A novel approach for linear interval regression models with fuzzy weights

摘要

The problem of the traditional linear interval regression models approach for interval-valued data is the incapability of interpreting local behavior of the estimated models. A novel approach, called the linear interval regression models with fuzzy weights, is proposed to deal with this problem. Although the proposed approach can indeed model local behavior of models better than the traditional approach, the proposed approach still has the problem of boundary effects, which may generate a large bias at the boundary and also need more time to calculate. In the proposed approach, the original interval-valued data set is separated into some of interval-valued data subset by the interval fuzzy c-means (IFCM) clustering algorithm. Then, the local interval regression models (LIRMs) are independently constructed by the traditional linear approaches for each interval-valued data subset. Finally, those LIRMs are combined by a fuzzy weighted mechanism to form the estimated output. Experimental results show that the proposed approach needs less computational time than the traditional approach and can also have better results than the traditional approach does.