<p>In multi-view machine learning problems we have multiple descriptions of the same objects. Multi-view clustering is about finding cluster structures from these multi-view descriptions. The goal is to integrate the complementary information that is found in these different views. In this work, we develop a graph-based approach for multi-view clustering. In particular, we introduce the Joint Multiple Efficient Neighbors and Graph (JMEG) learning method for multi-view clustering. Our objective is to find a good balance between sparsity and connectivity. To do so, we introduce a post-processing technique. Then, JMEG also employs partition space and consensus graph learning to uncover data structures effectively. The whole clustering process is defined in terms of an optimization problem, which is solved using an iterative alternating algorithm. We illustrate the usefulness of our approach with experiments that outperform state-of-the-art methods. We also discuss in this paper the computational cost of JMEG.</p>