EC Tissue-like P System Based Clustering Problem Research

Being a new research direction, membrane computing inspired by the way living cells processing compound, energy and information, then designs distributed and parallel computing models- P systems. Every cell in P systems is a reaction control unit which calculates independently, therefore P systems have the characteristics of distribution and highly parallelism,thus have high efficiency. At the theoretical level, such as mathematics and computer theory, it has gained fairly good development and provides system architecture for discrete and modular algorithm model, and implements some application by ordinary computer. The application of membrane computing has been involved in biology, bio-medical, computer graphics, economics,the approximate optimization, cryptography, and other fields. Thereby it shows great development potential.Data mining analysis is one of the critical phases of knowledge discovery. Different approaches can analyze data from different perspectives, such as exploratory, statistical,predictive, etc. In this way, the information obtained is visualized and interpreted in various forms. Clustering analysis is an important data mining technology. In the context of data stream,academia, business and government can find hidden valuable information from data and gain advantage with it. Clustering divides the datasets into clusters to make the intracluster similarity maximized and intercluster similarity minimized. The clusters adopted in the process can be used to understand the characteristics of data distribution. Various clustering methods are proposed and improved continuously, while different methods are suitable for different types of data.In this paper, we apply the Evolution-Communication Tissue P System to clustering analysis according to its structure features and parallel computing features.We focus our attention on the study of EC P systems in partition-based clustering and hierarchical clustering.We put forward a new EC Tissue P system of object controlling process, which utilize the flow characteristics of object through a channel to control the operations of rules, thus improve the computing efficiency.The main research work in this paper focuses on the followingaspects: Firstly, we modify the traditional EC P system in which object is viewed as control signal to conduct the rules execution flow. Secondly, the ranked K-medoids(RKM) algorithm based on the newly designed P system is constructed. We design the membrane structure,elements such as objects, rules, combine the advantages of RKM algorithm in dealing with globular clusters and distributed parallel computing characteristics of P system,which can achieve a better clustering effect. Thirdly, we propose a P system model to implement Hierarchical Euclidean Minimum Spanning Tree(HEMST) algorithm which clusters sub-trees iteratively. The MST clustering algorithm has advantages in finding irregular boundary clusters of which the constructed P system model can make use to accomplish better clustering. And we apply HEMST based P system model to high and new technology enterprises for hierarchical clustering analysis, and solve the realistic problems of enterprise level partitioning. The enterprises are divided into different development types. Then government and industry can execute regulation of different levels on enterprise, sequentially adopt corresponding tax, science and technology plan, finance and insurance, land policy and so on, therefore encourage and support the development of high and new technology industry. Fourthly, we construct an EC tissue P system based collaborative filtering recommendation model. Based on the users collaborative filtering recommender is a relatively successful recommender system, but also it is faced with the problem of scalability. With the number of users and commodities increasing, the performance of the system will be more and more low. The parallel computing feature of P systems can significantly improve the performance of the algorithm.