Predicting The Customer Purchase Intention Based on Perception of Risk Using Machine Learning Techniques

Abstract

Online shopping has become an indispensable aspect of modern life. Nevertheless, this convenience is not without its risks. Potential pitfalls include fraudulent activity, subpar product quality, and the security of personal and financial information. The purpose of this research paper is to predict how individuals react in terms of online purchase intention, guided by their unique risk perceptions. It also aims to explore important risks and sociodemographic factors influencing the levels of perceived risk. This investigation was conducted through a survey questionnaire administered to a sample of 308 participants from India. Various machine learning models were applied to predict the customer purchase intention based on the various risks. The results revealed that the CatBoost Classifier outperformed the other methods. Following closely, Random Forest and Gradient Boost Classifier also demonstrated strong performance. By utilizing Random Forest's feature importance, factors such as financial risk, delivery risk, perceived health risk, time loss risk, and cultural risk significantly impact purchase intention. Additionally, among demographic factors, occupation has the most significant influence. This prediction model will offer practical insights for e-commerce platforms, marketers, and policymakers, enabling them to tailor strategies.