Managing diversity of products in recommendation models based on an architecture with an attention mechanism (transformers)

Fedor Krasnov

Abstract


Millions of products are available for online buyers to choose from on Internet marketplaces. Despite all the advantages of diversity, a huge number of possible purchase options can stop and make it difficult to choose, as a result of which buyers leave without shopping at all. Such a situation does not benefit either buyers seeking to make a purchase, or sellers who have missed their benefits, or the online trading platform itself, which loses commission and attractiveness for advertisers. One of the main values of an Internet marketplace is to direct customers to the products that best match their interests and motivation. To do this, Internet marketplaces rely on recommendation systems. Using Data Science to empower sellers helps predict which products the buyer will add to the cart and place an order, and also allows you to improve the quality of customer service the next time you make purchases online. Modern recommendation systems consist of various models with different approaches, ranging from simple matrix factorization to deep artificial neural networks with an attention mechanism (transformer). However, there is no single model that could simultaneously optimize all the tasks of users, sellers and Internet marketplaces. In this study, the author focused on the task of a balanced selection of product recommendations, taking into account the interests of the user, the seller and the Internet marketplace. Improving recommendations will make navigating through seemingly endless search options easier and more attractive for buyers. To do this, the author examined the metrics of the variety of the product range such as Serendipity and showed the possibilities and limitations in optimizing diversity for models of recommendation systems based on an architecture with an attention mechanism (transformers).

Full Text:

PDF (Russian)

References


Fei Sun, Jun Liu, Jian Wu [et al] BERT4Rec: Sequential recommendation with bidirectional encoder representations from transformer // In Proceedings of the 28th ACM international conference on information and knowledge management. – 2019. – P. 1441–1450.

Kun Zhou, Hui Wang, Wayne Xin Zhao [et al] S3-rec: Self-supervised learning for sequential recommendation with mutual information maximization // In Proc. CIKM. – 2020. – P. 1893–1902.

Xinyan Fan, Zheng Liu, Jianxun Lian [et al] Lighter and better: low-rank decomposed self-attention networks for next-item recommendation // In Proc. SIGIR. – 2021. – P. 1733–1737.

Chang Liu, Xiaoguang Li, Guohao Cai [et al] Noninvasive Self-attention for Side Information Fusion in Sequential Recommendation // In Proc. AAAI. – 2021. – P. 4249–4256.

Kang, Wang-Cheng, Julian McAuley Self-attentive sequential recommendation // IEEE international conference on data mining (ICDM), 2018.

Wilm, T., Normann, P., Baumeister, S., Kobow, P. V. Scaling Session-Based Transformer Recommendations using Optimized Negative Sampling and Loss Functions // In Proceedings of the 17th ACM Conference on Recommender Systems. – 2023. – P. 1023-1026.

Ruihong Qiu, Zi Huang, Hongzhi Yin, and Zijian Wang Contrastive learning for representation degeneration problem in sequential recommendation // In Proc. WSDM. – 2022. – P. 813–823.

Channarong, Chanapa [et al] HybridBERT4Rec: a hybrid (content-based filtering and collaborative filtering) recommender system based on BERT //IEEE Access 10. – 2022. – P. 56193-56206.

Petrov A., Craig M. A systematic review and replicability study of bert4rec for sequential recommendation // Proceedings of the 16th ACM Conference on Recommender Systems, 2022.

Klenitskiy A., Vasilev A. Turning Dross Into Gold Loss: is BERT4Rec really better than SASRec? // Proceedings of the 17th ACM Conference on Recommender Systems, 2023.

Carlos García Ling, Elizabeth HMGroup, Frida Rim, inversion, Jaime Ferrando, Maggie, neuraloverflow, xlsrln // H&M Personalized Fashion Recommendations. Kaggle. https://kaggle.com/competitions/h-and-m-personalized-fashion-recommendations, 2022.

Wand An., Normann Ph., Baumeister S., Wilm T., Reade W., Demkin M. OTTO – Multi-Objective Recommender System // Kaggle. – URL: https://kaggle.com/competitions/otto-recommender-system (2022).

Mohammed A. A., Umaashankar V. Effectiveness of Hierarchical Softmax in Large Scale Classification Tasks // International Conference on Advances in Computing, Communications and Informatics (ICACCI), Bangalore, India. – 2018. – P. 1090-1094.

Wang R., Shivanna R., Cheng D. [et al]. Dcn v2: Improved deep & cross network and practical lessons for web-scale learning to rank systems // In Proceedings of the web conference. – 2021. – P. 1785-1797.

Carbonell J. G. and Goldstein J. The use of MMR, diversity-based reranking for reordering documents and producing summaries // in SIGIR. – ACM, 1998. – P. 335–336.

Kim J., Jeon H., Lee J., & Kang, U. Diversely regularized matrix factorization for accurate and aggregately diversified recommendation // In Pacific-Asia Conference on Knowledge Discovery and Data Mining. – 2023. – May. – P. 361-373. (Cham: Springer Nature Switzerland)


Refbacks

  • There are currently no refbacks.


Abava  Кибербезопасность MoNeTec 2024

ISSN: 2307-8162