基于簇中心群的时间序列数据分类方法

分类算法是时间序列数据挖掘中极为重要的任务和技术，该文提出一种基于簇中心群的时间序列数据分类方法。该方法根据时间序列训练数据集中的类别标签进行簇划分，利用近邻传播算法分别对每个簇进行中心代表点选择，构造出各代表点的代表对象集；然后借助基于动态时间弯曲的均值中心方法对各代表对象集实现中心群计算，结合改进后的K近邻算法实现时间序列数据的分类。数值实验结果表明，与传统方法相比，新方法具有更好的分类效果和计算性能。

Classification for Time Series Data Based on Center Sequences of Clusters

Classification algorithm is one of the important tasks and techniques in the field of time series data mining. A classification method for time series data based on center sequences of clusters is proposed in this paper. Time series in the training set are divided into several clusters according to their labels, and every cluster picks out the representation objects using affinity propagation clustering and constructs the representation subset. The barycenter averaging method based on dynamic time warping is used to calculate the center group in which the improved K nearest neighbors method is executed for time series classification. The experimental results demonstrated that the new method, compared to the traditional method, has better classification quality and calculation performance.