Mathematical Model for Constructing a Personalized Recommendation System Based on Object Features
Keywords:
RECOMMENDATION SYSTEM, CONTENT FILTERING, USER PROFILE, PRECISION@N METRIC, AVERAGE ACCURACY, SYNTHETIC EXPERIMENTAbstract
The model of a content-based filtering recommendation system considered in the article demonstrates its
efficiency and suitability for use in personalization tasks. The proposed approach relies on the use of object attributes and user profiles, which makes it possible to consider individual preferences when generating recommendations.
A synthetic experiment confirmed the model’s effectiveness: the Precision@5 and MAP metrics indicated that more than half of the recommended items were relevant.
The proposed model of a content-based recommendation system, despite its limitations, has a number of unique features that make it especially valuable for personalization tasks. These features determine its advantages and disadvantages, as well as the scenarios for its most effective use. It does not depend on the behavior of other users. Each user's profile is built exclusively on his personal interaction history and explicit preferences. This allows the system to provide very accurate and relevant recommendations for users with unique, niche, or specific tastes that may not coincide with general trends or preferences of the majority.
However, the model is sensitive to the choice of features and the quality of the user profile representation, and is prone to the «narrow specialization» effect. This underlines the importance of further research in the direction of
hybrid approaches that combine content-based and collaborative filtering, as well as testing the model on real-world data.
The development of the proposed methodology can contribute to the creation of more accurate and diverse recommendation systems with high user value and potential commercial impact.
References
Федоренко, В. И. Использование методов векторизации текстов на естественном языке для повышения качества контентных рекомендаций фильмов / В. И. Федоренко, В. С. Киреев // Современные наукоемкие технологии. – 2018. – № 3. – С. 102–106.
Сейдаметова, З. С. Системы рекомендаций в электронной коммерции / З. С. Сейдаметова // Ученые записки Крымского инженерно-педагогического университета. – 2018. – № 3(61). – С. 121–127.
Горелов, М. А. О подходах к построению моделей рекомендательных систем / М. А. Горелов, Ю. В. Бруттан // Математическое моделирование систем и процессов. – Псков : Псковский государственный университет, 2024. – С. 82–87.
Folorunso, Ojо. An Overview of Web Content Filtering Techniques / Ojо Folorunso. – Morgan State Univercity Maryland USA. – 2024. – 9 р.
Karthikeyan, V. K. T. Web Content Filtering Techniques : А Survey / V. K. T. Karthikeyan. – Текст : электронный // A Survey International Journal of Computer Science & Engineering Technology (IJCSET). – 2014. – Vol. 5. – № 03. – Р. 203–208. – URL: https://ijcset.com/docs/IJCSET14-05-03-038.pdf (дата обращения: 12.06.2025).
Черников, С. Ю. Использование системного анализа при управлении организациями / С. Ю. Черников, Р. В. Корольков // Моделирование, оптимизация и информационные технологии. – 2014. – № 2(5). – C. 1–6.
Преображенский, Ю. П. О методах создания рекомендательных систем / Ю. П. Преображенский, В. М. Коновалов // Вестник воронежского института высоких технологий. – 2019. – № 4(31) . – С. 75–79.
Hybrid quality-based recommender systems: A systematic literature review / В. Sabiri, А. Khtira, B. El Asri, М. Rhanoui. – Текст : электронный // Journal of Imaging. – 2025. – Vol. 11. – № 1. – DOI: 10.3390/jimaging11010012. – URL: https://www.researchgate.net/publication/387804221_Hybrid_Quality-Based_Recommender_Systems_A_Systematic_ Literature_Review (дата обращения: 12.06.2025).
Скворцов, В. А. Примеры метрических пространств / В. А. Скворцов. – Москва : МЦНМО. – 2002. – 24 с. – ISBN: 5-94057-002-Х.
