摘要： 提出了一种基于增长型分层自组织映射(GHSOM)的时间序列聚类方法，给出了该方法的基本原理和具体算法步骤，对实测时间序列数据进行了聚类验证和分析。研究结果表明，增长型分层自组织映射能根据对象特征无监督地对时间序列进行正确聚类，由于具有动态增长及分层特性，能分析对象内在的层次结构并实现由粗到精的聚类，可以扩展应用于大型乃至巨量时间序列数据库的模式发现。

关键词: 时间序列, 模式发现, 增长型分层自组织映射, 聚类

Abstract: A novel technique based on growing hierarchical self-organizing map (GHSOM) for clustering of time series is presented, and its algorithm is introduced. Experiment with clustering based on SOM and GHSOM is implemented on a real time series. The result shows that GHSOM is effective in the time series clustering; it can adapt its architecture during its unsupervised training process according to the particular requirements of the input data, so the training time is reduced. Furthermore, the representation of hierarchical relation in the data is provided.

Key words: Time series, Patterns recovery, Growing hierarchical self-organizing map, Clustering