering as weighting methods in the similarity calculation process in multi-criteria systems. With the proposed weighting methods, the aim is to decrease the calculated similarity as the number of users who rated both items decreases. Weighting methods are integrated into existing similarity calculation processes and it is aimed to increase neighbor selection and prediction performance. The proposed methods have been tested using three different versions of the Yahoo!Movies dataset. Experiments show that the proposed methods have greatly improved the prediction performance and coverage values compared to existing methods.

Keywords

• Multi-criteria
• Collaborative filtering
• Significance-weighting
• Jaccard similarity
• Item-based

Çok Kriterli Ürün-Tabanlı İşbirlikçi Filtrelemede Ağırlıklandırma Yöntemlerini Kullanarak Tahmin Performansının Arttırılması

Öz

Öneri sistemleri son yıllarda e-ticaret, turizm, film, müzik ve restoran gibi çeşitli alanlarda popüler olarak uygulanmaya başlanmıştır. Araştırmacılar çeşitli algoritmalar geliştirmelerine rağmen işbirlikçi filtreleme öneri sistemlerinde en yaygın kullanılan algoritmalardan biridir. İşbirlikçi filtreleme ile kullanıcıların geçmiş tercihleri göz önünde bulundurularak gelecekte kullanıcıların beğenebileceği ürünleri önermesi hedeflenir. Mevcut tek kriterli sistemlerde kullanıcıların ürünler hakkında tek bir derecelendirme vermesi beklenir. Fakat tek bir kriter kullanıcının ürünler hakkında fikrini yansıtmayabilir. Bu nedenle çok kriterli işbirlikçi filtreleme sistemleri geliştirilmiştir. Özellikle film, otel ve restoran gibi alanlarda kullanıcılar tek bir kritere göre derecelendirme vermek yerine birden çok kritere göre beğenilerini sunmaları onlara gelecekte yeni ürünler önermekteki başarıyı arttırabilir. Çok kriterli işbirlikçi filtreleme sistemlerindeki ilk aşama öneri isteyen bir kullanıcıya benzer en yakın kullanıcıları veya ürünleri bulmaktır. Literatürde benzerlik hesaplamak için yaygın kullanılan benzerlik-tabanlı metotlar mevcuttur. Bu metotlarda benzerlik hesaplanırken kullanıcılar / ürünler arasında ortak derecelendirilen ürünlerin / kullanıcıların verileri kullanılır. Fakat ortak derecelendirilen ürünlerin / kullanıcıların sayısı çok az olmasına rağmen aralarındaki korelasyon çok yüksek hesaplanabilir. Yüksek korelasyon değerleri her zaman en iyi komşular olduğunu garanti etmeyebilir. Bu dezavantajlar göz önüne alındığında yüksek korelasyon değerleri her zaman güvenilir ve doğru tahminler elde edilmesini engelleyebilir. Makalemizde çok kriterli ürün-tabanlı işbirlikçi filtrelemede benzerlik tabanlı yaklaşımlardaki performans artışını sağlamak için mevcut benzerlik hesaplamalarını iyileştirmeyi hedefliyoruz. İki ürünü oylayan ortak kullanıcı sayısı arttıkça, iki ürün arasında hesaplanan benzerlikte daha güvenilir olacaktır. Bu nedenle, geleneksel işbirlikçi filtrelemede kullanılan Jaccard ve önem ağırlıklandırma yöntemlerini çok kriterli sistemlerde benzerlik hesaplama sürecinde ağırlıklandırma yöntemleri olarak kullanılması önerilir. Önerilen ağırlıklandırma yöntemleri ile amaç, her iki ürünü de derecelendirme veren kullanıcı sayısı azaldıkça hesaplanan benzerliği azaltmaktır. Ağırlıklandırma yöntemleri, mevcut benzerlik hesaplama işlemlerine entegre edilerek komşu seçimi ve tahmin performansı arttırılması hedeflenir. Önerilen yöntemler Yahoo!Movies veri setinin üç farklı versiyonu kullanılarak test edilmiştir. Yapılan deneyler gösteriyor ki, önerilen metotlar mevcut metotlara göre tahmin performansını ve kapsam değerlerini büyük oranda arttırmıştır.

Anahtar Kelimeler

• Ürün-tabanlı
• Jaccard benzerliği
• Önem-ağırlıklandırma
• İşbirlikçi filtreleme
• Çok kriterli

References

1. Adomavicius, G., ve Tuzhilin, A. (2005). Toward the next generation of recommender systems: A survey of the state-of-the-art and possible extensions. IEEE transactions on knowledge and data engineering, 17(6), 734-749. doi: 10.1109/TKDE.2005.99.
2. Adomavicius, G., ve Kwon, Y. (2007). New recommendation techniques for multicriteria rating systems. IEEE Intelligent Systems, 22(3), 48–55. https://doi.org/10.1109/MIS.2007.58.
3. Adomavicius, G., Manouselis, N., ve Kwon, Y. (2011). Multi-Criteria Recommender Systems. Recommender Systems Handbook, 769–803. https://doi.org/10.1007/978-0-387-85820-3_24.
4. Aggarwal, C. C. (2016). Model-Based Collaborative Filtering. Recommender Systems, 71–138. https://doi.org/10.1007/978-3-319-29659-3_3.
5. Arazy, O., Kumar, N., ve Shapira, B. (2009). Improving Social Recommender Systems. IT Professional, 11(4), 38–44. https://doi.org/10.1109/mitp.2009.76.
6. Bag, S., Kumar, S. K., ve Tiwari, M. K. (2019). An efficient recommendation generation using relevant Jaccard similarity. Information Sciences, 483, 53–64. https://doi.org/10.1016/j.ins.2019.01.023.
7. Batmaz, Z., ve Kaleli, C. (2019). AE-MCCF: An Autoencoder-Based Multi-criteria Recommendation Algorithm. Arabian Journal for Science and Engineering, 44(11), 9235–9247. https://doi.org/10.1007/s13369-019-03946-z.
8. Bilge, A., ve Kaleli, C. (2014). A multi-criteria item-based collaborative filtering framework. 11th International Joint Conference on Computer Science and Software Engineering (JCSSE), 18–22. https://doi.org/10.1109/JCSSE.2014.6841835.

9. Bilge, A., ve Yargıç, A. (2017). Improving Accuracy Of Multi-Criteria Collaborative Filtering By Normalizing User Ratings. Anadolu University Journal Of Science And Technology A- Applied Sciences and Engineering, 18(1), 225–237. https://doi.org/10.18038/aubtda.273802.
10. Candillier, L., Meyer, F., ve Fessant, F. (2008). Designing specific weighted similarity measures to improve collaborative filtering systems. Advances in Data Mining. Medical Applications, E-Commerce, Marketing, and Theoretical Aspects, 242–255. https://doi.org/10.1007/978-3-540-70720-2_19.
11. Chae, D.K., Lee, S.C., Lee, S.Y., ve Kim, S.W. (2018). On identifying k -nearest neighbors in neighborhood models for efficient and effective collaborative filtering. Neurocomputing, 278, 134–143. https://doi.org/10.1016/j.neucom.2017.06.081.
12. Goldberg, D., Nichols, D., Oki, B. M., ve Terry, D. (1992). Using collaborative filtering to weave an information tapestry. Communications of the ACM, 35(12), 61-70. https://doi.org/10.1145/138859.138867.
13. Gomez-Uribe, C. A., ve Hunt, N. (2015). The Netflix Recommender System. ACM Transactions on Management Information Systems, 6(4), 1–19. https://doi.org/10.1145/2843948.
14. Herlocker, J. L., Konstan, J. A., Borchers, A., ve Riedl, J. (1999). An algorithmic framework for performing collaborative filtering. Proceedings of the 22nd Annual International ACM SIGIR Conference on Research and Development in Information Retrieval - SIGIR ’99. https://doi.org/10.1145/312624.312682.
15. Herlocker, J., Konstan, J. A., ve Riedl, J. (2002). An empirical analysis of design choices in neighborhood-based collaborative filtering algorithms. Information retrieval, 5(4), 287-310. doi: 10.1023/A:1020443909834.
16. Jannach, D., Karakaya, Z., ve Gedikli, F. (2012). Accuracy improvements for multi-criteria recommender systems. Proceedings of the 13th ACM Conference on Electronic Commerce- EC ’12, 674. https://doi.org/10.1145/2229012.2229065.
17. Kaleli, C. (2014). An entropy-based neighbor selection approach for collaborative filtering. Knowledge-Based Systems, 56, 273–280. https://doi.org/10.1016/j.knosys.2013.11.020.
18. Kim, T. H., ve Yang, S. B. (2007). An effective threshold-based neighbor selection in collaborative filtering. In European Conference on Information Retrieval, 712-715.
19. Konstan, J. A., Miller, B. N., Maltz, D., Herlocker, J. L., Gordon, L. R., ve Riedl, J. (1997). GroupLens: applying collaborative filtering to Usenet news. Communications of the ACM, 40(3), 77-87. https://doi.org/10.1145/245108.245126.
20. Koren, Y., Bell, R., ve Volinsky, C. (2009). Matrix factorization techniques for recommender systems. Computer, 42(8), 30-37. doi: 10.1109/MC.2009.263.
21. Linden, G., Smith, B., ve York, J. (2003). Amazon.com recommendations: item-to-item collaborative filtering. IEEE Internet Computing, 7(1), 76–80. https://doi.org/10.1109/mic.2003.1167344.
22. Ma, H., King, I., ve Lyu, M. R. (2007). Effective missing data prediction for collaborative filtering. Proceedings of the 30th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval- SIGIR ’07, 39. https://doi.org/10.1145/1277741.1277751.
23. Nilashi, M., bin Ibrahim, O., Ithnin, N., ve Sarmin, N. H. (2015). A multi-criteria collaborative filtering recommender system for the tourism domain using Expectation Maximization (EM) and PCA–ANFIS. Electronic Commerce Research and Applications, 14(6), 542–562. https://doi.org/10.1016/j.elerap.2015.08.004.
24. Nilashi, M., Esfahani, M. D., Roudbaraki, M. Z., Ramayah, T., ve Ibrahim, O. (2016). A multi-criteria collaborative filtering recommender system using clustering and regression techniques. Journal of Soft Computing and Decision Support Systems, 3(5), 24-30.
25. Ning, X., Desrosiers, C., ve Karypis, G. (2015). A Comprehensive Survey of Neighborhood-Based Recommendation Methods. Recommender Systems Handbook, 37–76., doi:10.1007/978-1-4899-7637-6_2.
26. Pérez-Marcos, J., ve Batista, V. L. (2017). Recommender system based on collaborative filtering for spotify’s users. In International Conference on Practical Applications of Agents and Multi-Agent Systems, 214-220. doi: 10.1007/978-3-319-61578-3_22.
27. Plantié, M., Montmain, J., ve Dray, G. (2005). Movies recommenders systems: automation of the information and evaluation phases in a multi-criteria decision-making process. Lecture Notes in Computer Science, 633–644. https://doi.org/10.1007/11546924_62.
28. Polatidis, N., ve Georgiadis, C. K. (2016). A multi-level collaborative filtering method that improves recommendations. Expert Systems with Applications, 48, 100–110. https://doi.org/10.1016/j.eswa.2015.11.023.
29. Polatidis, N., ve Georgiadis, C.K. (2017). A dynamic multi-level collaborative filtering method for improved recommendations. Computer Standards & Interfaces, 51, 14–21. https://doi.org/10.1016/j.csi.2016.10.014.
30. Resnick, P., Iacovou, N., Suchak, M., Bergstrom, P., ve Riedl, J. (1994). GroupLens: an open architecture for collaborative filtering of netnews. In Proceedings of the 1994 ACM conference on Computer supported cooperative work, 175-186. https://doi.org/10.1145/192844.192905.
31. Sarwar, B., Karypis, G., Konstan, J., ve Reidl, J. (2001). Item-based collaborative filtering recommendation algorithms. Proceedings of the Tenth International Conference on World Wide Web- WWW ’01. https://doi.org/10.1145/371920.372071.
32. Schick, A. G., Gordon, L. A., ve Haka, S. (1990). Information overload: A temporal approach. Accounting, Organizations and Society, 15(3), 199–220. https://doi.org/10.1016/0361-3682(90)90005-f.
33. Shambour, Q. (2016). A user-based multi-criteria recommendation approach for personalized recommendations. International Journal of Computer Science and Information Security, 14(12), 657.
34. Shambour, Q., ve Lu, J. (2011). A hybrid multi-criteria semantic-enhanced collaborative filtering approach for personalized recommendations. IEEE/WIC/ACM International Conferences on Web Intelligence and Intelligent Agent Technology, 71-78. https://doi.org/10.1109/WI-IAT.2011.109.
35. Shambour, Q., Hourani, M., ve Fraihat, S. (2016). An item-based multi-criteria collaborative filtering algorithm for personalized recommender systems. International Journal of Advanced Computer Science and Applications, 7(8), 274-279. https://doi.org/10.14569/IJACSA.2016.070837.
36. Shi, Y., Larson, M., ve Hanjalic, A. (2014). Collaborative filtering beyond the user-item matrix: A survey of the state of the art and future challenges. ACM Computing Surveys, 47(1), 1–45. https://doi.org/10.1145/2556270.