Current self-representation learning approaches, particularly Low-Rank Representation (LRR) and Sparse Subspace Clustering (SSC), frequently fail to account for the relationships between pseudo-labels while insufficiently exploiting the inherent connections within data. Pseudo-labels offer distinct advantages over raw data, including reduced dimensionality, enhanced generalization capability, and improved interpretability. To overcome these limitations, this work introduces a Low-Rank Representation based on Pseudo-Label Learning (PLLRR) clustering algorithm that effectively combines pseudo-label learning with LRR. The proposed method first decomposes the original data into symmetric latent feature matrices through affinity matrix analysis, utilizing these as pseudo-labels to guide feature selection in the low-rank representation space. Subsequently, it constructs an enhanced affinity matrix incorporating pseudo-labels to preserve inter-sample correlations, while employing Laplace embedding to maintain both local data structures and latent feature space geometry during low-rank projection. Experimental validation on benchmark datasets confirms the superior performance of PLLRR compared to existing approaches.