摘要： 提出了一种连续更新的过聚类和合并相结合的竞争学习的聚类算法(ONOPTOC)。该算法依据OPTOC的原则，构建原型点的动态邻接区域和动态背离区域，通过自分裂和聚合的策略，识别数据集中的天然聚类并给出聚类的数目。采用该算法对批过程中的历史数据进行分析，原型点的特征与对应的产品表达水平等级相符，有效地检测了该批次产品的表达水平，为当前的决策提供依据。

关键词: 聚类, 过聚类, 原型点, 邻接区域, 背离区域

Abstract:

A consecutively updated overclustering, self-splitting and merging algorithm based on one-prototype-take-one-cluster(OPTOC) is proposed. It is achieved by constructing a dynamic neighborhood and a distant neighborhood. And using overclustering and merging strategy, it can find natural clusters and give the correct number of clusters. The proposed method is applied to inspect historical datasets, and the results show that the prototypes accord with the corresponding expressions. The method is an effective method for inspect the level of products expression.

Key words: Clustering, Overclustering, Prototype, Neighborhood, Distant neighborhood

中图分类号: 

• TP391