A Hybrid Approach Optimized by Tissue-Like P System for Clustering

摘要

K-medoids algorithm is a classical algorithm used for clustering, it is developed from K-means algorithm and it is more robust compared to K-means algorithm for noises and outliers. But it takes more time to achieve a better result. In this paper, we proposed a hybrid algorithm to overcome the drawbacks. We combined K-means algorithm and an improved K-medoids algorithm together. Firstly, to have an elementary clustering results, we run K-means algorithm for the data set. Then, the improved K-medoids algorithm are used to optimize the results to make it more robust. Furthermore, we designed a Tissuelike P system for the proposed approach, the Tissue-like P system operates in a parallel way thus can improve time efficiency greatly. We tested the efficiency and effectiveness of our approach on some data sets of the well-known UCI benchmark and compared the approach with the K-means and the K-medoids algorithm.