Research And Application Of Multi-relation Clustering Algorithm Based On Membrane System

Membrane computing,as a new branch of natural computing,is a computational model that is abstracted from the collaboration of cell populations.Its purpose is to establish a distributed parallel computing model with good computational performance by drawing on and simulating the ways that the cells,tissues,organs or other biological structures deal with chemical substances.Due to the characteristics of distributed,maximal parallelism and strong fault tolerance,membrane computing has been applied in many fields,which has solved many practical problems.Traditional clustering methods usually assume that the data are independent of each other.However,most of the data are stored in a relational database in the form of multi-relation.The traditional clustering method can not meet the needs of the multi-relation data.In this paper,we study the problem of poor clustering quality and low clustering efficiency in multi relational clustering.The paper uses membrane system as the computing framework.Firstly,this paper improves the K-means clustering algorithm,And on this basis,proposes two efficient multi-relation clustering algorithms.Finally,the proposed algorithm is applied to collaborative filtering recommendation system.(1)K-means algorithm based on the optimal initial clustering center(OIK-means).First of all,this algorithm calculates the density of each object according to the similarity,selects candidate center according to the minimum distance between object and arbitrary high density object,and finally determines the K initial center point.The OIK-means algorithm is tested on the artificial data set and UCI data set,and compared with the traditional K-means algorithm.(2)Multi-relation clustering algorithm based on a comprehensive similarity(ISMC).The algorithm uses the idea of tuple ID propagation,which sets a weight for each table in the relation database,improves the traditional similarity calculation,and proposes a new similaritycalculation method.According to a certain weight,the inside-class similarity and the outside-class similarity of the object are integrated into a comprehensive similarity.Based on the comprehensive similarity method,the paper performes OIK-means clustering on the objects of the target table.The ISMC algorithm is tested on UCI data set Movie,and compared with TPC,ReCOM,and LinkClus algorithm.(3)Genetic K-means multi-relation clustering algorithm based on membrane system.The algorithm is based on the new view of the combination of membrane computing and multi-relation clustering algorithm,according to the improved similarity the algorithm ISMC proposed,improves the traditional K-means algorithm by genetic algorithm and the selection of K center point,makes full use of genetic algorithm,object attribute and relation information to improve the clustering accuracy.Then design the corrsponding membrane structure and membrane rules to help the realization of the multi-relation clustering algorithm.The GKM algorithm is tested on the UCI data set Movie,and compared with the ISMC,ReCOM and LinkClus algorithm.(4)An efficient collaborative filtering recommender algorithm based on membrane system and multi-relation clustering(MCMCF).The method makes full use of the characteristics of the maximum parallelism(Max)and the distributed execution of the membrane system.Comprehensive similarity calculation method can effectively solve the data sparsity problem.Multi-relation clustering can effectively reduce the nearest neighbor search scale.This algorithm improves the recommendation quality and operational efficiency.