Analysis and Research of Characteristics of Algorithms in Recommender Systems

Шевченко, Світлана Миколаївна and Жданова, Юлія Дмитрівна and Данилюк, О. С. (2024) Analysis and Research of Characteristics of Algorithms in Recommender Systems Вісник Херсонського національного технічного університету, 4 (91). pp. 367-376. ISSN 2078-4481

Text Shevchenko_S_Zhdanovа_Y_Danyliuk_O_KhNTU_4_91_2024.pdf - Published Version Download (630kB)

Abstract

The information environment is becoming more saturated and dynamic. Information grows according to a double exponential law. The process of searching, analyzing and filtering information is becoming more complicated every day, and the growing volume of information complicates the process of making informed decisions. Solving this problem is possible through the development and implementation of recommender systems. This article is devoted to the latest models of recommendation systems, namely: based on content filtering, collaborative filtering, hybrid filtering. Based on the analysis of scientific and technical literature, the underlying mechanism of the recommendation process was identified, and an overview of filtration methods and their characteristics was made. Presented are schemes for implementing recommendations and possible metrics for grouping relevant clients. The disadvantages of each recommendation process are characterized, namely: cold start, sparseness, "gray sheep" for collaborative filtering and is associated only with data on the subject of content filtering limitations. This justifies the superposition (combination) of these two filtering to create a hybrid filtering. The article considers the issue of the quality of recommender systems. In addition to traditional indicators of accuracy and completeness, modern approaches take into account diversity, novelty, unexpectedness of recommendations, robustness, interpretability, and fairness. Emphasis is placed on external destabilizing factors in the work of recommendation systems, in particular, the existence of a threat of violation of the confidentiality of the user's personal data and the threat of receiving incorrect recommendations as a result of a targeted attack on the recommendation system. As an example, the process of developing a recommender system for selecting sports equipment products is described: installing features before the operation of this system; the database of the characteristics of the sports equipment has been christened; the diagram of variants of the wiki was modeled; the security program was dismantled and protested by the Telegram bot. The results of the study can be implemented in the educational process of students 12 Information technology.

Item Type:	Article
Uncontrolled Keywords:	recommender system; content filtering; collaborative filtering; hybrid filtering; clustering methods
Subjects:	Статті у періодичних виданнях > Фахові (входять до переліку фахових, затверджений МОН)
Divisions:	Факультет інформаційних технологій та математики > Кафедра інформаційної та кібернетичної безпеки ім. професора Володимира Бурячка
Depositing User:	Ю. Д. Жданова
Date Deposited:	10 Mar 2025 09:40
Last Modified:	10 Mar 2025 09:40
URI:	https://elibrary.kubg.edu.ua/id/eprint/51299

Actions (login required)

View Item