使用蚁群优化和凝聚层次的混合聚类

为了获得全局最优的高质量层次聚类结果,针对智能蚁群优化算法改进凝聚层次聚类算法,以获得高质量的层次聚类结果,提出一种新的基于蚁群优化和凝聚层次聚类的混合聚类方法.该方法使用改进的凝聚层次聚类算法和新的目标函数生成聚类的系统树图,利用内部指标评估解决方案,用智能蚁群优化算法支持的信息素反馈和信息素挥发机制控制蚁群在解决方案空间中的搜索.由于使用了元启发式优化,加快了搜索过程,避免了局部最优.在加州大学欧文分校多个数据集上的实验结果表明,新方法具备一定的可行性.

Hybrid Clustering Using ACO and AHC

To study the use of intelligent ant-colony optimization (ACO) to improve agglomerative hierarchical clustering (AHC) and to attain high-quality cluster results of hierarchy, a hybrid clustering based on ACO and AHC (HCAA) is proposed. The modified AHC and a new objective function are used to generate the dendrogram of clusters and the internal index is utilized to evaluate the solution. The mechanism of pheromone feedback and pheromone volatilization supported by the ACO is employed to control the search of the ant colony in the solution space. The method will accelerate the search, avoiding the results of local optima because of using meta-heuristic optimization. Experiments on several datasets of university of California, Irvine verify the feasibility of this method.