Application of Boosting in Recommender Systems

In today’s digital era, recommender systems have gained a strong foothold, becoming an important tool for effectively managing information flows. Their demand is largely due to the dynamics of current society, namely, information overload and the need to personalize data. With the expansion of the scope of application of recommendation algorithms, many nonstandard cases have appeared, for which the use of classical approaches is not as effective. This paper examines one of these: a small number of objects with a relatively large number of users in conditions of high correlation between some objects. For modeling, it is proposed to use gradient boosting, a machine learning algorithm based on an ensemble of decision trees.