The paper introduces a novel method called Attribute-Missing Graph Clustering (AMGC) for deep graph clustering on graphs with missing node attributes. Traditional methods often assume all node attributes are complete, but in practice, many graphs have missing attributes due to privacy, copyright, or data acquisition issues. AMGC addresses this by integrating clustering and imputation processes in a unified framework, allowing the model to iteratively improve both tasks. The method uses cluster-oriented imputation to enhance data completion and a dual non-contrastive clustering loss to refine clustering distribution through model optimization. This approach improves the quality of graph embeddings and enhances clustering performance. The proposed method is evaluated on five datasets and outperforms existing methods in clustering accuracy. The results show that AMGC effectively handles attribute-missing graphs and achieves superior performance compared to other clustering methods. The method is also efficient and scalable, with linear time complexity relative to the number of nodes and edges. The experiments demonstrate that AMGC is robust to varying levels of attribute-missing ratios and outperforms other methods in clustering accuracy. The method is also effective in handling large-scale data and shows promising results in clustering tasks with limited visible attributes.The paper introduces a novel method called Attribute-Missing Graph Clustering (AMGC) for deep graph clustering on graphs with missing node attributes. Traditional methods often assume all node attributes are complete, but in practice, many graphs have missing attributes due to privacy, copyright, or data acquisition issues. AMGC addresses this by integrating clustering and imputation processes in a unified framework, allowing the model to iteratively improve both tasks. The method uses cluster-oriented imputation to enhance data completion and a dual non-contrastive clustering loss to refine clustering distribution through model optimization. This approach improves the quality of graph embeddings and enhances clustering performance. The proposed method is evaluated on five datasets and outperforms existing methods in clustering accuracy. The results show that AMGC effectively handles attribute-missing graphs and achieves superior performance compared to other clustering methods. The method is also efficient and scalable, with linear time complexity relative to the number of nodes and edges. The experiments demonstrate that AMGC is robust to varying levels of attribute-missing ratios and outperforms other methods in clustering accuracy. The method is also effective in handling large-scale data and shows promising results in clustering tasks with limited visible attributes.