一种近邻传播的层次优化算法

Affinity Propagation Hierarchical Optimization Algorithm

近邻传播算法是一种新的聚类算法,在许多领域有较好的应用。近邻传播算法倾向于生成多于真实数目的类,且先验值P对该算法结果优劣有很大影响。故提出了一种有效的近邻传播的层次优化算法——CAP算法。CAP算法利用CURE算法对近邻传播算法的结果进行优化,是一种半监督的聚类算法。在5个UCI数据集上进行了实验验证,结果显示该算法均取得比近邻传播算法更好的聚类结果质量且使得生成的类的个数更接近真实类个数;同时与K-means、Spectral、CURE算法进行比较,结果表明CAP算法能取得更优的结果。

Affinity propagation（AP）clustering algorithm is a new clustering algorithm,and it is used in many fields well.Affinity propagation clustering algorithm tends to generate more classes than the real data sets.Phas a great influence on the result.So this paper proposed an effective affinity propagation clustering’s hierarchical optimization algorithm called as CAP.CAP algorithm uses the CURE algorithm to optimize the result of AP algorithm,and CAP is a semi-supervised clustering algorithm.The result of experiment on five UCI data sets shows that CAP algorithm achieves higher quality than AP algorithm and the number of classes is much closer to the real number.At the same time,CAP also achieves much better clustering result than K-means,Spectral and CURE.