Incorporating knowledge graphs as auxiliary information into recommendation systems has become a research hotspot. Knowledge graphs contain a vast amount of inter-item relationship information, which can directly enhance item-side entity modeling and indirectly optimize user-side entity modeling. However, in previous research, this process has not fully utilized the information on the user side, and the information flow between users and items has been based solely on interaction relationships, leading to incomplete user representation modeling. To address these issues, we propose a recommendation model based on knowledge graphs and user representation enhancement. First, we integrate user attribute information and item knowledge graphs, constructing both a user collaborative graph and a project collaborative graph. The representations output by these dual graphs after knowledge propagation are then fused using an attention mechanism. The resulting representations incorporate both item-side and user-side auxiliary information. Second, to address the issue of simple connections between users and items, we enrich the connection forms between users and items using attribute spaces and model users’ interest fluctuation features. We assign additional time weights to users’ historical interaction data, allowing the model to learn item attribute features while reducing the influence of distant interactions that the user is no longer interested in. Experiments show that our model outperforms baseline models in terms of recommendation performance.