Аналіз методів машинного навчання для прогнозування поведінки користувачів в e-commerce

Abstract

In this study, machine learning methods for predicting user behavior in e-commerce systems are examined under conditions of rapidly increasing volumes of behavioral data. The main approaches to modeling customer behavior are reviewed, and a comparative evaluation of classification algorithms and ensemble methods is conducted. As an applied case, the prediction of a customer’s probability of product return is considered. The results demonstrate the effectiveness of ensemble methods for improving predictive accuracy and supporting decision-making in e-commerce.