基于蚁群优化算法的凝聚型层次聚类

Agglomerative hierarchical clustering based on ant colony optimization algorithm

传统的凝聚型层次聚类在分裂或合并类时如果没有很好地作出决定,就有可能导致低质量的聚类结果,针对这一缺点,提出一种基于蚁群优化算法的凝聚型层次聚类算法。该算法先利用蚁群优化算法的状态转移规则决定凝聚型层次聚类中下一个将要合并的数据点,再利用信息素更新规则寻找聚类的最优路径,最后获得全局最优的高质量层次聚类结果。该优化算法在人工数据集和UCI数据集上的仿真实验结果表明,相对于传统的聚类算法,该算法的准确率更高,聚类效果更好。

Traditional agglomerative hierarchical clustering may get low quality clustering result if it doesn＇ t make good decision when splits or merges clusters. According to this disadvantage, this paper proposed an agglomerative hierarchical clustering algorithm based on ant colony optimization. The algorithm used the state transition rule to determine points in the clustering hierarchy, and then used the pheromone update rule to find the optimal path, finally it could receive high quality hierarchical clustering results for global optimization. Simulation experiments and resuhs on the artificial data set and UCI data set show that compared with traditional clustering algorithm, the accuracy of the proposed algorithm is higher, clustering effect is better.