A Robust Method for Detecting Regression Change Points

Change-points detection is one of the important problems in data analysis. Traditional investigations on the detection of change-points considering little infection of noise always ignore the robust of the methods. In this paper, a highly robust regression-class mixture decomposition method is proposed for finding change-points in a large data set. By using this method, the problem of detecting change-point can be converted to determine the breakpoint of different regression classes. We can mine all of the regression classes first, and then determine the estimation of change-points by anglicizing the two joined regression-classes. So the change-points can be found with little prior information. The analysis of experiments shows that our method can detect change-points in a data set with a large proportion of noisy, which demonstrate that this method is very robust and effective in change points detection.