workshop

順序保存カーネルを用いた時系列データ分類

柏葉祐輝, 成澤和志, 篠原 歩

人工知能学会人工知能基本問題研究会, SIG-FPAI Vol.B502-04, pp.1-6 (2016)
開催日: 2016年1月21-22日

開催地: 仙台市
  • SIGFPAI
PDF DOI: 10.11517/jsaifpai.99.0_04

Abstract / 概要

In this paper, we propose a new similarity measure for time series data, that is called $k$-gram order preserving kernel. This kernel depends on the similarity of the shapes of data instead of that of the values. Moreover, we propose a new classification method using the kernel without adjusting the parameter $k$. Furthermore, we confirm the superiority of the proposed method by computer experiment.