摘要：The purpose of this paper is to propose a new way to solving clustering problems, which combines membrane computing with a k-nearest based algorithm inspired by chameleon algorithm. The new algorithm is defined as PKNBA, which can obtain the k-nearest graphs, complete the partition of subgraph through communication rules, evolution rules, dissolution rules and division rules in P system with active membranes. The whole process of PKNBA algorithm is shown by a 10 points test data set, which indicates the feasibility and less time consuming of the algorithm. All the processes are conducted in membranes. Cluster results via the famous iris and wine data set verify that the proposed PKNBA algorithm can cluster data set more accurate than k-means algorithm. The influences of parameters to the algorithm are illustrated also. The PKNBA provides an alternative for traditional computing.
