Psychological stress detection from online shopping

Abstract

The increasingly faster life pace in modern society makes people always feel stressful and it is of great significance to discover a users suffering stress in time. According to psychological study, shopping is chosen as an effective way for stress relief, especially for females. Compared with non-stress cases, a user may perform different shopping patterns when under stress. An interesting issue then arises: can we detect one’s psychological stress from online shopping data? By investigating stress-related outlier features from both content and behavior of online purchase orders, we learn a users stress status by classification. A real user study of 20 experienced female online customers aged 23–30 verifies the effectiveness of shopping based stress detection, achieving an F1- measure of more than 80% with J48 classifier. None of the features negatively affect the detection result. Feature combinations bring dramatic improvements than single feature. In total, shopping content features are proved to be more significant than behavior features.