Visual Communication and Fashion Popularity Contagion in Social Networks

Fast fashion has emerged as a prevalent retail strategy shaping fashion popularity. However, due to the lack of historical records and the dynamics of fashion trends, existing demand prediction methods do not apply to new-season fast fashion sales forecasting. We draw on the Social Contagion Theory to conceptualize a sales prediction framework for fast fashion new releases. We posit that fashion popularity contagion comes from Source Contagion and Media Contagion, which refer to the inherent infectiousness of fashion posts and the popularity diffusion in social networks, respectively. We consider fashion posts as the contagion source that visually attracts social media users with images of fashion products. Graph Convolutional Network is developed to model the dynamic fashion contagion process in the topology structure of social networks. This theory-based deep learning method can incorporate the latest social media activities to offset the deficiency of historical fashion data in new seasons.