With the advantages of content generation and information diffusion speed, social media has become an information sharing and online interaction tool used by more and more people every day. Different from the traditional instant interactive application, the online interaction of users in social media takes messages as the carrier, which makes the content and form of user interaction data extremely rich. Nowadays, social networks have become the main means for people to keep in touch and have fun. Users are the core of social networks, and their behavior is the starting point for understanding the operating mechanism of social networks, which is of great value to the analysis and research of user behavior in social networks. In this article, an analysis model of social media users' chat interaction behavior based on Hadoop and clustering algorithm is proposed to model social media users' interaction behavior and predict users' preferences, so as to provide users with various individual services. The analysis and modeling of high-quality user data can reveal the user preferences and behavior patterns behind users' online behavior, and provide more abundant supporting knowledge for personalized recommendation of social media.