Prediction of the Purchase Intention of Users on ECommerce Platforms using Gradient Boosting

In this paper, we propose a system that is able to forecast the purchase intention of users visiting e-commerce platforms from data collected as they browse on these websites. We use the Online Shoppers Purchasing Intention Dataset available at the University of California Irvine Machine Learning Repository. Thanks to some feature engineering methods, we deeply study the correlation between the various information. We also derive new information / features from the dataset by inference. The most relevant data is fed to gradient boosting, artificial neural networks and other algorithms in order to forecast whether or not a user intends to make a purchase. We evaluate the performances with the precision metric and the F1- Score. The experiments show that our gradient boosting model performs better than the state-of-the-art models thanks to the new features used. This also confirms that, in addition to being interpretable, some classic machine learning models such as gradient boosting can be very competitive compared to neural networks. This system thus conceived can allow e-commerce platforms to identify users intending to make a purchase. This gives them the possibility of offering personalized solutions to their potential customers in order to better attract them and guarantee their purchase, which will imply increased sales and better customer satisfaction