A Comparison of Machine Learning Classifiers for Evaluation of Remarketing Audiences in E-Commerce


Ekelik H., Emir S.

ESKISEHIR OSMANGAZI UNIVERSITESI IIBF DERGISI-ESKISEHIR OSMANGAZI UNIVERSITY JOURNAL OF ECONOMICS AND ADMINISTRATIVE SCIENCES, vol.16, no.2, pp.341-359, 2021 (ESCI) identifier

Abstract

In this study, user data of an e-commerce site operating in Turkey is examined. Users are those who have visited the site before, that is, they are in the remarketing audience pool. The main goal is to make accurate predictions for remarketing and thus offer customized ad packages for new visitors. Visitors are labeled as "Shoppers" and "Non-shoppers" based on their previous visits. The data set is divided into two portions that do not intersect with each other as training and test sets. Three classification models based on artificial neural networks, classification and regression trees (CART), and random forest are built to make predictions and then classification performances of these models are compared.