search for




 

Time series classification using wavelet transform
Journal of the Korean Data & Information Science Society 2021;32:943-52
Published online September 30, 2021;  https://doi.org/10.7465/jkdi.2021.32.5.943
© 2021 Korean Data and Information Science Society.

Se Eun Lim1 · Jonghwa Na2

1NH Nonghyup Finance
2Department of Information and Statistics, Chungbuk National University
Correspondence to: 1 Manager, NH Nonghyup Finance, Seoul 04517, Korea.
2 Professor, Department of Information & Statistics, Chungbuk National University, Chungbuk 28644, Korea. E-mail: cherin@chungbuk.ac.kr
Received July 22, 2021; Revised August 23, 2021; Accepted August 27, 2021.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/3.0) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.
Abstract
The application of data mining techniques to time-series data, which are being explosively produced from various digital devices, is a major concern. This includes the problem of classification for time series data. In this study, the method of using the discrete wavelet transform is introduced to the classification problem of time series data and applied to actual data analysis. Discrete wavelet transform is an advanced form of Fourier analysis that does not contain time information, and has the advantage of simultaneously increasing temporal resolution and frequency resolution, so it is effective in the classification problem of non-stationary time series data. Decision tree, naive Bayesian, k-NN, SVM, and random forest methods were used as classification models. And various measures, which include accuracy, kappa coefficient, F1-score, ARI, were used for performance comparison. As a result of applying these methods to time series data observing concentration of hydrochloric acid, it was confirmed that the method using discrete wavelets analysis performed better than the result of applying the classification model directly to the raw data or the result using the Fourier transform (or short time Fourier transform). The problem of selecting a mother wavelet, which is mainly handled in the process of applying the discrete wavelet transform, is also considered.
Keywords : Discrete wavelet analysis, fast Fourier transform, short time Fourier transform, time series classification.