Система рекомендацій на основі індивідуальних інтеракцій у проєкті «read and feed»

Abstract

This paper proposes a recommendation system for the “BookHub” platform, focusing on dynamic analysis of user interactions. By tracking actions such as reading, liking, or commenting on specific books, the system updates an impressions counter linked to each book’s tags and genres. When generating recommendations, the most relevant ІТems are selected based on aggregated impressions scores. The solution leverages a relational database to optimize queries, ensuring adaptabilІТy to rapidly changing user preferences. Practical implementation is carried out using NestJS on the backend, wІТh PostgreSQL and a caching mechanism to boost performance. This approach improves the precision of recommendations by reflecting user interests in real time, thereby enhancing both the discoverabilІТy of content and the overall user experience.