A learning methodology in uncertain and imprecise environments

A step by step methodology for learning fuzzy rules is presented. This methodology tries to be general enough to give a framework within which diierent learning methods in an environment of uncertainty and imprecision could be developed. The nal product will always be an uncertainty distribution on the diierent rules representing the behaviour of the system. The particular uncertainty distribution depends on the concrete functions selected in each step of the process. The learning approach can also be considered as a way to construct uncertainty measures from data.