A novel prediction method of complex univariate time series based on k -means clustering
- PDF / 2,176,440 Bytes
- 13 Pages / 595.276 x 790.866 pts Page_size
- 94 Downloads / 193 Views
(0123456789().,-volV)(0123456789(). ,- volV)
METHODOLOGIES AND APPLICATION
A novel prediction method of complex univariate time series based on k-means clustering Yunxin Liu1,2 • Shifei Ding1,2 • Weikuan Jia3
Ó Springer-Verlag GmbH Germany, part of Springer Nature 2020
Abstract Time-series prediction has been widely studied and applied in various fields. For the time series with high acquisition frequency and high noise, it is very difficult to establish a prediction model directly. Therefore, it is necessary to study how to obtain the change trend information of time series accurately, and then build a prediction model for its change trend. To obtain the change trend information of the original time series effectively and establish an accurate prediction model, this paper proposes a novel prediction method of complex univariate time series based on K-means clustering. This method first obtains the change trend information of the original time series based on the K-means clustering idea, and then, a gated recurrent unit based on the input attention mechanism is used to establish a prediction model for the obtained time-series change trend information. Extensive experiments on the electromagnetic radiation dataset we collected, the AEP_hourly dataset, and the Wind Turbine Scada dataset published online, demonstrate that our proposed K-means clustering method can effectively reduce noise interference and accurately obtain the time-series change trend information. Comparative experiments of different prediction models demonstrate that our prediction model has the best prediction accuracy, and our proposed complex univariate time-series prediction algorithm has great practical value. Keywords Time series Change trend prediction K-means clustering Attention mechanism Gated recurrent unit
1 Introduction Nowadays, with the development of the Internet of things, we can easily use sensors to collect large amounts of time series. However, due to the high acquisition frequency of the sensor, the collected time series has a high density and large data similarity, resulting in low utilization value. Moreover, due to the influence of factors such as complex environment, it is often mixed with considerable noise, so
Communicated by V. Loia. & Shifei Ding [email protected] 1
School of Computer Science and Technology, China University of Mining and Technology, Xuzhou 221116, China
2
Mine Digitization Engineering Research Center of Ministry of Education of the People’s Republic of China, Xuzhou 221116, China
3
School of Information Science and Engineering, Shandong Normal University, Jinan 250358, China
it is difficult to establish an accurate model directly, such as the electromagnetic radiation data collected in the coal mine. The change of electromagnetic radiation data can be used to monitor mine dynamic disasters (Wang et al. 2000; Morrison et al. 1994); when the magnitude of electromagnetic radiation reaches a certain value, there will be dangerous situations. Therefore, accurate prediction of its future m
Data Loading...
