Projected cross-view learning for unbalanced incomplete multi-view clustering aims to address the challenges of clustering data with missing samples across multiple views. The method introduces a novel approach called PCL_UIMVC, which effectively handles unbalanced incomplete multi-view data by integrating reconstruction terms, projection matrices, and graph regularization. The reconstruction term leverages existing samples to complete missing data, while the projection matrix harmonizes feature dimensions across views, reducing information imbalance. A graph regularization term preserves the geometric structure of the original data, and an iterative algorithm is developed to solve the model. Extensive experiments on eight standard datasets with varying missing rates validate the superior clustering performance of PCL_UIMVC. The method outperforms existing approaches in terms of clustering accuracy, normalized mutual information, purity, F-score, precision, recall, and AR. The proposed method is effective in handling both balanced and unbalanced incomplete multi-view data, demonstrating robustness and improved performance in clustering tasks.Projected cross-view learning for unbalanced incomplete multi-view clustering aims to address the challenges of clustering data with missing samples across multiple views. The method introduces a novel approach called PCL_UIMVC, which effectively handles unbalanced incomplete multi-view data by integrating reconstruction terms, projection matrices, and graph regularization. The reconstruction term leverages existing samples to complete missing data, while the projection matrix harmonizes feature dimensions across views, reducing information imbalance. A graph regularization term preserves the geometric structure of the original data, and an iterative algorithm is developed to solve the model. Extensive experiments on eight standard datasets with varying missing rates validate the superior clustering performance of PCL_UIMVC. The method outperforms existing approaches in terms of clustering accuracy, normalized mutual information, purity, F-score, precision, recall, and AR. The proposed method is effective in handling both balanced and unbalanced incomplete multi-view data, demonstrating robustness and improved performance in clustering tasks.