基于趋势特征表示的shapelet分类方法

发布时间：2018-03-21 03:25

本文选题：shapelet　切入点：趋势特征　出处：《计算机应用》2017年08期 　论文类型：期刊论文

【摘要】：Shapelet是一种具有辨识性的时间序列子序列,通过识别局部特征达到对时间序列准确分类的目的。原始shapelet发现算法效率较低,大量工作关注于提高shapelet发现的效率。然而,对于带有趋势变化的时间序列,采用典型的时间序列表示方法进行shapelet发现,容易造成序列中趋势信息的丢失。为了解决时间序列趋势信息丢失的问题,提出一种基于趋势特征的多样化top-k shapelet分类方法:首先采用趋势特征符号化方法对时间序列的趋势信息进行表示;然后针对序列的趋势特征符号获取shapelet候选集合;最后通过引入多样化top-k查询算法从候选集中选取k个最具代表性的shapelets。在时间序列的分类实验中,与传统分类算法相比,所提方法在11个数据集上的分类准确率均有提升;与Fast Shapelet算法相比,提升了运行效率,缩短了算法的运行时间,并在趋势信息明显的数据上效果显著。结果表明,所提方法能有效提高时间序列的分类准确率,提升算法运行效率。
[Abstract]:Shapelet is an identifiable subsequence of time series, which can accurately classify time series by recognizing local features. The original shapelet discovery algorithm is inefficient, and a lot of work is focused on improving the efficiency of shapelet discovery. In order to solve the problem of trend information loss in time series, the typical time series representation method is used for shapelet discovery, which is easy to cause the loss of trend information in time series. In this paper, a new method of top-k shapelet classification based on trend feature is proposed. Firstly, trend feature symbolization is used to represent the trend information of time series, and then the shapelet candidate set is obtained according to the trend feature symbol of the sequence. Finally, by introducing a variety of top-k query algorithms to select the most representative shapelets from the candidate set, the classification accuracy of the proposed method in 11 datasets is improved compared with the traditional classification algorithm in the experiment of time series classification. Compared with the Fast Shapelet algorithm, the algorithm improves the running efficiency, shortens the running time of the algorithm, and has a remarkable effect on the data with obvious trend information. The results show that the proposed method can effectively improve the classification accuracy of time series. Improve the efficiency of the algorithm.
【作者单位】：中国矿业大学计算机科学与技术学院;
【基金】：国家重点研发计划项目(2016YFC060908) 国家自然科学基金资助项目(61402482,61572505,52674255) 江苏省自然科学基金资助项目(BK20140192)~~
【分类号】：O211.61

【参考文献】