An Improved Spectral Clustering Based on Tissue-like P System

Generally, spectral clustering (SC) includes two steps. First, the similarity matrix is obtained from the original data, then perform k-means clustering based on the similarity matrix. For k-means algorithm, the choice of initial clustering center limits its clustering performance. To solve this problem, this paper proposes an improved spectral clustering algorithm based on tissue-like P system, called ISCTP. It replaces k-means algorithm in spectral clustering with k-means++ to improve the arbitrariness of initial point selection. k-means algorithm needs to artificially determine the initial clustering center, different clustering centers may lead to completely different results. While k-means++ can effectively refine this disadvantage, the basic idea of k-means++ is that the distance of different clustering centers should be as far as possible. In addition, we combine k-means++ with the tissue-like P system that has unique extremely parallel nature and can greatly improves the efficiency of the algorithm. The experimental results of UCI and artificial datasets prove the effectiveness of our proposed method.