关键词: 聚类分析, 遗传算法, 粗糙集, k-均值算法

Abstract: Traditional clustering methods use hard calculations to divide data objects，but in reality，the objects of different classes often do not have clear boundaries between different kinds of clusters.Rough set theory provides a method of dealing with uncertain boundary objects.Therefore，the rough theory and k-means method are combined.Meanwhile，the traditional k-means clustering method must be given in advance the number of clusters k，but in the actual cases，k is difficult to establish；In addition，traditional k-means algorithm has powerful local search capability，but easily falls into local optimum.Genetic algorithm can get the global optimal solution，but the convergence is fast.In view of this，this paper presents an improved rough clustering method based on genetic algorithm.The algorithm can dynamically generate the number of k-means clustering，using max-min principle to generate the initial cluster centers.Rough set theory’s upper and lower approximation set is combined to deal with the boundary object.Finally，the UCI’s Iris set is used to test the algorithm.The experimental results show that the algorithm has higher accuracy rate and more stable performance.

Key words: cluster analysis, genetic algorithm, rough set, k-means