DE-MC: A Membrane Clustering Algorithm Based on Differential Evolution Mechanism

A clustering algorithm using the framework of membrane computing is proposed in this paper. The P system used is a cell-like P system of two-layer nested structure: a skin membrane contains several elementary membranes. Each object in elementary membranes represents a group of cluster centers. Objects in the system evolve by using the differential evolution mechanism, and then the global optimal object in the skin membrane is updated by the best objects in all elementary membranes. The cell-like P system can automatically find the best cluster centers for a data set. The proposed DE-MC algorithm is evaluated on an artificial data set and a real-life data set and is further compared with classical k-means algorithm, GA-based clustering algorithm and DE-based clustering algorithm, respectively. The comparison results reveal that the proposed DE-MC algorithm is superior to the other three clustering algorithms in terms of clustering quality and robustness. Key-words: Membrane computing; P systems; Clustering algorithm; Differential evolution. DE-MC: A Membrane Clustering Algorithm 77