Modeling Online Customer Purchase Intention Behavior Applying Different Feature Engineering and Classification Techniques

Abstract

Online shopping, known as e-shopping, has added a new dimension to the business sector. The concept of consumption among customers has changed significantly as a result of the quick growth of computer technology and e-commerce. The purpose of this work is to investigate the outcomes of different machine learning techniques and predict online shoppers’ purchasing intentions from e-commerce sites. In this study, we collected a dataset of online shoppers' purchasing intentions from the University of California Irvine (UCI) data repository. Different feature transformation techniques were employed in the primary dataset and generated transformed datasets. Besides, transformed datasets were balanced and detected outliers from them. Then, we applied different feature selection methods into primary and transformed balanced datasets and again generated several feature subsets. Further, different classifiers were implemented into all of these generated datasets and obtained different outcomes for them. In this work, Random Forest performed the best accuracy of 92.39% for Z-Score and Gain Ratio Attribute Evaluation transformed dataset. In addition, this classifier also provided the highest AUROC of 0.974 for Square Root and Gain Ratio Attribute Evaluation dataset. Therefore, Random Forest is a more stable classifier to predict customer purchase intention than other models.