Predicting Implicit User Preferences with Multimodal Feature Fusion for Similar User Recommendation in Social Media

In social networks, users can easily share information and express their opinions. Given the huge amount of data posted by many users, it is difficult to search for relevant information. addition individual posts, would be useful if we recommend groups people with similar interests. Past studies on user preference learning focused single-modal features such as review contents or demographic users. However, usually not easy obtain in most media without explicit feedback. this paper, propose a multimodal feature fusion approach implicit prediction which combines text image from posts recommending media. First, use convolutional neural network (CNN) TextCNN models extract features, respectively. Then, these are combined using early late methods representation preferences. Lastly, list preferences recommended. The experimental results real-world Instagram show that best performance achieved when apply classification images texts, average top-k accuracy 0.491. This validates effectiveness utilizing deep fusing represent Further investigation needed verify different types