The Study Of Clustering Algorithm Based On Membrane Computing

Membrane computing, known as P system, is a kind of distributed parallel computingmodels. Membrane computing abstracts computing models from the structure and thefunctioning of living cells as well as from the cooperation of cells in tissues, organs, and otherhigher order structures such as colonies of cells. Membrane computing has several attractivefeatures such as distribution, maximum parallelism and non-determinism, and has beenwidely applied in various areas.Data clustering is an unsupervised learning process, which divides data points intoseveral groups or clusters such that patterns in same group are as similar as possible andpatterns form different groups are not similar as possible. Different from classification, dataclustering has not the class labels given in advance. The aim of clustering is helped to find thenature structure hidden in data.The paper uses membrane computing models as the computing framework and proposestwo clustering algorithms under membrane computing, which introduces genetic mechanismand simulated annealing mechanism into P systems as evolution rules of objects andcombines its communication mechanism:(1) GA-MC algorithm. Three genetic operators (selection, crossover and mutation) ingenetic mechanism are introduced to achieve the evolution of objects and the communicationmechanism of the tissue-like P systems is applied to share the objects between differentmembranes, which will accelerate the convergence of the algorithm. The GA-MC algorithm istested on several real-life data sets and artificial data sets and compares with k-meansalgorithm and GA-based k-means algorithm.(2) SA-MC algorithm. It is a clustering algorithm combining P systems and simulatedannealing mechanism, where simulated annealing mechanism is used as the evolution rules ofobjects, while the communication mechanism of P systems achieves the sharing of eliteobjects to promote the evolution of objects. The algorithm is also tested on several real-lifedata sets and artificial data sets and has a comparison with k-means algorithm and GA-basedk-means algorithm.