回归模型中变点识别的一种稳健方法

变点识别是数据分析中一个非常重要的研究内容.文中针对目前变点识别研究中忽略了方法的稳健性,未能充分考虑异常值的影响的不足,提出利用一种高度稳健的回归类混合分解算法来识别变点.该方法从混合回归模型的角度,将含有变点的回归模型看作回归类的混合,通过逐步挖掘数据集中的回归类,并对排序后的回归类进行分析,进而确定变点的位置及个数.数值模拟表明,在识别变点的过程中无须预先指定变点的数目,并且具有高度的稳健性和有效性.

A Robust Method for Change-Points Detection in Regression Model

Change-points detection is one of the most important problems in data analysis. Traditional investi-gations 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 - point in a large data set. In the method, based on the mixture of regression model, the regression model with change - point is regards as the mixture of many regression - classes. Different regression classes are mined sequentially in the data set, and then the number and the position of change - points can be determined by analysing the regression classes having been arranged. The experiments demonstrate that the number of change points which can be gotten automatically, will not be prespecified, and this method is very robust and effective in change point detection.