论文部分内容阅读
针对传统K-means算法过度依赖初始聚类中心、易陷入局部最优、不能处理边界对象及聚类精度低等问题,提出一种结合粒子群和粗糙集的聚类算法.此算法首先利用密度和最大距离积法初始化粒子群;然后采用线性递减与随机分布相结合的惯性权重、动态调整的学习因子和引入的随机粒子来避免陷入局部极值,使算法快速收敛于全局最优;最后结合粒子群和粗糙集来优化K-means算法.通过对几个常用UCI标准数据集的仿真实验表明,提出的算法不仅能减少对初始聚类中心的依赖、有效抑制局部收敛,而且聚类准确率更高,稳定性更强.
Aiming at the problems that the traditional K-means algorithm is overly dependent on the initial clustering center, easy to fall into local optimum, can not handle boundary objects and the clustering accuracy is low, a clustering algorithm combining particle swarm optimization and rough set is proposed. And the maximum distance product method to initialize the particle swarm. Then, the inertia weight, the dynamically adjusted learning factor and the introduced random particle are used to avoid falling into the local extremum by the linear reduction and random distribution, so that the algorithm converges to the global optimum quickly. Finally, Particle Swarm Optimization and Rough Set to optimize the K-means algorithm.According to several commonly used UCI standard data sets, simulation results show that the proposed algorithm can not only reduce the dependence on the initial clustering center, but also restrain the local convergence effectively, and the clustering accuracy Higher, more stable.