Distillation of Random Projection Filter Bank for Time Series Classification
Time series is widely found in various fields such as geoscience,medicine,finance,and social sciences.How to effectively extract the features of time series remains a challenge due to its potentially complex non-linear dynamics.Recently,Random Projection Filter Bank(RPFB)5 is proposed as a generic and simple approach to extract features from time series data.It generates the features by randomly generating numerous autoregressive filters that are convolved with input time series.Such numerous random filters inevitably have redundancy and lead to the increased computational cost of the classifier.In this paper,we propose a distillation method of RPFB,named D-RPFB,to not only maintain the high level of quantity of the filters,but also reduce the redundancy of the filters while improving precision.We demonstrate the efficacy of the features extracted by D-RPFB via extensive experimental evaluation in three different areas of time series data with three traditional classifiers(i.e.,Logistic Regression(LR)2,Support Vector Machine(SVM)14 and Random Forest(RF)8).
Random projection Filter bank Time series Feature extraction
Yufei Lin Sen Li Qianli Ma
School of Computer Science and Engineering,South China University of Technology,Guangzhou,China School of Computer Science and Engineering,South China University of Technology,Guangzhou,China;Guan
国际会议
广州
英文
586-596
2018-11-23（万方平台首次上网日期，不代表论文的发表时间）