TY - CONF
T1 - Prediction of Time Series Data Based on Transformer with Soft Dynamic Time Wrapping
AU - Ho, Kuo Hao
AU - Huang, Pei Shu
AU - Wu, I-Chen
AU - Wang, Feng-Jian
N1 - Funding Information:
This research is partially supported by the Ministry of Science and Technology (MOST) of Taiwan under Grant Number 107-2634-F-009-011, 108-2634-F-009-011, and 109-2634-F-009-019 through Pervasive Artificial Intelligence Research (PAIR) Labs, and also partially supported by CHANGING.AI company.
PY - 2020/9/28
Y1 - 2020/9/28
N2 - It is a challenge to predict the long-term future data from time series data. This paper proposes to use a Transformer with soft dynamic time wrapping for early stopping criteria, called a soft-DTW Transformer. Our experiment in an open-source dataset HouseTwenty shows that the average prediction error rate with soft-DTW Transformer is 27.79%, greatly reduced from 45.70% for using SVR, a common time series method.
AB - It is a challenge to predict the long-term future data from time series data. This paper proposes to use a Transformer with soft dynamic time wrapping for early stopping criteria, called a soft-DTW Transformer. Our experiment in an open-source dataset HouseTwenty shows that the average prediction error rate with soft-DTW Transformer is 27.79%, greatly reduced from 45.70% for using SVR, a common time series method.
UR - http://www.scopus.com/inward/record.url?scp=85098465311&partnerID=8YFLogxK
U2 - 10.1109/ICCE-Taiwan49838.2020.9258155
DO - 10.1109/ICCE-Taiwan49838.2020.9258155
M3 - Paper
AN - SCOPUS:85098465311
T2 - 7th IEEE International Conference on Consumer Electronics - Taiwan, ICCE-Taiwan 2020
Y2 - 28 September 2020 through 30 September 2020
ER -