Research On Optimization Of Soft Subspace Clustering Based On Flower Pollination Algorithm

With the development of information technology,data collection and storage technology,the scale of the data is gradually expanding,and the dimension is gradually increasing.Due to the sparsity of high dimensional data and the curse of dimensionality,traditional clustering algorithm can not achieve effective clustering.In order to solve the problem of high dimensional data clustering,soft subspace clustering analysis technology has emerged and has been widely concerned.Soft subspace clustering is aimed at finding a fuzzy subspace for each cluster by describing the uncertainty of the samples belonging to different clusters,which has better adaptability and flexibility,and is closer to the objective world.However,the existing soft subspace clustering algorithms mainly have the following two problems: the method of randomly selecting sample points to initialize cluster centers will lead to the clustering accuracy and stability of the algorithm depend on the initial cluster centers.And the local search strategy will make the algorithm easy to fall into local optimum in the search process.In order to solve the above problems,the main contents of this paper are as follows:Firstly,to solve the problem that the clustering result depends on the initial cluster centers,the Clustering by Fast Search and Find of Density Peaks(CFSFDP)is studied.By introducing the projection partition and the class merging technique,this paper proposes a new optimization of CFSFDP based on the Projection Partition and Merging Technique(PM-CFSFDP),which can be used to obtain more accurate cluster centers.The PM-CFSFDP is used as the initialization step in the soft subspace clustering to select more accurate cluster centers and reduce the dependence of the algorithm on the initial clustering centers.Secondly,to solve the problem of easy to fall into local optimal solution,the Flower Pollination Algorithm(FPA)is studied.By introducing the idea of shuffled frog leaping and adaptive gauss mutation strategy,this paper proposes a Flower Pollination Algorithm based on Adaptive Gauss Mutation and Shuffled Frog Leaping(AGM-SFLFPA).This method can avoid the local optimum in the search process,and the convergence speed is faster.The AGM-SFLFPA is used as a global optimization search strategy in soft subspace clustering to search the optimal weights and avoid the local optimum solution effectively.Finally,by introducing the two improved algorithms of PM-CFSFDP and AGM-SFLFPA,this paper proposes a Soft Subspace Clustering Algorithm based on Flower Pollination Algorithm(FPASC).The experimental results on the UCI standard datasets show that when dealing with high dimensional data,the proposed FPASC algorithm can effectively reduce the dependence on the initial cluster centers,avoid falling into the local optimum in the search process,and effectively improve the clustering accuracy and stability of the soft subspace algorithm.