Research On Mixed Clusterinig Algorithm Clustering Based On Partheno Genetic Algorithm

Clustering analysis, as an unsupervised classification method, is an important branch of data mining areas and is widely applied in various trades. K-means clustering algorithm is simple and understandable as one of the main algorithms of clustering analysis, but there are still some shortcomings:It is sensitive to initial center and falls into the local optimal value easily, and it needs to confirm cluster number in advance. Because of these disadvantages, this paper proposes a mixed cluster analysis algorithm based on partheno genetic algorithm and k-means clustering algorithm. The effectiveness of the proposed algorithm is verified through the simulation results.Innovative work of this article:1. The relevant properties and features of the cluster analysis are studied and the characteristic and improvement strategies of the k-means clustering algorithms are analyzed in detail; 2. How to solve clustering problem through genetic algorithm is analyzed in detail.3. The partheno genetic algorithm is presented to solve the problem of clustering division, and the coding method was improved in the paper. A kind of new population initialization methods is designed, which improves the efficiency of the algorithm. The new mutation operator is constructed and the k value and the clustering are optimized. K-means operating is introduced as the local optimal operator.4. The effectiveness of the algorithm is verified:It makes a performance test of k-means and the mixed clustering algorithm based on partheno genetic algorithm on the same data. Experimental results show that it has the high clustering accuracy and the convergence speed is also significantly improved in the mixed clustering algorithm based on partheno genetic algorithm, and the algorithm has obvious superiority.