Study Of Hybrid Clustering Algorithm Based On Attribute Weight

The goal of clustering is to collect data on the basis of similar classification,the difference between the various classes of data should be as large as possible,the difference between the within class data should be as small as possible,that is to minimize inter class similarity maximization within class similarity principle.At present,there are a large number of classical clustering methods,the choice of the algorithm depends on the type of data,the purpose of clustering and application.For example,K-means,BIRCH,CURE,DBSCAN,COBWEB,etc.,for the same data set,the use of different clustering algorithms may have different results.FCM algorithm is the most widely used clustering algorithm.The study found that the traditional FCM algorithm has two problems: first,the algorithm from the sample point,by optimizing the objective function calculation for the class membership of each sample point is the center,so as to achieve the purpose of classification,improper selection of initial value will lead to convergence to local minima;second,clustering analysis of each attribute contribution the sample data is not the same as the Euclidean distance FCM algorithm using standard calculation ignores the attribute weights of clustering results.Therefore,in essence,the FCM algorithm is a local search optimization algorithm.Based on the above analysis,this paper proposes a hybrid clustering algorithm based on attribute weights.The main research contents are as follows:(1)Improved particle swarm optimization algorithm introduces the concept of particle evolution and particle grouping,which provides the basis for the calculation of attribute weights.(2)Feature weight learning algorithm: the improved particle swarm optimization algorithm,the position vector of the particle as attribute weights,attribute weights using cross entropy as the evaluation function,using the gradient method to minimize the attribute weights by iterative evaluation function decline,obtain a set of optimal attribute weights.(3)Hybrid clustering algorithm: combining genetic algorithm and simulated annealing algorithm,using FCM clustering algorithm,the initial cluster center mapped to chromosome,the objective function as the fitness function of the genetic algorithm,through selection,crossover and mutation,calculate the clustering center,membership and individual fitness values by using the FCM clustering algorithm,simulated annealing algorithm is used to a certain probability of receiving new individuals through iteration,finally get the global optimal solution.