Clustering Analysis Of K-means Based On Improved Genetic Algorithm

The rapid development of science and the improved of storage technology that make date to get rapid expansion. The embarrassing situation of "Massive data, lack of knowledge," makes data mining came into being. Data mining is used to extract potentially valuable information to people from these massive data, bringing a lot of convenience for our production and life. Cluster analysis is an important data mining tools, the use of cluster analysis can identify potential relationships between data attributes. K-means clustering algorithm is a typical algorithm that has fast convergence, strong local search ability and so on. But K-means algorithm exist some shortcomings, such as the initial center highlighted a sense, easy to fall into local optimum. This paper is aimed at the shortage of K-means algorithm, making some improved methods, mainly for research in the following areas:On the basis of the traditional K-means clustering algorithm, this paper proposed a K-means clustering algorithm based on genetic algorithm. Genetic algorithms have good global optimization capability, using competition operator to save the best chromosomes, while crossover and mutation algorithms guaranteed to produce better chromosomes. Through a series of genetic manipulation to find the optimal initial cluster centers, and then perform K-means algorithm to find the final cluster centers. It has some advantage, such as improving the accuracy of clustering, accelerating the convergence speed, increasing the stability algorithm. Simulation results show the effectiveness of GA-Kmeans-clustering algorithm.Combining shuffled frog leaping algorithm with K-means algorithm, this paper presents a hybrid frog leaping algorithm based on K-means clustering algorithm. SFLA is an intelligent optimization algorithms, not only has the ability of global optimization algorithm with GA, but also has a local search capabilities. Chaos search optimization based on the initial solution and fitness variance to determine whether the population K-means algorithm operation. Improve the accuracy of the algorithm, and the algorithm more stable.