一种基于旋转最小-最大超盒的聚类算法

针对传统聚类算法无法解决复杂分布数据聚类的问题，本文提出了一种基于旋转最小-最大超盒的聚类算法。该算法首先对数据急归一化处理，然后借助旋转最小-最大超盒对样本空间进行划分，从而生成一系列的初始分类，最后利用类别间的可融合度概念将初始类别融合为最终的聚类。仿真结果表明，该算法在无需聚类数目的前提下，对复杂分布数据具有很好的聚类效果，其聚类性能与传统的FCA相比有极大的提高。

A Novel Clustering Algorithm Based-on Rotated Super-box

To resolve the problem that the traditional clustering algorithms cannot deal with the complicated data set, a novel clustering algorithm based\|on rotated super\|box is presented in this paper. This algorithm first uniforms the data set, then partitions the sample space with rotated super\|box to get a series of preliminary classes, finally fuses the preliminary classes with the fusible factors to form the clustering result. The simulation results show that this algorithm has better clustering performance than the traditional fuzzy C\|mean algorithm in dealing with the data set which has unknown pattern number and complicated distribution shapes.