提出一种改进的求解聚类问题的萤火虫群优化算法,该算法借鉴粒子群优化算法的思想，对聚类中心采用实数编码和解码方法；用线性递减的移动步长代替固定步长，萤火虫的更新位置由动态决策域和全局最优位置共同决定代替仅由动态决策域决定；并加入孤立点的移动策略，使得孤立点可以向最优值方向移动.将该算法与粒子群优化算法、基本的萤火虫群优化算法在UCI数据集上进行对比试验，结果表明改进的萤火虫群优化算法可以取得较好的聚类效果.

An improved glowworm swarm optimization algorithm (IGSOA)for clustering problem is propsed.In this algorithm,real-coded and decoded methods are used for the cluster center,linearly decreasing steps are adopted instead of fixed steps inspired by particle swarm optimization (PSO) algorithm,the position of the glowworm is updated based on both dynamic decision domains and global position in place of dynamic decision merely,and also the outliers′ movement strategy is added to the algorithm making its movement to the optima. The simulation result on UCI datasets demonstrates that the new method performs better than PSO and the basic GSO algorithms.