Change-Points Detection in Fuzzy Point Data Sets

Change-points detection is one of important problems in data analysis. Traditional change-points detection method is based on exact data sets which can't reflect prior information of data. In this paper, a new concept, called "fuzzy point data" which is defined by giving a fuzzy membership to the data in exact data sets, is proposed for helping us handle the confidence of data. We introduce regression-classes mixture decomposition method for Change-points detection in fuzzy point data sets. In the method, different regression classes are mined sequentially in fuzzy point data sets and the estimation of change-points are determined by the two joined regression-classes, the number of the change-points will not be pre-specified. Numerical experiments show that by using fuzzy data point data, important data can make much contribution to mining regression classes. This shows that the change-points we got in fuzzy data point sets are more meaningful than we got in exact data sets.