摘要
基于Shapelet的时间序列分类算法具有可解释性强、准确率高、速度快的优点,然而在Shapelet发现过程中存在Shapelet产生冗余和形式局限的缺点,严重制约了算法性能的提高。针对这一问题,提出一种基于优化Shapelet的时间序列分类算法,该方法首先利用K-means生成典型的Shapelet候选集,加速Shapelet的生成过程;然后,融合相似性和类标差异性提出Shapelet的选取模型,确保Shapelet的多样性和精简性;最后,提出优化策略获取最佳的Shapelet,并以此为基础实施时间序列分类。实验结果表明:该方法具有较高的分类准确率,并对位移和扭曲特征明显的数据集具有良好的分类效果。
The time series classification algorithm based on Shapelet has the advantages of strong interpretability,high accuracy,and fast speed.However,Shapelet redundancy and the limitation of Shapelet discovery seriously restrict the improvement of algorithm performance.To solve this problem,a time series classification algorithm based on the optimized Shapelet was proposed.Firstly,K-means was used to generate typical Shapelet candidate sets.Then,Shapelet optimization strategies were constructed by integrating similarity,class index difference,and classifier optimization to ensure the diversity and simplification of Shapelet.Finally,time series classification was implemented.Experimental results show that the proposed algorithm has higher classification accuracy and good classification performance on the data sets with obvious displacement and distortion characteristics.
作者
王威娜
胡佳利
任艳
WANG Wei-na;HU Jia-li;REN Yan(School of Information and Control Engineering,Jilin Institute of Chemical Technology,Jilin 132022,China;College of Automation,Shenyang Aerospace University,Shenyang 110136,China)
出处
《科学技术与工程》
北大核心
2023年第8期3345-3353,共9页
Science Technology and Engineering
基金
国家自然科学基金(62266046)
吉林省自然科学基金(YDZJ202201ZYTS603)
辽宁省自然科学基金(2020MS235)。