Title: Change point detection by distance covariance function
Authors: Konstantinos Fokianos - University of Cyprus (Cyprus) [presenting]
Abstract: The focus is on the problem of non-parametric change-point detection for multivariate time series data in the sense of discovering changes in the marginal distribution of the process. We employ a methodology that is based on properties of the distance covariance function. The main ideal is to investigate the behavior of a cumulative sum type of process in terms of the characteristic function. In fact, the test statistic we consider has a closed form expression and can be easily implemented to any given data stream. Some preliminary results show the validity of this approach.