<p>Knowledge graphs, as strong representations of complicated relationships among entities, have found significant acceptance in numerous domains. To successfully leverage these interconnected structures, embedding techniques have emerged as a potential technique for learning vector representations in low-dimensional entities and relations representations in knowledge graphs. These learned embeddings enable efficient computations and help numerous related tasks to knowledge graphs, including link prediction and entity classification. This paper provides a novel approach for deep clustering entities in knowledge graphs by integrating the RDF2Vec technique for generating embeddings, and convolutional autoencoders (CAEs) to reduce dimensionality and agglomerative clustering. The suggested method shows high performance on the FB15K, YAGO43K, and DB100K datasets, greatly outperforming other techniques. By exploiting the effective feature extraction capabilities of the convolutional autoencoder and the hierarchical clustering strengths of agglomerative methods, our technique delivers the greatest performance in clustering Accuracy, NMI, Purity and ARI metrics. The findings indicate that RDF2Vec embeddings are more successful for entity representation in knowledge graphs, especially when coupled with deep learning-based knowledge graph applications in the semantic web and data mining tasks.</p>