20 October 2024 in Scientific Messenger of LNU of Veterinary Medicine and Biotechnologies. Series: Economical Science
Today, recommender systems play a significant role in many areas related to the processing of large volumes of information, such as online stores, stock exchanges, services providing various services. The construction of recommendations is based on the analysis of information about the user's preferences obtained in the past. Often, such information is presented in the form of a matrix of users-objects. Most often, the work of recommender systems is based on generalization and analysis of preferences of other users, which includes consideration of various aspects of their behavior, and available information about the current user. However, there is a class of tasks for which recommendations are based on current user actions. Recommender systems were one of the first mass applications of data analysis in various fields. The reason is the user-friendly results (recommendations) and metrics for measuring the quality of their work. The work is devoted to recommender systems, which is one of the most widely used approaches in machine learning. The main task is to demonstrate different approaches to building a recommender system, such as associative rules and collaborative filtering. The work also proposes an algorithm for working with users who prefer unique objects. Considered popular information products that use the data of their users to improve their recommendation algorithms, which leads to the improvement of the service and the provision of more optimized and relevant recommendations for the presented product or the display of advertisements. Using the Python programming language, an algorithm for generating synthetic information simulating customer transactions in retail trade was implemented to more correctly demonstrate the work of the recommender system. Based on the capabilities of the Python programming language and its powerful libraries, the development direction for Python was chosen with the possibility of creating a recommender system for demand predictions. The presented system allows you to build recommendations based on associative rules and by means of visual comparison, to select the necessary parameters for building rules that correspond to specific applied tasks. By applying the synthetic information generation algorithm, a database was generated, on the basis of which associative rules were built. Based on the obtained results, the functionality of the algorithms is demonstrated and the possibilities for analysis and implementation of the system in applied tasks are reviewed.