A Comparative Study on Change-Point Detection Methods in Time Series Data

Abstract

The Time-series data is a sequence of data points at regular time intervals indexed in time order. It is also known as time-stamped data. These sequential data characteristics might change during the process. Change points in time series data are substantial statistical property changes in the data. Many applications rely on the detection of these changes for appropriate modeling and prediction. Many vital activities can be monitored with the help of Change-Point Detection (CPD) algorithms, and appropriate actions can be made as a response. There are a variety of methods for detecting CPD in time series, which are divided into supervised and unsupervised categories. This comparative study compares all of the algorithms that have been published in the literature. Many novel algorithms based on the results of deep learning are also evaluated. Finally, we give the community some challenges to ponder.