The Research And Application Of Partitioning And Clustering Combination Algorithm Based On Non-metric Multidimensional Scaling

With the continuous development of science and technology,clustering,as one of the important analysis techniques in the field of machine learning artificial intelligence,is far from meeting the actual needs in the face of increasingly complex social and practical problems.This paper verify the effectiveness of the improved algorithm,which is improved from three aspects:expanding the data type to non-metric high-dimensional data,effectively combining the change rate function based on individual contour coefficient to determine the optimal number of clusters,and fully integrating the advantages of ant colony algorithm and particle swarm optimization to optimize the initial clustering center.The main contents and results of this paper are as follows:1.In terms of being solved non-metric high-dimensional data problems urgently,preprocess data,explain the non-metric multidimensional scaling technology,the principle and algorithm flow.In the one hand,non-metric multidimensional scaling is used to solve the problem of non-metric high-dimensional data.On the other hand,combining non-metric multidimensional scaling expands the adaptation range of K-Means clustering algorithm.2.Based on the clustering combination algorithm of non-metric multidimensional scaling,what is used for determining K value adaptively.The change rate of individual contour coefficient is used as the evaluation function.When the function converges,the optimal clustering number K is obtained by maximizing the change value of contour coefficient between adjacent K pairs.3.Has been successfully applied to various optimization problems about the effective fusion of particle swarm optimization and ant colony algorithm,so that integrate the respective advantages of PSO-ACO to optimize the initial clustering center.Firstly,the initial pheromone distribution is obtained by using particle swarm optimization with global and fast performance,and then the accurate solution is obtained by using ant colony algorithm with positive feedback and parallelism,after multiple iterations of the algorithm,the global optimal solution can be obtained with a large probability.The improved partitioning clustering algorithm is applied to the analysis of actual data and simulation data respectively.The experiment shows that the new algorithm can improve the solving ability and time efficiency effectively.In addition,the cluster performance experiment further verifies the advantages of the new algorithm based on large data platform.However,the different number of clusters in the particle swarm optimization algorithm in the fusion strategy,there may be a slight gap between the initial cluster centers.therefore,the iterative algorithm optimization problem needs further improvement and discussion..