Recommender systems synthesis and analysis for goods and services sale with cold start problem

Abstract

The article develops a general approach to solving the problem of cold start with the possibility of flexible operation of the algorithm. An analysis of the features of the construction of a recommender system was carried out, as a result of which it was found that today there are a number of problems, the implementation of which requires taking into account many parameters depending on the specifics of the subject area, which served as an incentive for further analysis. It is noted that the process of collecting user data is a rather complex and time-consuming process that has several implementation principles. This is due to the fact that not every user provides sufficient information for further work, which in fact creates further complications due to the insufficient amount of information. One of the ways to solve this problem is to apply intelligent methods to construction, namely, machine learning methods. The main aspects of algorithms and methods of their improvement are briefly described. The hybrid method implementation for a system construction, as well as its performance testing in comparison with the classical k-means algorithm, is carried out. Scalability and ways to improve are also taken into account during implementation.