Degree of Personalization in Recommender Systems: a Comparative Analysis of Approaches and Selection Criteria

This article presents a comparative analysis of approaches to building recommender systems based on the degree of personalization of the content they provide. The relevance of this research topic stems from a fundamental contradiction that inevitably arises during the design phase of such systems: the need to simultaneously ensure high accuracy of individual predictions while maintaining an acceptable level of computational complexity, interpretability, and implementation cost.

The goal of this study was to develop a system of criteria and practical tools to support an informed choice of an architectural strategy for a recommender system based on its classifi by degree of personalization. To achieve this goal, we identifi and substantively characterized four classes of systems, developed a set of comparable evaluation parameters, and developed a design decision-making procedure.

The scientific significance of this study lies in its systematization of concepts regarding gradations of personalization and the substantiation of a set of seven interrelated criteria reflecting the key tradeoff between accuracy and complexity in this subject area. The practical value is determined by the development of a flowchart and checklist-tools applicable in the initial design stages and allowing for iterative refinement of the architecture as the service evolves.

The main results of the work indicate that none of the considered classes has an absolute advantage. The choice is determined by the specific project conditions, and the optimal development trajectory is a consistent progression from simple solutions to more complex ones as data accumulates. The novelty of the proposed approach lies in its shift from the problem of choosing a recommendation architecture to one of structured comparison based on explicit criteria.

1. Jannach D., Zanker M., Felfernig A., Friedrich G. Recommender Systems. Cambridge: Cambridge University Press, 2010. URL: https://www.researchgate.net/publication/235910467 .

2. Ricci F., Rokach L., Shapira B., Kantor P.B. Recommender Systems Handbook. New York: Springer, 2015. URL: https://www.researchgate.net/ publication/227268858_Recommender_Systems_Handbook.

3. Меньшикова Н.В., Портнов И.В., Николаев И.Е. Обзор рекомендательных систем и возможностей учета контекста при формировании индивидуальных рекомендаций // Academy. 2016. № 6 (9). С. 20–22.

4. Briand L., Salha-Galvan G., Bendada W., Morlon M., Tran V. A Semi-Personalized System for User Cold Start Recommendation on Music Streaming Apps // arXiv.org. 2021. 7 June. URL: https://arxiv.org/abs/2106.03819 .

5. Burke R. Hybrid Recommender Systems: Survey and Experiments // User Modeling and User-Adapted Interaction. 2002. Vol. 12. № 4. P. 331–370. https://doi.org/10.1023/A:1021240730564.

6. Фальк К. Рекомендательные системы на практике / Пер. Д.М. Павлов. М.: ДМК Пресс, 2020. 448 с.

7. Чжан Л. Анализ компромисса между точностью и сложностью идентифицированных моделей динамических систем // Системы анализа и обработки данных. 2024. № 2 (94). С. 85–93.

8. Чжан Чжибо, Афанасьев Г.И. Основные технологии и перспективы эволюции персонализированных рекомендательных систем // E-Scio. 2022. № 4 (67). С. 309–320.