Abstract: With the rapid growth in the number of intelligent terminal devices and wireless multimedia applications, mobile communication traffic has exploded. The latest report from Cisco Visual Networking Index (CVNI) indicates that by 2022, global mobile data traffic will have grown to three times that in 2017, which will exert tremendous pressure on the backhaul link. One key approach to solve this problem is to cache popular content at the edges (base stations and mobile devices) and then bring the requested content from the edges close to the user, instead of obtaining the requested content from the content server through backhaul networks. Thus, by obtaining the required content of mobile users locally, edge caching can effectively improve network performance and reduce the pressure on the backhaul link. However, owing to the limited storage capacity of the edge nodes and the diversification of user requirements, the edge nodes can neither cache all the content in the content server nor randomly cache the content. To solve these problems, an edge-caching mechanism based on user-awareness was proposed. First, using an implicit semantic model, we predicted popular content in a macro cell in terms of the users’ interests. Small base stations within identical macro cells cache data cooperatively, which update local popular content based on the dynamic content preference of users. To further reduce the delay in content delivery, we helped users to ascertain their top communities of interest based on their content preferences. At the same time, the most appropriate user equipment (UE) is selected considering the caching willingness and caching ability to cache data for other UEs in identical communities of interest. Results show that the proposed mechanism outperforms the random cache approach and the most popular content-caching algorithm; it improves the cache hit rate and reduces the transmission delay while enhancing the quality of user experience.