Fuzzy Soft Subspace Clustering Considering Between-cluster Balance Of Information And Separation

Clustering technology has been widely studied and applied in many fields.Firstly,this paper summarizes the background of relevant research for the clustering problem,and the commonly used clustering algorithms are analyzed and introduced to some extent.We mainly research k-nn algorithm and k-means algorithm.And we proprosed an feature-weighted clustering algorithm.In clustering problem,fuzzy C means clustering algorithm(FCM)is one of the most commonly used data clustering techniques.It improves the traditional C means clustering algorithm by introducing the concept of fuzzy,so that the performance of the clustering algorithm has been greatly improved.Although a series of deformation algorithm of FCM are proposed from different angles,the clustering performance is improved for different clustering problems.However,there are still three key problems that affect the clustering performance of FCM algorithm.In these problems for specific performance:1.it is sensitive to the prior distribution of clustering data cluster.2.it is sensitive to the prior probability of clustering data cluster.3.it is easy to form the partition result of data cluster with poor differentiation.When faced with such a clustering problem,such as the prior distribution characteristics of the data cluster forming clustering is differ or when the prior distribution is not balanced,the performance of FCM algorithm is not satisfactory.In order to overcome this problem,this paper proposes a fuzzy soft subspace clustering considering between-cluster balance of information and separation algorithm(EWFCM).The advantages of our proposed algorithm are mainly embodied in the following aspects:In this paper,the distance between data clusters is introduced to highlight the separability between different data clusters.Based on this,we propose the distribution degree of information equilibrium data cluster in this paper,and a method to calculate the separability of different data clusters;And the soft subspace is introduced to improve the description ability of data clusters with different distribution features.In the process of modeling introduced the distance between different data clusters,the discernibility degree between subspace description model and the subspace can be considered synthetically,and with unified optimization.In this paper,out proposed method fuzzy soft subspace clustering considering between-cluster balance of information and separation algorithm will be compared with the traditional FCM algorithm and some other relevant algorithms,and some comparative experiments will be conducted.We test our method and other methods by using artificially generated data sets with different shapes and data distributions,synthetic images,real images,and Iris data set.The experimental results showed that the performance of our proposed method is superior to both traditional FCM and most of enhanced methods.