Fast nonparametric algorithm for change point detection in stochastic process

Abstract

The problem of constructing algorithms for change point detecting in a ran- dom process is considered. We investigate processes with change of spectral density or correlation function of the process. The algorithms use emission sta- tistics of random processes: the average number of positive and negative emis- sions above a given level and rapid procedure sliding nonparametric hypothesis testing with the estimation of the moments of change points appearance.