基于扰动免疫粒子群和K均值的混合聚类算法

针对传统K均值聚类算法对初始化敏感和容易陷入局部最优的缺点，提出了一种基于扰动免疫粒子群和K均值的混合聚类算法。该算法采用K均值将粒子群进行分类，选择平均适应度值最高的聚类域用于产生疫苗，在粒子更新过程中采用疫苗接种机制和免疫选择机制提高粒子的多样性。当个体极值和全局极值连续停滞代数超过所设置的阀值时，算法使用扰动算子改变粒子群的运动方向，提高算法跳出局部极值的能力。当扰动次数达到设置的最大值时，对各个粒子进行K均值操作，提高收敛精度。实验结果表明，该算法具有较高的正确率和较好的稳定性。

Hybrid clustering algorithm based on disturbance immune particle swarm optimization and K-means

After analyzing the disadvantages of initialization sensitive and local extremum of the K-means algorithm, this paper proposes a hybrid clustering algorithm based on disturbance immune particle swarm optimization and K-means. The new clustering algorithm uses K-means to divide the particles into several categories and then chooses the optimal clustering domain to produce vaccine. After that, it adopts the vaccination and immune selection to improve the diversity of the particles. Meanwhile, in the algorithm, the disturbed arithmetic operators is introduced to break away from the local extremum by changing the movement of the particles when the times of the continuous stagnation exceed the threshold. The K-means clustering algorithm is employed to improve the convergence precision of the algorithm when the times of the disturbance meets the maximum. The experimental results show that the convergence accuracy and stability of the algorithm are good.