Detecção de pontos de mudança em series temporais utilizando dados biomédicos advindos de relógios inteligentes

Abstract

This final paper addresses the topic of detecting change points in time series, using biomedical data from smartwatches. With the increasing availability of these devices and the wealth of information they capture, detecting changes in temporal patterns is the first challenge for interpreting and understanding data. The study proposes an analysis of this domain, integrating theoretical concepts about time series, segmentation and specific change point detection algorithms. Furthermore, the singularities and complexity inherent to biomedical data collected by smart watches are explored, aiming to contribute to the advancement of research in this field. The methodology includes the implementation and evaluation of four change point detection algorithms, considering three different types of simulations in order to identify the effectiveness of these techniques in detecting changes in biomedical time series. The results obtained are discussed in terms of performance and practical applicability evaluated on a standard metric. This final paper aims not only to expand the understanding of the detection of change points in biomedical time series, but also to provide practical and applicable contributions through the development of a tool for annotating time series that suggests change points for a given time series.