Proceedings of the International scientific and practical conference ―Modern Science: Challenges and Perspectives‖ (February 9-11, 2026) / Publisher website: www.naukainfo.com. - London, United Kingdom, 2026. - 121 p.

50 methods, such as boosting or stacking, to integrate different recommendation algorithms in order to enhance the generalisation capability of the system. The engineering advantages of hybrid recommender models are primarily manifested in their ability to address the cold-start problem by leveraging content features for new users and items, as well as in reducing the negative impact of interaction matrix sparsity [4]. The integration of multiple information sources enhances prediction accuracy, recommendation stability, and the diversity of suggested items, which is critical for improving conversion rates and user retention in e-commerce systems. In the practice of large e-commerce platforms, hybrid recommender systems are employed for homepage personalisation, the construction of dynamic product catalogues, cross-item recommendations, and personalised marketing campaigns. In particular, large-scale e-commerce ecosystems adopt hybrid architectures that integrate user behavioural data, semantic content features, and contextual factors (such as time, location, and device) into multi-layer neural models, thereby enabling adaptive real-time personalisation [5]. To systematise contemporary algorithmic approaches to recommender systems, it is appropriate to conduct a comparative analysis of the main classes of algorithms according to key engineering and algorithmic criteria, including recommendation accuracy, scalability, data requirements, result interpretability, and the complexity of integration into industrial information systems (Table 1). Such an analysis makes it possible to assess the suitability of specific algorithms for deployment in large-scale e-commerce platforms [6].