As people’s aesthetic awareness improves and the younger generation’s consumption concept rises, personalized clothing consumption has become a new trend. However, current personalized clothing recommendation algorithms often do not consider incomplete clothing attribute descriptions in real-world scenarios and do not effectively combine user preference features with clothing attribute information. To address this, This paper proposes a personalized clothing recommendation method based on knowledge graph. The method uses LightGCN to complete the information of the clothing to be recommended based on the user-clothing interaction graph, applies the self-attention mechanism to extract user personalized preference attributes, and adopts the KGCN model in the clothing knowledge graph to aggregate the neighborhood entity information of the to-be-recommended clothing and calculate the recommendation probability comprehensively. Experimental results show that the method can achieve an accuracy of 92.4% and provide more interpretable recommendations.