The Study Of Black Hole Algorithm Based On Membrane Computing

Nature can always bring us endless inspiration and rich imagination. When encounter obstacles in solving problems, people are more likely to ask nature for help which can bring them unexpected results. In the past few decades, a variety of heuristic algorithms are simulated for dealing with complex and optimization problems inspired by the nature. Based on the behavior of animals or the physical phenomena in nature, the heuristic algorithms are developed as a calculation model to provide a method for solving problems, such as genetic algorithms(GAs), simulated annealing algorithm(SA) and particle swarm optimization(PSO). The black hole algorithm which will be mentioned in this thesis is also inspired from the physical black hole phenomenon in the universe and developed as a calculation model which used for data clustering proposed by Abdolreza Hatamlou in 2013. Because of the simple structure and the better clustering effect, this model provides a new direction for people on the clustering problem analysis.As a new branch of natural computation, the aim of membrane computing is to abstract a kind of calculation model from cells or organ. Because of the distributed and parallel characteristics, membrane computing has got a lot of attention since it was put forward. Some experiments show that the membrane computing can solve the SAT problem in linear time effectively and may be beyond the limitations of “Turing machine” on the computing power.This thesis proposes a new calculation model called BH-MC combined the above two algorithms. Based on the calculation model of membrane computing which has the distributed parallel characteristic, BH-MC put the black hole algorithm into the framework of membrane computing to promote the convergence speed of the BH algorithm. In order to verify the performance of BH-MC, we make a comparison on the clustering effect between it and other classic algorithms by using three datasets. In order to highlight the advantage on the convergence speed than BH algorithm, a comparison between BH algorithm and BH-MC algorithm with 2, 4, and 8 elementary membrane respectively. The final experimental results show that BH-MC algorithm has a better advantage on the clustering effect than some other algorithms and a better effect on the convergence speed than BH algorithm.