It is essential for many research fields to find the period of a data set. Many algorithms have been derived for solving related problems. Recently, scholars have reported that deep neural networks can achieve a performance similar to a human on image classification. In this paper, we report a period classification algorithm based on the convolutional neural networks (CNNs). We test its performance on the randomly-generated periodic time series data sets (PTSDs) that consist of periodic and polynomial components. Our results show that the algorithm can achieve 100% out-of-sample accuracy when the polynomial component of a PTSD does not dominate.
Paper
Full text
Periodic Time Series Data Classification By Deep Neural Network
Semantic Scholar · Computer Science · 2019
Abstract
It is essential for many research fields to find the period of a data set. Many algorithms have been derived for solving related problems. Recently, scholars have reported that deep neural networks can achieve a performance similar to a human on image classification. In this paper, we report a period classification algorithm based on the convolutional neural networks (CNNs). We test its performance on the randomly-generated periodic time series data sets (PTSDs) that consist of periodic and polynomial components. Our results show that the algorithm can achieve 100% out-of-sample accuracy when the polynomial component of a PTSD does not dominate.