Predicting customer purchase patterns in online retail using a cnn-based deep learning model

Abstract

Accurately predicting customer purchase patterns in online retail enables personalized recommendations, targeted marketing, and improved business decision-making. However, challenges such as high-dimensional transactional data, class imbalance, and the limitations of traditional Machine Learning (ML) models often hinder predictive performance. In this study, a Convolutional Neural Network (CNN) based model was designed to predict customer purchase behavior from online retail transaction data. CNNs are particularly effective at learning complex patterns and feature relationships, making them well-suited for structured data representation. The experiment was conducted on an online retail dataset comprising customer purchase patterns obtained from the University of California, Irvine repository, one of the most widely used benchmark datasets for evaluating ML algorithms. The performance of the CNN model was evaluated using accuracy, precision, recall, F1-score, and the Area Under the Curve of the Receiver Operating Characteristic (AUC-ROC), achieving 93.6% accuracy, 100.0% precision, 91.1% recall, 95.4% F1-score, and an AUC-ROC of 0.98. These results demonstrate that deep learning can effectively model customer purchasing behavior, offering valuable insights for online retail platforms aiming to anticipate customer actions and optimize engagement strategies.