Change Point Detection in Non-Stationary, Multivariate IoT Time Series Data for Prescriptive Analytical Models

Prescriptive analytics is a major advancement in the field of IoT analytics which enhances decision-making and increases the effectiveness of processes. IoT data is referred to as dynamic, non-stationary, and multivariate time series data since it continuously varies over time and dramatically increases in volume with multiple attributes. To help identify IoT data trends and change points in data streams so that the right actions may be linked to them, time-series data analysis is needed. The temporal data analysis presented in this paper aims to find patterns in the data and understand their evolution. An effort is made in this manner to express time-series data by shortening its length through the piecewise aggregate approximation method without loss of any key information and then the sliding window approach is applied to detect change points by analyzing the cumulative statistics of time-series distribution.