[Purposes]Given that traditional traffic information recommendation methods lack the ability to depict individual user differences, making it difficult to achieve precise and personalized information recommendation, this study proposes a traffic information targeted recommendation method driven by user profile and graph contrastive learning. [Methods]Firstly, a multi-source traffic data is integrated to construct a multi-dimensional label system encompassing user behavior, travel attributes, and information preferences, enabling user profile modeling. Secondly, a user-label bipartite graph structure is constructed, node weights are designed, and spatiotemporal context features are aggregated through a multi-layer Graph Convolutional Network (GCN). Dynamic traffic information and user preference labels are mapped to a unified metric space for representation learning. Additionally, an adaptive perturbation and cross-layer contrastive mechanism is introduced to enhance feature robustness, and prototype contrastive enhancement is incorporated to strengthen group semantic expression. Finally, a comprehensive loss function is constructed to optimize the model and generate personalized recommendation results. [Findings]Compared to the optimal results of various benchmark models, the model achieves an improvement of 4.56%, 4.56%, 2.42%, and 6.44% in terms of Hitratio@10, Precision@10, Recall@10, and Normalized Discounted Cumulative Gain (NDCG)@10, respectively. Simultaneously, the model demonstrates optimal comprehensive performance when the number of information recommendations is 5, and the contributions of each module to the recommendation task are validated in ablation experiments. [Conclusions]This study provides a new research direction for personalized traffic information recommendation.