The Research Of Sparse Subspace Clustering Method For Multi-view Data

Clustering analysis is one of the important research directions in the field of data mining.In recent years,with the development of big data analysis and mining research,clustering analysis has attracted great attention.The purpose of clustering is to segment the data set into several clusters,so that the data points in the same cluster are as similar as possible,and the data points between different clusters are as different as possible.Subspace clustering method is a hot topic in the research of clustering analysis.It assumes that the data are drawn from multiple low-dimensional subspaces,and the significance of subspace clustering is investigated from different subspaces.Sparse Subspace Clustering(SSC)is one of the representative methods,the basic idea of these methods is to learn an affinity matrix from the given data by their sparse or low-rank self-representation and then obtain the clustering result by applying spectral clustering algorithms.Sparse subspace clustering method has been successfully applied in the clustering of face,handwriting and texture images.Although sparse subspace clustering method performs well in many clustering applications,these methods mainly focus on the clustering of single view data.In the real world,many data could be observed in different views or can be represented by multiple features,which is often referred to as multi-view data.Multi-view data not only contains the general information,but also contains the complementary information of different views.Therefore,it is more suitable for clustering analysis of data.The existing multi-view clustering method is used to study the consistency information or the diversity information of multi-view data,but it has not considered the integration and utilization of the consistent and diversiform information.To solve the problem of multi-view data clustering,this paper proposes a new multi-view subspace clustering method based on consistency and diversity constraints.In this method,the commonness and difference of data are considered in the subspace clustering simultaneously.The adaptive adjustment mechanism of model parameters has also been discussed.The main contributions of this paper are summarized as follows:(1)In this paper,a multi-view subspace clustering method is proposed to solve the problem of multi-view data clustering,this paper presents a fusion framework with consistency and diversity constraints of multi-view data.(2)In this paper,the parameter adjustment mechanism of consistency and diversity constraint in the model is studied,a multi-view subspace clustering with adaptive weight based on data attributes and a multi view subspace clustering algorithm based on dynamic weight iteration are proposed to adaptive the weight of the consistency and diversity ConstraintsIn order to test the proposed clustering method,the clustering experiments are carried out on the open multi-feature data sets and multi-angle of view data sets.The experimental results are compared with the related clustering methods.The experimental results show the effectiveness of the proposed method.