Аннотация:The purpose of this paper is to describe the method for determining the similarity of the information items through the analysis of user preference data. The method is an implementation approach known as Item-Item CF (collaborative filtering based on the similarity of the information items), which in turn is one of the most popular approaches to the construction of modern recommender systems. Initial data for collaborative filtering are the data about users’ activity when they are browsing web resources. Data can be collected as explicit (evaluations, surveys, ratings) and implicit (logging of users’ actions). The proposed method solves the problem of cold start using implicit data about the routes of other users. The method was tested on real data from existing online store Thaisoap, which confirmed the possibility of its applicability in the framework of the task. A unique identifier of the project supported by the Ministry of education and science of the RF is RFMEFI60414X0139.