Research On Fuzzy Clustering Algorithm Based On Feature Weighting

The importance difference of each dimensional feature was ignored by the traditional fuzzy c-means clustering algorithm.However,in practical application,this difference is often more important.To solve this problem,the fuzzy clustering algorithm based on feature weight was proposed.This kind of algorithm can identify the impact of each feature on the clustering results,but it still has some shortcomings: on the one hand,it is sensitive to the initialization process,on the other hand,it is difficult to ensure the global optimization of the clustering results.Under this influence,the clustering results are very dependent on the selection of clustering centers,and the clustering center search based on gradient descent is easy to make the clustering algorithm fall into local optimization.This topic focuses on the fuzzy clustering algorithm based on feature weight,the research contents are as follows:In order to reduce the sensitivity of initialization process and improve the robustness of the algorithm,a feature weighted fuzzy clustering algorithm combined with multi strategy fusion gray wolf optimization is proposed.Firstly,aiming at the problems of low precision and slow convergence in the process of seeking the ideal optimal solution of gray wolf optimization algorithm,a variety of optimization strategies are introduced.Then it is applied to the centroid updating process of feature weighted fuzzy clustering algorithm.In order to verify the effectiveness of the algorithm,it is compared with other clustering algorithms on six data sets.The experimental results show that the algorithm has better clustering accuracy and robustness.Because bayesian fuzzy clustering algorithm does not consider the importance difference of different features,bayesian inference and markov chain monte carlo random sampling technology are introduced into feature weighted fuzzy clustering,and feature weighted bayesian fuzzy clustering is proposed.Firstly,the algorithm combines probability with fuzzy method to establish the equivalent relationship between them;Then,it breaks through the limitation that the value of parameter m must be greater than1 in traditional fuzzy clustering;Finally,the feature weighting mechanism is introduced to improve the clustering accuracy.In order to test the performance of the algorithm,experiments are carried out with 2 synthetic data sets and 16 UCI data sets.The experimental results show that the four indexes of F-measure,RI,Accuracy and Purity of the experimental results of the algorithm are better than the other four comparison algorithms.In order to avoid falling into local optimization,an improved feature weighted bayesian fuzzy clustering algorithm is designed,and the feature weight of samples is obtained by MCMC sampling method.The experimental results show that the improved feature weighted bayesian fuzzy clustering algorithm has better clustering effect.There are 6 figures,15 tables and 78 references.