Uso de técnicas de modelagem preditiva para prever o comportamento de compras em E-Commerces

Abstract

With the rapid expansion of electronic commerce, companies have faced increasingly significant challenges related to customer retention, particularly due to the ease with which consumers can switch between providers. Customer turnover, or churn, has emerged as a critical factor affecting financial performance, as it entails potential revenue loss and higher costs associated with acquiring new customers. In this context, the application of machine learning techniques proved to be a promising approach for predicting user behavior and generating strategic insights to support churn-mitigation actions. With this purpose, the present study evaluated predictive models designed to estimate the probability that a product viewed in an e-commerce environment would be effectively purchased by the customer. The experiments were conducted using a public dataset available on the Kaggle platform. The study encompassed preprocessing, feature engineering, model training, and validation stages. The results obtained were compared with those reported in the literature, aiming to contribute to the improvement of predictive solutions applied to the electronic commerce sector