具有全局优化能力的K均值聚类算法

传统的K均值聚类算法是确定性的迭代算法,具有探索能力弱、容易陷入局部最优的缺点.在聚类中心的更新过程中加入系数因子线性递减的随机项,使改进的迭代算法在前期具有强的探索能力,而在后期保持良好的局部搜索能力,同时保持了传统K均值聚类算法结构简单的特点.实例说明,增加了随机项的K均值聚类算法具有良好的全局优化能力.

On K-Means Clustering Algorithm with Global Optimization Ability

The traditional K-mean clustering algorithm is a deterministic iterative algorithm .Its exploration ability is week ,and it is easy to find locally optimal solution .In the process renewing centers of cluste-ring ,a random term with coefficient factor linearly decreasing is added .So ,the improving iterative algo-rithm has strong exploring ability in the early stage ,and it has good local search ability in the later stage . At the same time ,the improving algorithm holds the simple algorithm structure possessed by the tradition-al K-mean clustering algorithm .An example indicates that the random K-mean clustering algorithm has excellent global optimization ability .