Research And Application Of Evolutionary Clustering Algorithm

As a very effective data analysis method in the field of data mining,clustering has been widely used in the fields of pattern recognition,image processing and data compression.Clustering is the grouping of data objects into multiple classes or clusters.The principle of partitioning is that there is a high degree of similarity between objects in the same cluster,and the objects in different clusters are different.The traditional clustering algorithm can only deal with some static data,but it is still worthy for scholars to research the evolutionary data which has aroused the focus of machine learning and data mining in recent years.The evolution of data is that the distribution of data will change as time goes on,and the new data will appear the old one will disappear,so how to make every moment on the clustering performance as well as possible and reflect the data distribution of each moment basically.Through clustering to explore the evolution of data,such as the emergence,change,split and disappear of clustering.In order to make the clustering results as smooth as possible in time and the clustering results of the present time are similar to those of the previous time,a small number of scholars have been studied them.In this paper,we focus on the clustering problem of evolutionary data,and study two unsupervised evolutionary clustering algorithms and semi-supervised(constrained)evolutionary clustering algorithm,which have been applied simply.Specific research work and achievements are as follows:(1)In this paper,an evolutionary clustering framework based on temporal smoothness is proposed,which is based on the online framework proposed by Chakrabarti.In addition,this paper also defines the formula for the similarity matrix between data.And the similarity calculation includes two parts,that is,the similarity between the current time data and the similarity on the time series.Finally,the framework is applied to the standard spectral clustering,and two new evolutionary spectral clustering algorithms are proposed.(2)In this paper,we propose an evolutionary semi-supervised clustering algorithm via two-level random walk,which is used to deal with evolutionary clustering with constraint information.In the process of dealing with the continuously changing data,the original static double layer random walk semi supervised clustering algorithm will spend a lot of time and memory,and cannot achieve very good results.Based on the double-walk semi-supervised clustering algorithm,this paper makes good use of the information of the previous time,and obtains the old data information directly by solving the two similarities between the components when traveling at high level,which greatly reduce the computational time.At the same time,the semi supervised data can be processed better.(3)An Evolutionary Face Clustering System is designed,which is based on the application of evolutionary clustering algorithm proposed in this paper.The main functions of the system include three parts:data processing(evolution clustering),results display of recognition and file management.