As e-commerce applications evolve from breadth to depth, personalization has become an important direction for e-commerce service innovation. In order to accurately predict consumers’personalized demand, this paper proposes a personalized demand prediction method with a limited preference constraint by integrating information of product description texts and display images. Inspired by the hypothesis of limited attention, this paper models the limited personalized preference of consumers, and combines the image and text features to construct a Sparse Text-Image Linked Topic Model with a limited preference constraint. The model predicts the individual demands of consumers through group interest modeling, individual preference modeling, and purchase decision modeling. Experiments on the Amazon public data set show that the proposed model can effectively predict group interests and personalized preferences of consumers. The integration of text and image information improves the interpretability of personalized demand prediction.