ERGONOMIC ANALYSIS OF RECOMMENDER SYSTEMS BASED ON THE COLLABORATIVE FILTERING METHOD

Goryachkin Boris Sergeevich, Semenov Ilya Alexandrovich

Problem statement. Modern recommender systems play a crucial role in personalizing user experiences. However, one of the main challenges is selecting and evaluating methods that ensure high recommendation accuracy, fast query processing, and high performance under scalability conditions. Goal. To conduct a theoretical and practical analysis of the performance of a recommender system based on the Collaborative Filtering method to evaluate its effectiveness across key metrics: accuracy, performance, and query generation time. Results. An analysis of data from the MovieLens dataset was conducted, considering a 25% random sample of ratings and movies. Theoretical values for accuracy, performance, and query generation time were calculated. Practical implementation allowed for the evaluation of deviations from theoretical predictions and the identification of factors affecting system performance. The results demonstrate that Collaborative Filtering provides high accuracy, reasonable query generation time, and acceptable performance. Practical significance. The proposed approaches and calculations can be applied to further optimize recommender systems, improving their scalability and accuracy. This is particularly important for platforms dealing with large datasets and focused on enhancing the user experience.

Keywords:Recommender systems, Collaborative Filtering, accuracy, performance, query generation time