Title: Changepoint detection for home activity count data: A functional approach
Authors: Israel Martinez-Hernandez - Lancaster University (United Kingdom) [presenting]
Abstract: Many companies aim to improve health and care of people by taking advantage of new technologies. We use a dataset from Howz; a company that helps older people to improve their health by identifying changes in their behavior over time. We propose a novel methodology to detect change points in the daily activities of people. Our goal is to detect changes across periods (days), and changes within each period are considered normal changes in daily activities. To this end, we model the cumulative activities for each period with a Cox process. Then, the sequence of the stochastic intensity functions results in a functional time series. Thus, our proposal uses this functional time series to detect changepoints across periods.