Quantifying the Impact of Psychological and User Experience Factors on Online Shopping Satisfaction Via Predictive Models

Abstract

This study examines the joint influence of psychological and user experience (UX) factors on online shopping satisfaction using predictive modelling. The importance of this research is to address a gap in previous e-commerce research, in which psychological drivers and UX attributes are usually analysed independently. Understanding the combined effect of them is crucial to designing customer-centric digital commerce environments. The research has gathered data from online shoppers in Coimbatore, India, using a structured questionnaire. A total of 238 responses were collected, and after data cleaning, 212 valid observations were kept for modelling. Psychological constructs, including trust, fear of missing out (FOMO), price sensitivity, social influence, and emotional motivation, were measured using multi-item Likert-scale questions adapted from the consumer behaviour literature. These responses were transformed into composite numerical features through aggregation and normalisation. The machine learning models used in the research were Random Forest, Gradient Boosting, LightGBM and ElasticNet. Class imbalance was addressed using K-Means SMOTE resampling. Model evaluation revealed that the Random Forest model performed best, with a Mean Absolute Error (MAE) of 0.379 and an ???????? of 0.179. Explainable AI analysis using SHAP values showed that approximately 1.5 times more of the prediction results were attributed to psychological factors than to UX features. These results show that internal psychological drivers are much more important than those related to the interface experienced to determine online shopping satisfaction. The research is a contribution to the field in that it combines behavioural theory, predictive modelling and explainable AI as an analytical framework. Future research can extend this framework to other geographic contexts and incorporate longitudinal behavioural data.