Research On Evolutionary Multi-objective Clustering Algorithm In Cloud Computing Environment

Clustering technology in data mining is an important means of data analysis.Due to the different distribution characteristics of different data sets,the traditional single target clustering can not adapt to the effective processing of different data sets,so multi-objective clustering has gradually become a research hotspot.With the development of evolutionary multi-objective optimization technology,evolutionary multi-objective optimization technology has provided an effective solution for multi-objective clustering analysis.Cloud computing has the characteristics of distribution and big data.The appearance of cloud computing meets the needs of mass data processing,and the parallel computing of data has been applied and developed in all walks of life.Because the distribution characteristics of the data set processed by clustering is not known,the limitation of the single clustering evaluation makes the calculation result not very ideal.Through the study of evolutionary multi-objective optimization technology,this paper adds the clustering evaluation function to the traditional single objective clustering K-Means algorithm so that it can adapt to the processing of different data sets.Based on the framework of evolutionary multi-objective optimization technology,the artificial bee colony evolution model is introduced,and the multi-objective artificial bee colony algorithm model is constructed.We use multi-objective artificial bee colony algorithm to find the K-Means clustering center,weaken the dependence of K-Means on initial cluster centers,and get the desired clustering results.Different single target and multi-objective clustering algorithms are selected to test and evaluate the algorithm proved in this papert on different data sets.Finally,the convergence speed,accuracy and different performance indexes of the algorithm have been proved to have higher index and accuracy.The parallelization scheme of multi-objective artificial bee colony clustering algorithm are designed on this paper,builds the experimental environment with the Hadoop open source cloud computing platform,and implements the parallel function of the algorithm by using the MapReduce programming model.In the experiment,we choose the dataset with large amount to test the parallel algorithm.The experiment shows that the better effect of the parallel clustering algorithm on the data set with the larger amount from the algorithm running time,accuracy rate and acceleration ratio under different nodes.