基于云模型改进的粒子群K均值聚类算法

研究粒子群K均值聚类算法问题，针对传统粒子群K均值算法容易陷入局部最优解，出现早熟收敛的缺点，提出一种基于云模型改进的粒子群K均值聚类算法．使用X条件云发生器自适应地调整粒子个体惯性权重的方法．保证惯性权重会逐渐减小而又不失随饥性。根据个体适应度的优劣将粒子群分为三个子群，在每次迭代时都保证仍有一个子群的粒子在进行全局搜索，避免算法陷入局部最优和早熟收敛。在典型数据集上的仿真结果表明，改进算法相比其他聚类算法得到较好的聚类准确率和较快的收敛速度，是一种行之有效的方法。

K-Means Clustering Algorithm of Improved Particle Swarm Based on Cloud Model

Analyzes the PSO K-means clustering algorithm, in order to overcome the shortcomings of traditional PSO K-means,such as falling into local optimal solution easily and premature convergence, proposes a modified K-means algorithm. The algorithm uses cloud model to ad- just individual particle inertia weighl adaptively, ensuring that the inertia weight will be gradually reduced without losing the randomness. Dividing particle swarm into three subgroups based on the fitness of each particle, it is ensured that there are particles from one subgroup conducting a global search in each iteration, avoiding falling into local optimal solution and premature convergence. Simulating results on the UCI of data sets show that CPSO K-means has better clustering accuracy and faster convergence rate compared with other clustering algorithms.