An anti-accuracy rule rooting in information diffusion techniques

Reviewing the information diffusion techniques, in this paper we give an anti-accuracy rule: a model for estimating a function would be more rough if the data used by the model is incomplete. This paper shows some anti-accuracy phenomena. The rule is from the method of information distribution and the model of normal diffusion, where the roughness of a model is defined by using step length and diffusion coefficient, the incompleteness of a given sample is defined by using the size of the sample. It implies that a fuzzy model can be used and only can be used to deal with data that has imprecise or incomplete information. We can ensure our fuzzy model if and only if it is employed to deal with fuzzy information.