Accurate classification of time series based on a new robust similarity measure

作者：He Xiaoxu^*; Shao Chenxi; Xiong Yan

来源：Journal of Computational Information Systems, 2013, 9(9): 3521-3528.

DOI：10.12733/jcis5882

• Comprehensive comparisons
• Conventional classifier
• Experimental evaluation
• Nearest neighbor rule
• Similarity measure
• Testing effectiveness
• Time series classifications
• Time series data mining

Among different time series data mining, time series classification is one of the most important aspects. How to design a suitable similarity measure of similarity is a burning issue for accurate time series classification. In this paper, we propose a new similarity measure based on feature exaction of the original time series. The new similarity measure is used in classification with the nearest neighbor rule. In order to provide a comprehensive comparison, we conducted three sets of experiments, testing effectiveness on different time series datasets from a wide variety of application domains. Experimental evaluations show that the proposed similarity measure can tolerate the most distortions than other three typical similarity; besides, the new classifier is not only superior to other conventional classifier, but also more excellent in 1nn classifier than other similarity measure.

• 出版日期2013
• 单位中国科学技术大学