Enhancing Personalized Recommendations through Optimized Parameters in Collaborative Filtering

The influence of the main parameters of the collaborative filtering algorithm on the quality of personalized recommendations is investigated. The number of users in the system and the number of nearest neighbors involved in the calculation are used as independent variables in the experiment. For each combination of these parameters, the mean absolute error is measured, which reflects the accuracy of the recommendations, as well as the algorithm's operation time. Based on the obtained data, heat maps are created that allow visualizing the estimates obtained. Optimal areas are identified in which high accuracy of recommendations can be achieved with low time costs. The obtained data can be used to more precisely tune the algorithms of recommendation systems.