Real-Time Change-Point Detection : A Deep Neural Network Based Adaptive Approach for Detecting Changes in Multivariate Time Series Data

The behavior of a time series may be affected by various factors, the most common of which are changes in the mean, variance, frequency, and auto-correlation. Change-Point Detection (CPD) aims to track down abrupt statistical characteristics changes in time-series data, which can benefit many applications of different domains. Deep learning approaches have the potential to identify more subtle changes, as demonstrated by recent deep learning CPD methodologies. However, due to improper handling of data and insufficient training, these methodologies generate more false alarm rates and are inefficient in detecting change-points with more precision. In real-time CPD algorithms, prepossessed data plays a vital role in increasing the algorithm’s efficiency and minimizing false alarm rates. Therefore, pre-processing of data should be a part of the algorithm, but in the existing methods, pre-processing of data is done, and then the whole data is passed to the CPD algorithm. To address these issues, a solution has been presented in which the pre-processing phase is combined with the CPD algorithm to make the entire process adaptive. Through experiments on real-world data sets and artifacts, it has been demonstrated that our proposed strategy is superior to existing methods and consistently outperforms them