Personalization in E-Commerce: Optimizing Recommendations for Multimodal Content

Аннотация

This article examines modern approaches to VOLUME Vol.07 Issue 05 2025 the personalization of multimodal content in e- CITATION commerce, driven by the growing complexity of user Condé Mariame requests and the evident need to adapt recommendations to diverse data formats and content COPYRIGHT modality. The relevance of this topic is underscored by © 2025 Original content from this work may be used under the terms of the creative commons attributes 4.0 License. the increasing volume of information associated with the article, including text, images, video, and audio, which necessitates the application of specialized methods for precise customization and improved personalization. The purpose of the study is to develop original proposals for optimizing recommendation algorithms based on multimodal information, enabling the consideration of both context and individual user preferences. The research reveals contradictions in the literature—while many studies focus on specific aspects of personalization, such as textual data or visual elements, integrative approaches to the analyzed content are insufficiently addressed. The author proposes solutions combining deep learning methods and behavioral model analysis to achieve more accurate results in predicting audience interests. The materials presented in this work will be useful for e- commerce professionals, developers of recommendation systems, and researchers focused on evaluating behavioral patterns. Keywords: algorithms, deep learning, multimodal content, personalization, user behavior, recommendations, e-commerce. Introduction: E-commerce, which has become one of the The American Journal of Applied Sciences 57 https://www.theamericanjournals.com/index.php/tajas The American Journal of Applied Sciences leading forms of interaction between sellers and the perception of information. Ya. Zhu and his consumers, is currently characterized by rapid growth in colleagues [10] detail a training model for analyzing data volumes and content representation formats. videos considering affective content, which is relevant for enhancing user experience in e-commerce. As multimodal information representation technologies (texts, images, videos, audio files, etc.) evolve, Certain researchers focus on personalizing interfaces. A. numerous additional challenges arise, associated with Wasilewski and G. Kolaczek [8] emphasize that a adapting to the individual needs of users. universal interface design does not suit all users, proposing a multivariate approach with a focus on The relevance of this article's topic is explained by the adaptation to individual preferences. dynamic development of digital commerce combined with the increasing consumer expectations for Contemporary studies also address understanding the personalized experiences. In the context of intense impact of multimodal content on trust. I. Syed and competition and information overload, businesses are colleagues [6] examine the effects of facial features, compelled to adapt their approaches to data processing voice, and text messaging on trust levels, which is by integrating textual, visual, and audio information to valuable for fostering deeper customer engagement. P. create accurate, relevant recommendations. Thangavel and R. Lourdusamy [7] describe approaches Multimodal content enables the consideration of to emotional analysis using lexicon-based methods, numerous factors influencing preferences, making the enhancing personalization capabilities. process more targeted and effective. A number of authors concentrate on creating The research problem lies in the need to develop recommendation systems to improve user experience. effective personalization methods that formulate S. Silvester and Sh. Kurain [5] provide an overview of the recommendations accounting for not only user Dual-Blend Insight system, which integrates multimodal preferences but also the context of the described content and analytical data to improve recommendation content's consumption. accuracy. Z. Zhang and colleagues [9] explore the methodological foundation for synthesizing visual