Machine Learning-Driven Prediction of Consumers' Pre-purchase Safety Behaviors in Online Shopping Malls

As online shopping malls play an increasingly crucial role in consumers' daily lives, large amounts of consumer and transactional data have become available.While current machine learning applications in e-commerce focus primarily on enhancing customer experience, increasing sales, and providing personalized recommendations, the analysis of consumer risk behaviors remains underexplored.This study addresses that gap by predicting consumers' pre-purchase safety behaviors to enable the development of personalized safety education programs, ultimately helping prevent unsafe or non-compliant product purchases.We utilize an online survey dataset, which includes consumer demographics, newly defined safety knowledge levels, and reported safety practices-such as checking reviews, monitoring public alerts, and verifying sellers.Five machine learning models were compared: Linear Regression, Random Forest, Neural Network, XGBoost, and SVM.Results from the model comparison indicate that SVM outperforms the other methods, achieving the lowest mean absolute error in numerical predictions and the highest accuracy and AUC in binary classifications of safety behaviors.These findings highlight the influential role of consumer safety knowledge and demographics in shaping pre-purchase risk decisions.Based on the SVM model's predictions, we propose personalized consumer safety education initiatives, such as pre-purchase pop-up or e-mails, that online mall operators can implement to promote safer purchasing decisions.The study demonstrates the feasibility and effectiveness of machine learning in identifying high-risk consumers, offering valuable insights for enhancing product safety awareness and fostering safer e-commerce environments.