Search academic texts

Information × Registration Number 2123U006689, Article popup.category Препринт Title popup.author Luchka Yurii popup.publication 01-01-2023 popup.source_user Український католицький університет popup.source https://hdl.handle.net/20.500.14570/4818 popup.publisher Description This thesis examines the issue of user selection bias in recommender sys- tems. This problem is especially acute when recommendations are built on limited data (for example, only transactional data is present, but no data on item categories, customer demographics, etc.). In this work, the solution to the problem of minimiz- ing the recommendation selection bias is proposed by building an algorithm, which will be based on finding the best next offer relative to the customer’s purchase se- quence and similar behaviour of other customers. Patterns in customer behaviour are proposed to be determined using K-Means clustering, based on the similarity between customers (namely, their sequences) found using Dynamic Time Warping. The next best offer is searched for the target variable, the basis of which is the transi- tion probability calculated using the transition matrix and the individual score of the item (in this case, it is calculated as a complex variable based on recency, frequency, and monetary). popup.nrat_date 2025-05-09 Close